Guideline for Practical Use of Methods for Testing the ...cristina/RREst/Aulas_Apresentacoes/07... · Guidelines for practical use of methods for testing the resistance of concrete

Guideline for Practical Use of Methods for Testing the Resistance of Concrete

to Chloride Ingress

Deliverable D23 CONTRACT N°: G6RD-CT-2002-00855

PROJECT N°: GRD1-2002-71808

ACRONYM: CHLORTEST

DURATION: January 2003 – December 2005

CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress – From laboratory tests to in-field performance

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

PROJECT COORDINATOR: SP Swedish National Testing and Research Institute (SP) S PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E

University of Alicante (UoA) E

Chalmers University of Technology (Chalmers) S

Skanska Norge AS (Selmer) NO

Swedish National Road Administration (SNRA) S

Electricité de France (EDF) F

Netherlands Organisation for Applied Scientific Research (TNO) NL

Hochschule Bremen (HSB) D

Slovenian National Building and Civil Engineering Institute (ZAG) SI

Queens University Belfast (QUB) UK

Laboratório Nacional de Engenharia Civil (LNEC) P

Icelandic Building Research Institute (IBRI) IS

National Institute of Applied Science (INSA) F

Laboratoire Central des Ponts et Chaussées (LCPC) F

Valenciana de Cementos, S.A. CEMEX (VCLC) E

Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT: The present document is a deliverable of Workpackage 6 – “Conclusions”. All the consortium members were involved in the work of this part of the project.

This document was prepared by Tang Luping (SP)

FURTHER INFORMATION: Regarding this document: Dr Tang Luping SP Swedish National Testing and Research Institute Box 857 S-501 15 BORÅS, Sweden Tel. +46-33 165138; Fax: +46-33 134516 e-mail: [email protected]

Regarding the project in general Dr Tang Luping SP Swedish National Testing and Research Institute Box 857 S-501 15 BORÅS, Sweden Tel. +46-33 165138; Fax: +46-33 134516 e-mail: [email protected]


TABLE OF CONTENTS

Page

1 BACKGROUND 5

2 BRIEF DESCRIPTION OF TEST METHODS 6 2.1 European method EN 13396 6 2.2 Nordtest method NT BUILD 443 (Immersion test) 6 2.3 Nordtest method NT BUILD 492 (Rapid migration test) 7 2.4 INSA steady state migration test 7 2.5 Multi-regime migration test 8 2.6 Resistivity test 9

3 PROPOSED METHODS FOR STANDARDISATION 9

4 INTERPRETATION OF TEST RESULTS 9 4.1 Results from Immersion test 9 4.2 Results from Rapid migration test 10 4.3 Results from Resistivity test 11

5 ACCEPTANCE CRITERIA 13

REFERENCES 13



1 BACKGROUND Concrete exposed to chloride environment such as seawater or roads where de-icing salts are used in very cold weathers may have durability problems because of the chloride-induced reinforcement corrosion. The premature deterioration of concrete structures is increasingly demanding methods for better prediction of the distresses and for evaluation of the suitability of concrete mixes to the desired service life. In order to recommend reliable methods for testing the resistance of concrete to chloride ingress, the European Commission funded the research project CHLORTEST, in which 17 partners from 10 European countries participated. Chloride ingress into concrete involves complex physical and chemical processes. The complexity comes at least from three sources:

a) The external environment is not constant. In marine environments the amount of chlorides in contact with concrete depends on whether the structure is placed fully submerged or in the tidal zone, or only in contact with marine fog, while in road environments, the intermittent use of de-icing salts in very cold weathers will make difficulty in calculating the amount of chlorides sprayed to the structures;

b) The material concrete is constituted by different types of cement and binder, with different mix proportions, which makes the concrete to be not a single material but many different ones which in addition evolve in properties with age; and

c) The mechanisms of chloride penetration are not single (simple diffusion) but combined with convection (absorption), chemical and physical binding, interaction of other coexisting ions, etc. Changes in temperature, rain and sunshine introduce variations that should also be taken into account.

Owing to its important role with regard to durability of concrete structures, many methods have been proposed for testing chloride ingress in concrete, although the above complexity has up to now hindered to reach a general agreement on a single test method. A collection of more than ten different test methods is available through an international committee, RILEM TC-178 TMC [1]. These methods can be categorised into three categories: diffusion tests, migration tests, and indirect tests based on resistivity or conductivity. Among those many methods, six of them were evaluated in this project, that is, European method EN 13396 (immersion test for repair products and systems), Nordtest methods NT BUILD 433 (Immersion test) and NT BUILD 492 (Rapid migration test), INSA steady-state migration test [2], Multi-regime migration test [3], and Resistivity test [4]. After the pre-evaluation [5], four methods, that is, NT BUILD 433, NT BUILD 492, Multi-regime migration test, and Resistivity test, were selected for inter-comparison evaluation, in which 15 laboratories participated in order to produce reliable precision data [6]. Based on the evaluation results, three methods, that is, NT BUILD 433, NT BUILD 492 and Resistivity test, are recommended for further European standardisation.


2 BRIEF DESCRIPTION OF TEST METHODS 2.1 European method EN 13396 This method was proposed by CEN/TC 104/SC 8, with the scope of testing the resistance to chloride penetration in repair products and systems for the protection and repair of concrete. This method gives chloride contents at different depths after different exposure durations, but does not provide any transport parameter. The method is also time-consuming and takes about half a year for the full test. At the start of CHLORTEST project, the available version of this method is prEN 13396:2002, while the latest formal version is EN 13396:2004. The main difference between prEN 13396:2002 and EN 13396:2004 is that the former specifies the chloride immersion at 40 °C, while the latter at 23 °C. In this project, the former version was used in the evaluation. Specimens: 6 specimens of diameter ≥ 100 mm and length ≥ 60 mm, with the trowelled surface as the test surface Pre-conditioning: Vacuum saturation with demineralised water Testing: Immersing specimens in the 3% NaCl solution and measuring chloride contents in two specimens at the depths 0∼2, 4∼6 and 8∼10 mm, after 28 days, 3 months and 6 months Test duration: about 6 months Test results: Chloride contents at three depths and three exposure durations Precision: Average repeatability COV (Coefficient of Variation) 14% and reproducibility COV 36% according to the pre-evaluation results [5] 2.2 Nordtest method NT BUILD 443 (Immersion test) This method is based on natural diffusion under a very high concentration gradient. The test gives values of Dnssd (non-steady state diffusion coefficient) and Cs (surface total chloride content) by curve-fitting the measured chloride profile to an error-function solution of Fick’s 2nd law. From the values of Dns and Cs, the parameter KCr, called penetration parameter, can be derived. The test is relatively laborious and takes relatively long time (more than 35 days). Specimens: 3 specimens of diameter ≥ 75 mm and length ≥ 60 mm, with the cut surface as the test surface and epoxy coating on all non-exposure surfaces Pre-conditioning: Natural immersion in the saturated lime-water until a constant weight is reached Testing: Immersing specimens in the solution of 165 g NaCl per litre for at least 35 days, and measuring chloride penetration profiles by grinding the specimen successively from the exposed surface and titration-analysing total chloride content in each powder

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress Test duration: at least 35 days Test results: Chloride penetration profiles and curve-fitted parameters Dnssd and Cs, as well as the derived parameter KCr Precision: Average repeatability COV 20%, 18% and 9% for parameters Dnssd, Cs and KCr, respectively, and reproducibility COV 28%, 22% and 14% for parameters Dnssd, Cs and KCr, respectively, according to the final evaluation results [6], noting that KCr is proportional to the square root of Dnssd, implying that the deviation of KCr is theoretically a half of that of Dnssd 2.3 Nordtest method NT BUILD 492 (Rapid migration test) This method is a non-steady state migration test using an external electrical field for accelerating chloride penetration. The test gives values of Dnssm (non-steady state migration coefficient). The test is relatively simple and rapid with the test duration in most cases 24 hours. Specimens: 3 specimens of diameter 100 mm and thickness 50 mm, with the cut surface as the test surface Pre-conditioning: Vacuum saturation with saturated lime-water Testing: Imposing a 10∼60 V external potential across the specimen with the test surface exposing in the 10% NaCl solution and the oppose surface in the 0.3 M NaOH solution for a certain duration (in most cases 24 hours), then splitting the specimen and measuring the penetration depth of chlorides by using a colourimetric method Test duration: 6∼96 (in most cases 24) hours depending on the quality of concrete Test results: Non-steady state migration coefficient Dnssm calculated from the average penetration depth Precision: Average repeatability COV 15% and reproducibility COV 24% for parameter Dnssm, according to the final evaluation results [6] 2.4 INSA steady state migration test In this steady state migration test the chloride flux is determined by measuring the concentration changes in the upstream chloride solution. The test gives values of Ds (steady state migration coefficient). The test is relatively laborious due to many samples for chloride analysis. The reproducibility of this test method seems not satisfactory. Specimens: 3 specimens of diameter 100 mm and thickness 20 mm, with the cut surface as the test surface Pre-conditioning: Vacuum saturation with demineralised water

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress Testing: Imposing a 12 V external potential across the specimen with the test surface exposing in the 1 M NaCl solution (upstream cell) and the oppose surface in the demineralised water (downstream cell) and measuring the chloride concentration in the upstream cell at a certain interval until a constant decrease in chloride concentration can be obtained from the concentration-time curve Test duration: a few days up to about two weeks depending on the quality of concrete Test results: Steady state migration coefficient Ds calculated from the constant flux of chloride ions Precision: Average repeatability COV 19% and reproducibility COV 87% for parameter Ds, according to the pre-evaluation results [5] 2.5 Multi-regime migration test In this test the chloride flux is determined by measuring the conductivity changes in the downstream solution. The test gives values of Dssm from the flux and Dnssm from the time-lag. The test is relatively simple due to the indirect measurement of chloride concentration through the simple conductivity measurement. The reproducibility of this test method seems not satisfactory. Specimens: 3 specimens of diameter 100 mm and thickness 20 mm, with the cut surface as the test surface Pre-conditioning: Vacuum saturation with demineralised water Testing: Imposing a 12 V external potential across the specimen with the test surface exposing in the 1 M NaCl solution (upstream cell) and the oppose surface in the demineralised water (downstream cell) and measuring the conductivity, which can be converted to chloride concentration, in the downstream solution at a certain interval until a constant increase in conductivity can be obtained from the concentration-time curve Test duration: a few days up to about two weeks depending on the quality of concrete Test results: Steady state migration coefficient Ds calculated from the slope of the constant portion of the concentration-time curve (the constant flux) and non-state migration coefficient Dns calculated from the intersection on the time-axis of the constant portion of the concentration-time curve (time-lag) Precision: Average repeatability and reproducibility COV 22% and 76%, respectively, for parameter Ds according to the final evaluation results [6], and average repeatability and reproducibility COV 24% and 45%, respectively, for parameter Dns according to the pre-evaluation results [5].


2.6 Resistivity test The resistivity test is an indirect measurement of the transport property of concrete, because the electrical resistance of concrete is related to the pore structures and ionic strength in the pore solution. Specimens: 3 specimens of diameter 100 mm and thickness 50 mm, with the cut surface as the test surface Pre-conditioning: Vacuum saturation with distilled or demineralised water. Testing: Imposing a constant alternative current across the specimen and measuring the potential response for calculating the resistance using Ohm’s law Test duration: a few seconds or minutes for the measurements Test results: Resistivity ρ Precision: Average repeatability COV 11% and reproducibility COV 25% for resistivity according to the final evaluation results [6] 3 PROPOSED METHODS FOR STANDARDISATION Based on the evaluation results, the CHLORTEST consortium proposes the following three methods for further standardisation at the European level:

• Immersion test (based on NT BUILD 443) for determination of non-steady state diffusion coefficient, Dnssd, and surface total chloride content, Cs;

• Rapid migration test (based on NT BUILD 492) for determination of non-steady state migration coefficient, Dnssm, under the standardised laboratory exposure condition; and

• Resistivity test (based on the version used in [6]) for determination of resistivity ρ as an indirect measurement of the transport property of concrete

All the above three proposed methods have the precision in an acceptable range, that is, repeatability COV ≤20% (11%∼20%) and reproducibility COV ≤30% (24%∼28%). Therefore, they are suitable for data exchanges and industrial applications. The descriptions of the above proposed methods with certain revisions and modifications for transfer to standards are given in [7]. 4 INTERPRETATION OF TEST RESULTS 4.1 Results from Immersion test The immersion test provides coupled values of Dnssd and Cs by curve-fitting the measured chloride profile to an error-function solution of Fick’s 2nd law, which is under the assumption

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress of constant chloride binding capacity. In the reality, chloride binding capacity is non-linearly dependent on free chloride concentration and also dependent on type of cementitious binder [8-10]. The total chloride content is a sum of the free chlorides in the pore solution and the bound chlorides on the surfaces of hydrates. Therefore, the Cs value is dependent on the porosity and the type of binder. Even under the same exposure condition, that is, in the solution of the same chloride concentration, different types of binder will give different values of Cs.

Since Dnssd is coupled with Cs in the curve-fitting, the value of Dnssd alone may not reflect the actual resistance of concrete to chloride ingress. To properly interpret the test results, both values of Dnssd and Cs should be taken into account. The penetration parameter, KCr, combines influences of Dnssd and Cs and, therefore, better facilitates comparison of the results. An example is shown in Figure 1, where Mix A reveals a lower value of Dnssd and a higher value of Cs than Mix B, but both mixes have the same value of KCr.

It should be noted that the parameter KCr with a dimension of mm/√yr is mainly for facilitating comparison, but not necessarily means the actual penetration depth per square root of year.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 2 4 6 8 10

x, mm

Cl,

mas

s% o

f con

cret

e

Mix A Mix B

D nssd = 2 x10-12 m2/sC s = 1.2% of concreteK Cr = 23 mm/√yr

<>=

D nssd = 2.5 x10-12 m2/sC s = 0.7% of concreteK Cr = 23 mm/√yr

Figure 1 – Example of coupled values of Dnssd and Cs. 4.2 Results from Rapid migration test The rapid migration test provides value of Dnssm, which is also under the assumption of constant chloride binding capacity during the test. Different from the immersion test, this assumption may better hold in the rapid migration test, owing to the strong external electrical field and short testing duration, both of which tend to reduce the amount of bound, especially

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress physically bound, chlorides. Therefore, the parameter Dnssm describes the property of chloride transport under a condition of reduced chloride binding [11]. Since Dnssm and Dnssd are from completely different testing conditions, their values may not be necessarily comparable. However, experimental results from the CHLORTEST project [5,6] and some other previous projects [12,13] show that these two diffusion coefficients are coincidentally quite comparable, as shown in Figure 2. Considering the measurement uncertainties of the test methods, it is reasonable to conclude that both the test methods measure the similar transport parameters.

0

10

20

30

40

0 10 20 30 40

D nssd, x10-12 m2/s

Dns

sm, x

10-1

2 m

2 /s

Ref [5]

Ref [6]

Ref [12]

Ref [13]

x ± dx*

y ± dy*

* dx, dy: reproducibility standard deviation of D nssd and D nssm, respectively.

Note: The values of Dnssd were from the exposure for a target of 35 days.

Figure 2 – Relationship between chloride transport parameters Dnssm and Dnssd.

4.3 Results from Resistivity test Theoretically, resistivity is inversely proportional to diffusivity. Practically, however, the measured resistivity is contributed by all ions, especially hydroxides, in the pore solution. A concrete with the binder containing high alkali will resulted in a low resistivity, while a concrete with pozzolanic additions containing low alkali will often resulted in a high resistivity. To convert resistivity to chloride diffusion coefficient, the chloride transference number (the ratio of chloride ions’ intensity to the intensity of total ions) in the pore solution should be known. It is not an easy task to estimate the intensity of total ions in the pore solution. Therefore, the resistivity test can only be taken as an indirect measurement of chloride transport property. Owing to its rapidity and simplicity, this test is a very efficient tool for quality control in the real production of concrete. Calibration is needed in order to

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress establish the empiric relationships between resistivity and chloride diffusion coefficient, as exampled in Figure 3.

Figure 3 – Example of empiric relationships between resistivity and diffusivity.

0

1

2

3

4

5

0 5 10 15 20 25 30

1/ρ, mS

Ds,

x10-1

2 m2 /s

Data from [5]

Data from [6]

Acc. to [14]

k = 0.213

k = 0.152

D = k /ρ

k = 0.12

0

10

20

30

40

50

60

0 5 10 15 20 25 30

1/ρ, mS

Dns

sm, x

10-1

2 m2 /s

Data from [5]

Data from [6]

Data from [12]

Data from [15]

Data from [16]

k = 0.86

k = 0.98

D = k /ρ k = 3.04

Data excludedfrom regression


5 ACCEPTANCE CRITERIA In normal cases chloride itself does not directly result in any damage of concrete, but induce corrosion of steel in concrete. The service-life of reinforced concrete structures exposed to chloride environments includes the periods of corrosion initiation and propagation. The former is related to chloride ingress, while the latter is related to corrosion rate. The period of corrosion initiation is a function of chloride transport property, concrete cover, threshold chloride level, exposure environment, etc. Obviously, chloride transport property is only one of the several parameters regarding initiation of corrosion. A concrete with a relatively high diffusivity can be compensated with a thicker cover to reach the desirable resistance to corrosion initiation. Therefore, acceptance criteria for the values from the proposed tests are dependent on many factors, and can be expressed by the following function:

( )envCrLmin ,,, KCtxfD = where D is the value from a proposed test, f denotes a function, xmin is the minimum thickness of concrete cover, tL is the desired period of corrosion initiation, CCr is the threshold chloride level, Kenv is the environmental factor including chloride load (e.g. surface concentration or content) and micro climate (temperature, humidity, precipitation, etc.). To solve the above function, proper models are needed. Different prediction models for chloride ingress have been evaluated in the CHLORTEST project with the infield data collected from short time (0.5 year) up to 42 years’ exposures [17]. The results show that one model, Model 5 (ClinConc [18,19]), which was previously calibrated with the 10 years’ traceable data from a field exposure site, reveal reasonably good benchmarks [17]. This indicates the importance of calibration with the reliable long-term data. When compared with over 100 years’ service life, however, these 10 years’ traceable data are still not enough to assure the verification of models for actual service life prediction. Therefore, it is always the user’s responsibility to use the D values obtained from the proposed tests for modelling of service life of a particular concrete structure. REFERENCES [1] RILEM TC-178 TMC: “Testing and Modelling Chloride Ingress into Concrete”,

Internal document to be published in the near future. [2] Truc, O., Ollivier, J.P.and Carcassès, M, “A new way for determining the chloride

diffusion coefficient in concrete from steady state migration test”, Cem. Concr. Res., 30(2) 217-226 (2000).

[3] Castellote M., Andrade C.and Alonso C., “Measurement of the steady and nonsteady state chloride diffusion coefficients in a migration test by means of monitoring the conductivity in the anolyte chamber - Comparison with natural diffusion tests”, Cem. Concr. Res., 31(10) 1411-1420 (2001).

[4] Andrade, C., “Determination of electrical resistivity in concrete specimens – Direct method”, A translation of UNE 83XXX, 2004.

[5] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855, WP2 Report: “Pre-evaluation of different test methods”, 2005.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 14 Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress [6] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-

field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855, WP5 Report: “Final evaluation of test methods”, 2005.

[7] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855, Deliverable 22: “Testing Resistance of Concrete to Chloride Ingress – A proposal to CEN for consideration as EN standard”, 2005.

[8] Tritthart, J. “Chloride binding in cement: II. The influence of the hydroxide concentration in the pore solution of hardened cement paste on chloride binding”, Cem. Concr. Res., 19(5) 683-691 (1989).

[9] Byfors K., “Chloride-initiated Reinforcement Corrosion - Chloride binding”, Swedish Cement and Concrete Research Institute (CBI), Stockholm, CBI Report 1:90, 1990.

[10] Tang, L. and Nilsson, L-O. “Chloride binding capacity and binding isotherms of OPC pastes and mortars”, Cem. Concr. Res., 23(2) 347-353 (1993).

[11] Tang, L., “Chloride Transport in Concrete - Measurement and prediction”, Doctoral thesis, Publication P-96:6, Dept. of Building Materials, Chalmers Universities of Technology, Gothenburg, Sweden, 1996.

[12] Frederiksen, J.M., Sørensen, H.E., Andersen, A. & Klinghoffer, O., ‘HETEK, The effect of the w/c ratio on chloride transport into concrete - Immersion, migration and resistivity tests’, HETEK Report No. 54, ed. by J.M. Frederiksen, published by the Danish Road Directorate, 1997.

[13] Tang L. and Sørensen, H.E., “Precision of the Nordic Test Methods for Measuring the Chloride Diffusion/Migration Coefficients of Concrete”, Materials and Structures, 34 479-485 (2001).

[14] Andrade, C., Alonso, C., Arteaga, A. & Tanner, P., “Methodology based on the electrical resistivity for the calculation of reinforcement service life”. Supplementary papers of the proceedings of the Fifth International CANMET/ACI Conference on Durability of concrete. Barcelona, Spain, 4-9 June 2000, pp 899-915.

[15] DURACRETE, “Probabilistic Performance Based Durability Design of Concrete Structures”, EU Brite-EuRam III project DuraCrete (BE95-1347), Deliverable R8, “Compliance testing for probabilistic design purposes”, 1999.

[16] Romer, M., “TC 189-NEC Comparative Test - Part I - Comparative test of 'penetrability' methods”, Materials and Structures, 38, 895-905 (2005).

[17] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855, Workpackage 4 Report: “Modelling of chloride ingress”, 2005.

[18] Tang, L. “Engineering Expression of the ClinConc model for prediction of free and total chloride ingress in submerged marine concrete”, submitted to Cem. Concr. Res., 2005.

[19] Tang, L., “Service-life prediction based on the rapid migration test and the ClinConc model”, Proceedings of International RILEM Workshop on Performance Based Evaluation and Indicators for Concrete Durability, 19-21 March 2006, Madrid, Spain.

Testing Resistance of Concrete to Chloride Ingress – A proposal to CEN for consideration as EN standard

Deliverable D22 CONTRACT N°: G6RD-CT-2002-00855


ACRONYM: CHLORTEST







Selmer Skanska AS (Selmer) NO













ACKNOWLEDGEMENT: The present document is a deliverable of Workpackage 6 – “Conclusions”. All the consortium members were involved in the work of this part of the project. Comments and suggestions from Jens M. Frederiksen at Birch & Krogboe A/S, Denmark, and Joost Gulikers at RWS, The Netherlands, are specially appreciated.




A proposal to CEN for consideration as EN standard

3

Contents

Foreword......................................................................................................................................................................5 1 Scope ..............................................................................................................................................................5 2 Normative references....................................................................................................................................5 3 Terms and definitions ...................................................................................................................................5 4 Test specimens in general............................................................................................................................6 5 Immersion test ...............................................................................................................................................6 5.1 Principle..........................................................................................................................................................6 5.2 Reagents and equipment..............................................................................................................................7 5.3 Preparation of the test specimen.................................................................................................................7 5.4 Test procedures.............................................................................................................................................8 5.5 Expression of results ....................................................................................................................................9 5.6 Test report ....................................................................................................................................................10 6 Rapid migration test ....................................................................................................................................11 6.1 Principle........................................................................................................................................................11 6.2 Reagents and equipment............................................................................................................................11 6.3 Preparation of the test specimen...............................................................................................................13 6.4 Test procedures...........................................................................................................................................14 6.5 Expression of results ..................................................................................................................................16 6.6 Test report ....................................................................................................................................................17 7 Resistivity test .............................................................................................................................................18 7.1 Principle........................................................................................................................................................18 7.2 Equipment ....................................................................................................................................................18 7.3 Preparation of the test specimen...............................................................................................................18 7.4 Test procedures...........................................................................................................................................19 7.5 Expression of results ..................................................................................................................................19 7.6 Test report ....................................................................................................................................................20 8 Precision data ..............................................................................................................................................20 9 Annex............................................................................................................................................................21 Figure 1 — Example of regression analysis for curve-fitting ......................................................................................10

Figure 2 — One arrangement of the rapid migration set-up.......................................................................................12

Figure 3 — Stainless steel clip ...................................................................................................................................12

Figure 4 — Plastic support and cathode ....................................................................................................................13

Figure 5 — Rubber tube assembled with specimen, clips and anode .......................................................................13

Figure 6 — Illustration of measurement for chloride penetration depths ...................................................................16

Figure 7 — Arrangement for the measurement of resistivity by the direct method....................................................19

Tables

Table 1 — Recommended depth intervals (in mm) for powder grinding......................................................................9

Table 2 — Test voltage and duration for concrete specimen with normal binder content. ........................................15

Table 3 — Precision data of various test methods.....................................................................................................20

EU-Project CHLORTEST G6RD-CT-2002-00855 Testing resistance of concrete to chloride ingress

4


5

Foreword

This is a proposal CEN for consideration as EN standard.

This document is based on the results from the EU-project “Chlortest” under the 5th Frame Programme (GRD1-2002-71808/G6RD-CT-2002-00855).

Introduction

Chlorides can induce reinforcement corrosion. Reinforced concrete structures exposed to environments containing chlorides need to be durable and to have an adequate resistance to the ingress of chlorides.

Owing to its important role with regard to durability of concrete structures, many different test methods have been developed. Some of the commonly used test methods were evaluated in the recently closed EU-project “Chlortest” under the 5th Frame Programme (GRD1-2002-71808/G6RD-CT-2002-00855). Based on the evaluation results from this EU-project, the ChlorTest consortium recommends three test methods for testing resistance of concrete to chloride ingress, that is, immersion test, rapid migration test and resistivity test.

It is desirable, especially in the case of new constituents or new concrete compositions, to test the resistance of concrete to chloride ingress. This also applies to concrete mixes, concrete products, precast concrete, concrete members or concrete in situ.

1 Scope

This European prestandard describes the testing for the resistance of concrete to chloride ingress. It can be used either to supply quantitative data to durability design or to assess the quality of concrete during the production or construction.

2 Normative references

This proposal incorporates by dated or undated reference, provisions from other publications. These normative references are cited at the appropriate places in the text and the publications are listed hereafter. For dated references, subsequent amendments to or revisions of any of these publications apply to this proposal only when incorporated in it by amendment or revision. For undated references the latest edition of the publication referred to applies.

EN 12390-2, Testing hardened concrete — Part 2: Making and curing specimens for strength tests.

prEN 14629, Production and systems for the protection and repair of concrete structures – Test methods – Determination of chloride content in hardened concrete.

ISO 5725, Accuracy (trueness and precision) of measurement methods and results.

3 Terms and definitions

For the purposes of this proposal, the following terms and definitions apply.

3.1 As-cast surface The surface of a concrete structure exposed to the chloride environment


6

3.2 chloride penetration The ingress of chlorides into concrete due to exposure from external chloride sources 3.3 diffusion The movement of molecules or ions under a concentration gradient or, more strictly, chemical potential, from a high concentration zone to a low concentration zone. 3.4 maturity-day A concrete of 28 maturity-days has developed a maturity corresponding to curing in 28 days at 20 °C.

3.5 migration The movement of ions under the action of an external electrical field

3.6 profile grinding Grinding off concrete powder in thin successive layers from a test specimen using a dry process

3.7 resistivity The electrical resistance per unit length and per unit reciprocal cross-sectional area of concrete at a specified temperature

3.8 surface-dry condition A surface condition achieved by drying the water-saturated test specimen with a clean cloth or similar leaving the test specimen damp but not wet.

4 Test specimens in general

Drilled cores or cast cylinders can be used as test specimens. They must be representative of the concrete and/or structure in question. The diameter of test specimens is in normal cases 100 mm, or at least 4 times as large as the maximum size of aggregates. At least three test specimens should be used in the test. If cast cylinders or cores drilled from cast cube are used as specimens, the casting and curing procedures should be in accordance with EN 12390-2.

In normal cases the concrete should be hardened to at least 28 maturity-days for testing. Since the concrete age has significant effect on chloride transport, the date of manufacture of concrete and the date of testing shall always be noted in the report. If the concrete temperature during hardening was outside the range of 10~30 ºC, this must also be noted in the report.

5 Immersion test

5.1 Principle

A water-saturated concrete specimen is on one plane surface exposed to water containing sodium chloride. After a specified exposure time thin layers are ground off parallel to the exposed face of the specimen and the chloride content profile is measured. The apparent chloride diffusion coefficient and the chloride content at the exposed surface are calculated by curve-fitting the measured profile to the error function solution to Fick’s 2nd law. A penetration parameter combining the influence of diffusion coefficient, surface chloride content, initial chloride content and a reference chloride content is calculated for facilitating comparison of the test results from different types of concrete.


7

5.2 Reagents and equipment

5.2.1 Reagents

5.2.1.1 Distilled or demineralised water.

5.2.1.2 Calcium hydroxide: Ca(OH)2, technical quality.

5.2.1.3 Sodium chloride: NaCl, chemical quality.

5.2.1.4 2-component solvent free (chloride-ion and water vapour diffusion-proof) polyurethane or epoxy-based paint (membrane).

5.2.1.5 Chemicals for chloride analysis as required by the test method employed (see 5.4.6).

5.2.2 Equipment

5.2.2.1 Equipment for cutting specimens such as a water-cooled diamond saw.

5.2.2.2 Balance: with accuracy better than ±0.01g.

5.2.2.3 Thermometer or thermocouple: with readout device capable of reading to ±1 ºC.

5.2.2.4 Plastic container with tight-fitting lid.

5.2.2.5 Slide calliper with a precision of ±0.1 mm.

5.2.2.6 Ruler with a minimum scale of 1 mm.

5.2.2.7 Equipment for grinding off and collecting concrete powder from thin concrete layers (less than 2mm).

5.2.2.8 Standard sieve, mesh width 1.0 mm.

5.2.2.9 Equipment for chloride analysis as required by the test method employed (see 5.4.6).

5.3 Preparation of the test specimen

5.3.1 Test specimen

5.3.1.1 If drilled cores are used, cut about 70 mm thick slice from the outmost portion of each core as the test specimen. The surface 10 mm below the as-cast surface is the one to be used for exposure (see 5.3.2.4).

5.3.1.2 If cast cylinders are used, cut an about 70 mm thick slice from the central portion of each cylinder as the test specimen. The end surface that was nearer to the as-cast surface is the one to be used for exposure (see 5.3.2.4).

5.3.1.3 Cut a slice with a thickness of at least 20 mm from the remainder of the drilled core or cast cylinder for the measurement of initial chloride content (see 5.4.6.2).

NOTE 1 The term 'cut' here means to saw perpendicularly to the axis of a core or cylinder, using a water-cooled diamond saw.

NOTE 2 It is very important that the test is made on the concrete between the surface and the layer of reinforcement because it is here the protection against chloride penetration is needed. Furthermore, the quality of the concrete in this particular area can deviate from the rest of the concrete. The outermost approximately 10 mm of concrete shall be removed (see 5.3.2.4) to ensure that the measurement is made in an area with relatively constant cement matrix content and unaffected by surface treatments, lack of curing, etc.

5.3.2 Preconditioning and coating

5.3.2.1 The test specimen is immersed in a saturated Ca(OH)2 solution at 20~25 °C in a tightly closed plastic container. The container must be filled to the top to minimize carbonation of the liquid. The next day the mass in surface-dry condition is determined by weighing the test specimen.


8

5.3.2.2 The storage in the saturated Ca(OH)2 solution continues until the mass of specimen under the surface-dry condition does not change by 0.1 mass% when measured at an interval of at least 24 hours.

5.3.2.3 Dry all surfaces of the test specimen except the one to be used for exposure at room temperature to a stable white-dry condition and given a thick epoxy or polyurethane coating on all white-dry surfaces. It must be ensured that the method of application and hardening prescribed by the supplier of the coating material is observed.

5.3.2.4 When the coating has hardened, cut off the outermost approximately 10 mm thick layer from the surface to be used for exposure and then immerse the test specimen in the Ca(OH)2 solution for overnight.

5.4 Test procedures

5.4.1 Exposure liquid

An aqueous NaCl solution is prepared with a concentration of 165g ± 1g NaCl per litre solution. This exposure liquid is used for 5 weeks and then replaced by a new pure NaCl solution. The NaCl concentration of the solution must be checked at least before and after use.

5.4.2 Temperature

The temperature of the water bath shall be kept between 20~25 °C, with a target average temperature of 23 °C during the immersion. The temperature shall be measured at least once a day.

5.4.3 Exposure

5.4.3.1 Place the surface-dry specimens in the container in such a way that the exposure surface is vertically exposed.

5.4.3.2 Fill the container with the exposure liquid. It is important that the container is completely filled with the exposure liquid and closed tightly. The ratio of the exposed area in cm2 to the volume of exposure liquid in litre shall be between 20~80.

5.4.3.3 Store the container in the room at the temperature as specified in 5.4.2.

5.4.3.4 The exposure shall last for at least 35 days, but in normal cases not exceed 40 days. The date and time of exposure start and termination is recorded to ±10 minutes.

5.4.4 Profile grinding

5.4.4.1 After the exposure, slightly rinse the specimen with tap water.

5.4.4.2 Wipe off excess water from the surfaces of the specimen.

5.4.4.3 Grind off the material in layers parallel to the exposed surface. The grinding area should be within a diameter approximately 10 mm less than the full diameter of the core from the exposure surface and can be gradually reduced in small steps depending on what is appropriate for the grinding tool as the depth increases, but should always be more than 3 times as large as the maximum size of aggregate. This obviates the risk of edge effects and disturbances from the coating.

5.4.4.4 At least eight layers must be ground off. The thickness of the layers should be adjusted according to Table 2 or the expected chloride profile so as to assure minimum 6 points in the profile having chloride content higher than the initial value (see 5.4.6.2). However, the outermost layer shall always have a thickness of minimum 1.0 mm.

5.4.4.5 It must be ensured that a sample of at least 5 g of dry concrete dust is obtained from each layer. For each sample of concrete dust collected, the depth below the exposed surface is calculated as the average of four uniformly distributed measurements using a slide caliper.


9

Table 1 — Recommended depth intervals (in mm) for powder grinding.

water/binder 0.25 0.30 0.35 0.40 0.50 0.60 0.70

Depth #1 0∼1 0∼1 0∼1 0∼1 0∼1 0∼1 0∼1

#2 1∼2 1∼2 1∼2 1∼3 1∼3 1∼3 1∼5

#3 2∼3 2∼3 2∼3 3∼5 3∼5 3∼6 5∼10

#4 3∼4 3∼4 3∼5 5∼7 5∼8 6∼10 10∼15

#5 4∼5 4∼6 5∼7 7∼10 8∼12 10∼15 15∼20

#6 5∼6 6∼8 7∼9 10∼13 12∼16 15∼20 20∼25

#7 6∼8 8∼10 9∼12 13∼16 16∼20 20∼25 25∼30

#8 8∼10 10∼12 12∼16 16∼20 20∼25 25∼30 30∼35 NOTE For concrete with pozzolanic additions such as fly ash, slag, and silica fume, the depth intervals in the column one place left should

be applied, e.g. for slag cement concrete with w/b = 0.4, the depth intervals in the column for w/b = 0.35 should be applied.

5.4.6 Measurement of chloride content

5.4.6.1 The acid-soluble chloride content of the samples is determined to three decimals in accordance with prEN 14629 or by a similar method with the same or better accuracy.

5.4.6.2 From the concrete slice (see 5.3.1.3), a representative sample of approximately 20 g is prepared by grinding or crushing until the material passes the 1 mm standard sieve. The initial chloride content is determined using the same method as used in 5.4.6.1.

5.5 Expression of results

The values of the apparent non-steady state chloride diffusion coefficient Dnssd and the apparent surface chloride content Cs are determined by curve-fitting the measured chloride profile to equation (1) according to the principle of least squares, as illustrated in figure 1:

⎟⎟

⎠

⎞

⎜⎜

⎝

⎛

tDx )C- C(- C = t)C(x, iss

4erf

nssd

(1)

where:

Dnssd: apparent non-steady state diffusion coefficient, m2/s;

x: average depth where the chloride sample was ground, m;

t: immersion duration, seconds;

C(x, t): chloride content at depth x, mass% of sample;

Cs: apparent surface chloride content, mass% of sample;

Ci: Initial chloride content, mass% of sample;

erf: error function as defined by equation (2).

( ) ∫ −

π

z u uz0

2de 2 = erf (2)

The penetration parameter, KCr, is calculated using equation (3), with the above obtained values of Ci, Cs, and Dnssd, and the assumption of Cr = 0.05 mass% of sample, unless another value is required.


10

⎟⎟⎠

⎞⎜⎜⎝

⎛

CCCC D2 =K nssd

is

rs1-Cr -

- erf (3)

where: erf-1 is the inverse of error function.

NOTE KCr is defined only when Cs > Cr > Ci.

Figure 1 — Example of regression analysis for curve-fitting

5.6 Test report

The test report shall contain at least the following information:

a) Reference to this standard method.

b) Origin, size and marking of the specimens.

c) Concrete identification.

d) Date of manufacture of concrete

e) Any deviation from the test method.

f) Test results, including the specimen dimensions, concrete age at exposure, exposure temperatures, chloride concentration of the exposure liquid, exposure duration, data of chloride content vs depth, the curve-fitted diffusion coefficient and surface content, and the calculated penetration parameter.

g) Any observation or suspicion of defects (cracks, large pore holes, foreign bodies etc.) in the specimen.


11

6 Rapid migration test

6.1 Principle

An external electrical field is applied axially across the specimen and forces the chloride ions outside to migrate into the specimen. After a certain test duration, the specimen is axially split or dry-cut and a silver nitrate solution is sprayed on to one of the freshly split or cut sections. The chloride penetration depth can then be measured from the visible white silver chloride precipitation, after which the chloride migration coefficient can be calculated from this penetration depth.

6.2 Reagents and equipment

6.2.1 Reagents

6.2.1.1 Distilled or demineralised water.

6.2.1.2 Sodium chloride: NaCl, chemical quality.

6.2.1.3 Sodium hydroxide: NaOH, chemical quality.

6.2.1.4 Silver nitrate: AgNO3, chemical quality.

6.2.1.5 Chemicals for chloride analysis as required by the test method employed (optional, see 6.4.6).

6.2.2 Equipment

6.2.2.1 Equipment for cutting specimens such as water-cooled saw.

6.2.2.2 Vacuum container: Capable of containing at least three specimens.

6.2.2.3 Vacuum pump: Capable of maintaining a pressure of less than 50 mbar (5 kPa) in the container, such as a water-jet pump.

6.2.2.4 Migration set-up: One design is shown in figure 2, which includes the following parts:

a) Silicone rubber tube: inner diameter suitable to the diameter of specimens, thickness 5∼7 mm, long about 150 mm.

b) Clip: stainless steel with 20 mm wideness and diameter range suitable to the rubber tube (figure 3).

c) Catholyte reservoir: plastic box, 370 × 270 × 280 mm (length × width × height).

d) Plastic support: (figure 4).

e) Cathode: stainless steel plate (figure 4), about 0.5 mm thick.

f) Anode: stainless steel mesh or plate with holes (figure 5), about 0.5 mm thick.

NOTE Other designs are acceptable, provided that temperatures of the specimen and solutions during the test can be maintained in the range of 20 to 25 °C (see 6.4.2).


12

Figure 2 — One arrangement of the rapid migration set-up

6.2.2.5 Power supply: capable of supplying 0~60 V DC regulated voltage with an accuracy of ±0.1 V.

6.2.2.6 Ammeter: capable of displaying current to ±1 mA.

6.2.2.7 Thermometer or thermocouple with readout device capable of reading to ±1 ºC.

6.2.2.8 Any suitable device for splitting (such as compression machine) or dry-cutting (such as air cooled diamond saw) the specimen.

6.2.2.9 Spray bottle.

6.2.2.10 Slide calliper with a precision of ±0.1 mm.

6.2.2.11 Ruler with a minimum scale of 1 mm.

6.2.2.12 Equipment for chloride analysis as required by the test method employed (optional, see 6.4.6).

Figure 3 — Stainless steel clip

+-

Potential (DC)

a. Rubber tube b. Anolyte c. Anode d. Specimen

e. Catholyte f. Cathode g. Plastic support h. Plastic box

a

b

c

d

e

f

g

h


13

Figure 4 — Plastic support and cathode

* The diameter of anode is 5 mm less than that of the specimen.

Figure 5 — Rubber tube assembled with specimen, clips and anode


6.3.1 Test specimen

6.3.1.1 If drilled cores are used, the outermost approximately 10 mm thick layer should be cut off and the next (50 ± 2) mm thick slice should be cut from each core as the test specimen. The end surface that was nearer to the outermost layer is the one to be exposed to the chloride solution (catholyte).

6.3.1.2 If cast cylinders are used, cut a (50 ± 2) mm thick slice from the central portion of each cylinder as the test specimen. The end surface that was nearer to the as-cast surface is the one to be exposed to the chloride solution (catholyte).

6.3.1.3 Measure the thickness with a slide calliper and read to 0.1 mm.

NOTE The term 'cut' here means to saw perpendicularly to the axis of a core or cylinder, using a water-cooled diamond saw.

Cathode

Plastic support

Plastic stud

15~20 mm

32°

A-A

Silicone rubber tube Anode* (stainless steel)

Stainless steel clips

Specimen

A-A


14

6.3.2 Preconditioning

After cutting, brush and wash away any burrs from the surfaces of the specimen, and wipe off excess water from the surfaces of the specimen. When the specimens are surface dry, place them in the vacuum container for vacuum treatment. Both end surfaces must be exposed. Reduce the absolute pressure in the vacuum container to a pressure in the range of 10~50 mbar (1~5 kPa) within a few minutes. Maintain the vacuum for three hours and then, with the vacuum pump still running, fill the container with the distilled or demineralised water so as to immerse all the specimens. Maintain the vacuum for a further hour before allowing air to re-enter the container. Keep the specimens in the solution for (18 ± 2) hours.

6.4 Test procedures

6.4.1 Catholyte and anolyte

The catholyte solution is 10% NaCl by mass in tap water (100 g NaCl in 900 g water, about 2 N) and the anolyte solution is 0.3 M NaOH in distilled or demineralised water (approximately 12 g NaOH in 1 litre water). Store the solutions at a temperature of 20~25 °C.

NOTE It is important to use distilled or demineralised water for the anolyte solution to obviate the corrosion damage of anode.

6.4.2 Temperature

Maintain the temperatures of the specimen and solutions in the range of 20~25 °C during the test.

6.4.3 Preparation of the test

6.4.3.1 Fill the catholyte reservoir with about 12 litres of 10 % NaCl solution.

6.4.3.2 Fit the rubber tube on the specimen as shown in figure 4 and secure it with two clips. If the curved surface of the specimen is not smooth, or there are defects on the curved surface which could result in significant leakage, apply a line of silicone sealant to improve the tightness.

6.4.3.3 Place the specimen on the plastic support in the catholyte reservoir (see figure 2).

6.4.3.4 Fill the tube above the specimen with about 300 ml anolyte solution (0.3 M NaOH) in order to cover the specimen surface with at least 3 mm liquid layer.

6.4.3.5 Immerse the anode in the anolyte solution.

6.4.3.6 Connect the cathode to the negative pole and the anode to the positive pole of the power supply.

6.4.4 Migration test

6.4.4.1 Turn on the power, with the voltage preset at 30 V, and record the initial current through each specimen.

6.4.4.2 Adjust the voltage if necessary (as shown in Table 2). After adjustment, note the value of the initial current again.

6.4.4.3 Record the initial temperature in each anolyte solution, as shown by the thermometer or thermocouple.

6.4.4.4 Choose appropriate test duration according to the initial current (see Table 2).

6.4.4.5 Record the final current and temperature before terminating the test.


15

Table 2 — Test voltage and duration for concrete specimen with normal binder content.

Initial current I30V

(with 30 V) (mA)

Applied voltage U

(after adjustment) (V)

Possible new initial current I0

(mA)

Test duration t (hour)

I0 < 5 60 I0 < 10 96 or longer

5 ≤ I0 < 10 60 10 ≤ I0 < 20 48

10 ≤ I0 < 15 60 20 ≤ I0 < 30 24

15 ≤ I0 < 20 50 25 ≤ I0 < 35 24

20 ≤ I0 < 30 40 25 ≤ I0 < 40 24

30 ≤ I0 < 40 35 35 ≤ I0 < 50 24

40 ≤ I0 < 60 30 40 ≤ I0 < 60 24

60 ≤ I0 < 90 25 50 ≤ I0 < 75 24

90 ≤ I0 < 120 20 60 ≤ I0 < 80 24

120 ≤ I0 < 240 15 60 ≤ I0 < 120 24

240 ≤ I0 < 600 10 80 ≤ I0 < 200 24

I0 > 600 10 I0 > 200 6 NOTE 1: The values in the table are based on the specimens with 100 mm diameter. For the specimens with another diameter,

d (mm), correct the measured current by multiplying a factor of (100/d)2 in order to be able to use the above table.

NOTE 2: For specimens with a special binder content, such as repair mortars or grouts, correct the measured current by multiplying a factor (approximately equal to the ratio of normal binder content to actual binder content) in order to be able to use the above table.

6.4.5 Measurement of chloride penetration depth

6.4.6.1 Remove the specimen by following the reverse of the procedure in 6.4.3. A wooden rod is often helpful in removing the rubber tube from the specimen.

6.4.6.2 Rinse the specimen with tap water.

6.4.6.3 Wipe off excess water from the surfaces of the specimen.

6.4.6.4 Split or dry-cut the specimen axially into two pieces. Choose the piece having the split or cut section more nearly perpendicular to the end surfaces for the penetration depth measurement, and keep the other piece for chloride content analysis (optional). If there is no requirement for chloride content analysis, both the pieces can be used for the penetration depth measurement.

6.4.6.5 Spray 0.1 M silver nitrate solution on to the freshly split or cut section.

6.4.6.6 When the white silver chloride precipitation on the split or cut surface is clearly visible (after about 15 minutes), measure the penetration depth, with the help of the slide calliper and a suitable ruler, from the centre to both edges at intervals of 10 mm (see figure 6) to obtain seven depths. Read the depth to 0.1 mm.

NOTE 1 The term 'dry-cut' here means to saw the specimen using an air-cooled diamond saw.

NOTE 2 If the penetration front to be measured is obviously blocked by the aggregate, move the measurement to the nearest front where there is no significant blocking by aggregate or, alternatively, ignore this depth if there are more than five valid depths.

NOTE 3 If there is a significant defect in the specimen that results in a penetration front much larger than the average, ignore this front as indicative of the penetration depth, but note or photograph it and report the condition.

NOTE 4 To obviate the edge effect due to a non-homogeneous degree of saturation or possible leakage, do not make any depth measurements in the zone within about 10 mm from the edge (see figure 6).


16

Figure 6 — Illustration of measurement for chloride penetration depths

6.4.6 Measurement of surface chloride content (optional)

6.4.6.1 From the other axially split or cut specimen, cut an approximately 5 mm thick slice parallel to the end surface that was exposed to the chloride solution (catholyte).

6.4.6.2 Determine the chloride content in the slice in accordance with prEN 14629 or by a similar method with the same or better accuracy.

NOTE 1 Information about chloride binding capacity of the tested material may be estimated from the surface chloride content.

NOTE 2 The thickness of the slice should always be less than the minimum penetration depth.


Calculate the non-steady state migration coefficient from equation (4):

DRTz FE

x xtnssm

d d= ⋅− α

(4)

where:

EU

L=

− 2 (5)

α = ⋅ −⎛

⎝⎜

⎞

⎠⎟−2 1

21

0

RTz FE

cc

erf d (6)

Dnssm: non-steady state migration coefficient, m2/s;

z: absolute value of ion valence, for chloride, z = 1;

F: Faraday constant, F = 9.648 ×104 J/(V·mol);

U: absolute value of the applied voltage, V;

xd6 xd4 xd2 xd1 xd3 xd5 xd7

10 10 10 10 10 10 mm

L

Specimen

Ruler

10 mm 10 mm Measurement zone


17

R: gas constant, R = 8.314 J/(K·mol);

T: average value of the initial and final temperatures in the anolyte solution, K;

L: thickness of the specimen, m;

xd: penetration depth, m;

t: test duration, seconds;

erf-1: inverse of error function;

cd: chloride concentration at which the colour changes, cd ≈ 0.07 N for OPC concrete;

c0: chloride concentration in the catholyte solution, c0 ≈ 2 N.

Since erf− −×⎛

⎝⎜⎞⎠⎟

1 12 0 07

2. = 1.28, the following simplified equation can be used:

( )( )

( )D

T LU t

xT L x

Unssm dd=

+−

−+−

⎛

⎝⎜⎜

⎞

⎠⎟⎟

0 0239 2732

0 0238273

2.

. (7)

where:

Dnssm: non-steady-state migration coefficient, ×10-12 m2/s;

U: absolute value of the applied voltage, V;

T: average value of the initial and final temperatures in the anolyte solution, °C;

L: thickness of the specimen, mm;

xd: penetration depth, mm;

t: test duration, hour.

6.6 Test report







f) Test results, including the specimen dimensions, concrete age at testing, applied voltage, initial and final currents, initial and final temperatures, testing duration, average and maximum penetration depth, average and maximum migration coefficient, and individual data of penetration depths.

g) Any observation or suspicion of an abnormal penetration front due to defects (cracks, large pore holes, foreign bodies etc.) in the specimen.

h) Optional information about surface chloride content.


18

7 Resistivity test

7.1 Principle

The test is based on the measurement of the electrical resistance of a concrete sample by means of current lines parallel to the basement of the sample.

7.2 Equipment

7.2.1 Equipment for cutting specimens such as a water-cooled diamond saw.

7.2.2 Vacuum container: Capable of containing at least three specimens.

7.2.3 Vacuum pump: Capable of maintaining a pressure of less than 50 mbar (5 kPa) in the container, such as a water-jet pump.

7.2.4 Resistivimeter: able to apply a stable alternating voltage or current at a frequency between 50∼100Hz to the test specimen and to measure respectively the current or voltage drop generated.

NOTE It is acceptable to use an alternating current source with two external multimeters to measure the current and voltage drop.

7.2.5 Electrodes: metallic nets (openings < 2mm) or plates, made of steel, copper or any other good conductive metal and free of superficial impurities (deposits, rust, oxides, etc.), with a diameter equal to that of the specimen.

7.2.6 Contact sponges: two pieces of thin sponges (thickness < 5mm) with a diameter equal that of electrodes.

7.2.7 Weight object: made of non-conductive material with a mass of about 2 kg.

7.2.8 Ruler with a minimum scale of 1 mm.


7.3.1 Test specimen

7.3.1.1 If drilled cores are used, the outermost approximately 10 mm thick layer should be cut off and the next (50 ± 2) mm thick slice should be cut from each core as the test specimen.

7.3.1.2 If cast cylinders are used, cut a (50 ± 2) mm thick slice from the central portion of each cylinder as the test specimen.

7.3.1.3 Measure the thickness with a slide calliper and read to 0.1 mm.

NOTE The term 'cut' here means to saw perpendicularly to the axis of a core or cylinder, using a water-cooled diamond saw

7.3.2 Preconditioning

After cutting, brush and wash away any burrs from the surfaces of the specimen, and wipe off excess water from the surfaces of the specimen. When the specimens are surface dry, place them in the vacuum container for vacuum treatment. Both end surfaces must be exposed. Reduce the absolute pressure in the vacuum container to a pressure in the range of 10~50 mbar (1~5 kPa) within a few minutes. Maintain the vacuum for three hours and then, with the vacuum pump still running, fill the container with the distilled or demineralised water so as to immerse all the specimens. Maintain the vacuum for a further hour before allowing air to re-enter the container. Keep the specimens in the solution for (18 ± 2) hours.

NOTE If the specimens are in whole time cured in the water and no apparent drying has happened during transport, drilling, cutting, and other preparation procedures, the vacuum procedure can be skipped.


19

7.4 Test procedures

7.4.1 Testing room

The testing room shall have a temperature of (20 ± 2) ºC and a relative humidity RH ≥ 45%.

7.4.2 Measurement arrangement

The measurement arrangement is as shown in figure 7.

Figure 7 — Arrangement for the measurement of resistivity by the direct method

7.4.3 Measurement of the ohmic resistance

7.4.3.1 Wet the sponges with tap water and slightly drip them to remove the excess of water.

7.4.3.2 Wipe away the exceed water from the surfaces of the samples.

7.4.3.3 Place the sponges, specimen, electrodes and the 2 kg weight in such a way as shown in figure 7.

7.4.3.4 Impose a current, I, of about 40 mA from the resistivimeter and record the correspondent potential drop to 1 mV as soon as the value becomes stable (after around 5 seconds) , noted as Us+sp.

7.4.3.5 Turn off the current and remove the specimen but keep the sponges, electrodes and the 2 kg weight in the measurement arrangement.

7.4.3.6 Impose a current, I, of about 40 mA from the resistivimeter and record the correspondent potential drop to 1 mV as soon as the value becomes stable (after around 5 seconds) , noted as Usp.

NOTE The value of Usp/I should not be higher than 100 mV/mA. Otherwise the sponges shall be rewetted and the measurement repeated.


Calculate the resistivity from equation (8):

IUU

Ld spsps −

⋅π

=ρ +

4000

2 (8)

Resistivimeter

Holder

Specimen

Electrodes Sponges

2 kg weight


20

where:

ρ: resistivity of concrete, Ω·m;

d: diameter of the specimen, mm;

L: thickness of the specimen, mm;

Us+sp: absolute value of the potential drop recorded with specimen and sponges, mV;

Usp: absolute value of the potential drop recorded with sponges only, mV;

I: Imposed current, mA.

7.6 Test report







f) Test results, including the specimen dimensions, concrete age at testing, temperature, imposed current, potential drops, and resistivity.

g) Any observation or suspicion of defects (cracks, large pore holes, foreign bodies etc.) in the specimen.

8 Precision data

Based on the inter-laboratory test organised in the EU-project “Chlortest” (GRD1-2002-71808/G6RD-CT-2002-00855) with 3 replicates and 12 laboratories, the values of repeatability and reproducibility of various test methods are listed in Table 3.

Table 3 — Precision data of various test methods

Test method Parameter Repeatability sr Reproducibility sR

Rapid migration test Dnssm sr = 0.152Dnssm (R2 = 0.86) sR = 0.236Dnssm (R2 = 0.93)

Resistivity test ρ sr = 0.105ρ (R2 = 0.97) sR = 0.251ρ (R2 = 998)

Immersion test

Dnssd

Cs

KCr

sr = 0.201Dnssd (R2 = 0.94)

sr = 0.177Cs (R2 = 0.14)

sr = 0.090KCr (R2 = 0.76)

sR = 0.283Dnssd (R2 = 0.99)

sR = 0.223Cs (R2 = 0.41)

sR = 0.135KCr (R2 = 0.88)

Steady state migration test Ds

Dns

sr = 0.220Ds (R2 = 0.92)

sr = 0.237Dns (R2 = 0.76)

sR = 0.759Ds (R2 = 0.95)

sR = 0.454Dns (R2 = 0.92)


21

9 Annex

Major modifications in the proposed methods for standardisation are summarised in the following table.

Proposed Method Method Used in ChlorTest

Immersion Test

• Cut off the outmost about 10 mm layer after lime-water saturation and coating to prevent the exposure surface from calcium densification

• Exposure duration “at least 35 days, but in normal cases not exceed 40 days” to reduce the age effect in the normal tests

NT BUILD 443

• Cut off the outmost about 10 mm layer before lime-water saturation and coating

• Exposure duration “at least 35 days”

Rapid Migration Test

• Vacuum saturation with distilled or demineralised water to simplify the procedure, because during this short saturation period there should not be remarkable leaching problem

• After the migration test, “split or dry-cutting the specimen”, to facilitate the laboratory that may have no machine for splitting and also to make the depth measurement easier

NT BUILD 492

• Vacuum saturation with saturated lime-water

• After the migration test, “split the specimen”

Resistivity test

• Thickness of specimen is specified to (50 ± 2) mm to reduce the size effect

• Vacuum to a pressure of “less than 50 mbar (5 kPa) in the container”, to be more practical, especially when specimens are moist or water has been filled in the container

UNE 83XXX (Carmen Andrade’s translation)

• No size specification

• Vacuum to a pressure of “less than 1 mmHg”*

* Corresponding to 0.76 mbar or 0.076 kPa. This pressure was specified in ASTM C1202, but can hardly be achieved without a special vacuum pump, especially when specimens are moist or water has been filled in the container.

WP2 REPORT – PRE-EVALUATION OF TEST METHODS

Deliverables D5-8 CONTRACT N°: G6RD-CT-2002-00855


ACRONYM: CHLORTEST




EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 28 WP2 REPORT – Pre-Evaluation of Test Methods

















ACKNOWLEDGEMENT: The present document is deliverables of Workpackage 2 – “Pre-Evaluation of Test Methods”. The consortium members IETcc, SP, TNO, QUB, LNEC, IBRI, INSA and LCPC were involved in the work of this part of the project. The work was led by IETcc and assisted by SP.

This document was prepared by Marta Castellote and Carmen Andrade (IETcc)

FURTHER INFORMATION: Regarding this document: Dr Carmen Andrade Consejo Superior de Investigaciones Cientificas Serrano Galvache s/n ES-28033 MADRID, Spain Tel. +34-91 3020440; Fax: +34-91 3020700 e-mail: [email protected]



TABLE OF CONTENTS

Page

1 INTRODUCTION 5

2 PARTICIPATING LABORATORIES 5

3 TEST METHODS SELECTED 6

4 MIXES 6

5 TEST RESULTS AND PRECISION ANALYSIS 4

6 STATISTICAL ANALYSIS OF THE DATA 13

7 DISCUSSION 18

8 CONCLUDING REMARKS AND SUGGESTIONS 27

REFERENCES 28



1 INTRODUCTION According to work-plan of the Chlortest project, the Objectives to be reached by the Workpackage 2 – Pre-evaluation of different test methods – are:

• Pre-evaluation of five to six promising test methods selected from WP1 • Identify about three best methods for final evaluation in WP5

To reach these objectives, and according to the work plan, six test methods were selected from WP1 and their results have been evaluated. From the results, four methods have been identified for final evaluation in WP5 on a basis of their values, accuracy, information that supply and practical convenience. 2 PARTICIPATING LABORATORIES The laboratories that have participated in this WP are listed in table 1, where it can be seen that, in addition to the seven labs involved in WP2 of the project, LCPC was a volunteer laboratory to perform also the tests, what was accepted from the consortium. In table 1, the different laboratories have been listed in alphabetical order, not being the number that appears in the first column of the table valid for identification of the results of each lab. On this purpose, each of them has been given randomly a different number. Table 1: Laboratories that took part in the WP2. Number Laboratory Country

1 IBRI Iceland 2 IETcc Spain 3 INSA France 4 LCPC France 5 LNEC Portugal 6 QUB United Kingdom 7 SP Sweden 8 TNO Netherlands


3 TEST METHODS SELECTED During the kick- off meeting, the participants agreed upon the following six test methods for pre-evaluation:

• D2: NT BUILD 443 [1]: Immersion in 165g NaCl per litre at 23 °C for 35d. • R1: Resistivity test [2,3]: Measuring resistivity of concrete and calculation of the

effective diffusion coefficient. • M3: INSA steady-state-migration [4],: Measuring concentration changes in the

upstream cell. • M4: NT BUILD 492 [5]: Non-steady state migration under 10~60 V for 24 h.

• M6: Multi-regime method [6]: Calculation of steady and non-steady state diffusion

coefficients by measuring conductivity change in the downstream cell. • PrEN: prEN 13396 [7]: Immersion in 3% NaCl at 40 °C for 28, 90 and 180 days

A detailed description of all the selected six methods for pre-evaluation are available at the references 1 to 7 and at the RILEM TC 178’s homepage. 4 MIXES It was agreed at the kick-off meeting to cast concrete specimens in different countries with a typical mixture design used in the respective country. Upon the agreement, Chalmers (Sweden) prepared the concrete specimens with 5% silica fume and w/b 0.4. The IETcc (Spain) prepared the OPC (CEM I) concrete specimens with w/c 0.45. LNEC (Portugal) prepared the concrete specimens with CEM IV (39% fly ash) and w/b 0.45 and TNO (Netherlands) prepared the concrete specimens with CEM III (76% slag) and w/b 0.45. These four mixes were cast, cured, drilled and the cores delivered according to the procedures given in the quality assurance procedure (QAP). The mixture proportions of the concrete manufactured for WP2 are summarized in Table 2. 5 TEST RESULTS As has been commented, each laboratory has been labelled with a number that has been randomly assigned in order to maintain the confidentiality of the results corresponding to each laboratory.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Tables 3 to 7 present the individual results for every test. It can be noticed that most of the laboratories, according to the recommended procedure, have performed the different tests by triplicate. Only a few numbers of laboratories have not reported the whole set of results. Table 2 Mixture proportions of concrete manufactured for WP2.

Kg/m3 SF OPC FA SL Cast by CHALMERS IETcc LNEC TNO Cement 399

I-42.5 N V/SR/LA

400 I-42.5 R/SR

340 IV/B 32.5 R

350 III/B 42.5 LH

HS Silica fume 21 (slurry) ----- -----

Fly ash ----- ----- ----- Slag ----- ----- -----

Water 168 180 153 157.5 Sand 842.5 (0-8

mm) 742 (0-6 mm) 62 (0-2 mm)

603 (0-4 mm) 70 (0-1 mm) 790 (0-4 mm)

Coarse aggregate

842.5 (8-16 mm)

1030 (6-16 mm)

619 (4-12 mm) 555 (12-25

mm)

1040 (4-16 mm)

Total aggregate 1685 1772 1823 1830 Super-

plasticisers 3.4

Cementa 92M 4.8

Melcret 222

4.1 Rheobuild 1000

3.9 Cretoplast

Air content 6% ----- 1.5 W/C 0.42 0.45 W/B 0.40 0.45 0.45 0.45

Strength (MPa) 63 45 52.6 Slump (mm) > 150 Casting date November 2002 May 2003 June 2003 October 2003

Delivered May 2003 August 2003 October 2003 April 2004 Concerning the results for prEN 13396, it has been considered that the increment nº 2 was the most representative for the analysis, and only these increment, for each time of exposure, has been taken into account in the statistical treatment the results.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 3: Individual results for method D2 corresponding to the Dns “effective chloride

transport coefficient” and K, “penetration parameter”.

Table 4: Individual results for methods M3 (INSA steady-state-migration) and M4 (NT

BUILD 492).

MethodLaboratory Mix 1 Mix 2 Mix 3 Mix 4

Lab 1, test 1 0,62 0,79test 2 0,61 1,1test 3 0,81 0,88

Lab 2, test 1 1,56 1,928 1,63 1,55test 2 1,58 1,524 1,54 1,68test 3 1,64 1,553 1,71 1,64

Lab 3, test 1 0,093 0,2 0,24 0,08test 2 0,136 0,25 0,16 0,06test 3 0,23 0,13 0,1


Lab 5, test 1 1 5,2 1,5 0,7test 2 0,8 5,8 1,5 0,4test 3 0,8 4,7 2,1 0,9

Lab 6, test 1 1,37 2,8test 2 1,53 3,02test 3 1,1 2,92


Lab 8, test 1 1,41 2,39 1,31test 2 1,6 2,83 1,94test 3 2,26 2,14 0,97

M3 (Ds x 10-12 m2/s) MethodLaboratory Mix 1 Mix 2 Mix 3 Mix 4

Lab 1, test 1 2,6 16,3 7,3 2,4test 2 2,7 15,9 7,3 3,2test 3 3 14,1 8,6 2,4

Lab 2, test 1 2,2 15,5 8,3 1,8test 2 2,7 17 6,7 2,34test 3 3,4 15,6 7,5 1,68

Lab 3, test 1 2,7 15,1 3,8 2,3test 2 2 17 6 2,8test 3 2,7 15,3 4,6 3,6

Lab 4, test 1 1,8 12,8 4,3 2,5test 2 1,9 18,6 3,5 2test 3 2 14,1 3,3 2,1

Lab 5, test 1 2,8 17,7 3,2 2,2test 2 4,3 17,2 3,2 2,5test 3 3,4 15,7 6 2,3

Lab 6, test 1 1,59 4,4 2,4test 2 2,9 4,7 81,1test 3 2 3,1 3,6

Lab 7, test 1 3,1 14,9 4,8 2,2test 2 3 14,8 5,7 2,9test 3 3 14,9 5,4 2,9

Lab 8, test 1 2 15,4 3,4test 2 2,6 21,2 6,5test 3 2,3 13,6 4

M4 (Dns x 10-12 m2/s)


Lab 1, test 1 2,01 16,9 9 1,8test 2 3,73 18,2 5,5 2,5test 3 2,33 16,4 9,8 2








D2 (Dns x 10-12 m2/s) MethodLaboratory Mix 1 Mix 2 Mix 3 Mix 4

Lab 1, test 1 21 62 47 21test 2 28 62 37 24test 3 23 60 50 22





Lab 6, test 1 28,26 232 23,39test 2 24,34 287 24,91test 3 25,88 192 22,46

Lab 7, test 1 22 54 38 23test 2 21 64 36 23test 3 44 38 25

Lab 8, test 1 23 44 62test 2 23 49 53test 3 23 51 43

D2 (K mm/year1/2)

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 5: Individual results for method M6 (Multi-regime method) corresponding to the Ds

(steady state diffusion coefficient) and Dns (non-steady state diffusion coefficient).

Table 6: Individual results for method R1 (Resistivity in ohm.cm) and for method prEN

13396 (% Cl) for the increment nº 2 at the duration of the tests of 28 days.




Lab 3, test 1 0,72 0,08test 2 0,15 0,87 0,09test 3 0,28 0,93 0,54 0,156


Lab 5, test 1 3,5 5,3 2,7 1test 2 2,7 5,7 2,9 0,8test 3 5,6 2,2 1

Lab 6, test 1 3,04test 2 4,49test 3 3,84


Lab 8, test 1 0,89 4 0,4test 2 0,92 3,81 0,8test 3 50 2,62 0,86

M6 (Ds x 10-12 m2/s) MethodLaboratory Mix 1 Mix 2 Mix 3 Mix 4

Lab 1, test 1 4,8 16,4 17,6 3,9test 2 4,5 15,1 45 3,8test 3 7,6 17,4 10,5 4,1

Lab 2, test 1 4,2 13,6 9,27 2,28test 2 4,89 18 8,48 2,57test 3 4,67 3,7

Lab 3, test 1 27 1,55test 2 4,46 18,68 1,54test 3 5,4 18,82 6,79 1,54

Lab 4, test 1 8 22,7 8,98test 2 17,2 23,5 11,5test 3 6,27 22,2


Lab 6, test 1test 2test 3


Lab 8, test 1 54 49test 2 141test 3 65 84

M6 (Dns x 10-12 m2/s)


Lab 1, test 1 0,086 0,211 0,167 0,15test 2 0,096 0,174 0,137 0,2test 3





Lab 6, test 1 0,14 0,043 0,292test 2 0,095 0,083 0,352test 3


Lab 8, test 1 0,123 0,237test 2 0,141 0,237test 3

prEN 13396 - 28 d - increment 2 (%)MethodLaboratory Mix 1 Mix 2 Mix 3 Mix 4








Lab 8, test 1 41850 8700 30680test 2 42100 8350 28090test 3 37760 8410 27560

R1 (ohm.cm)

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 7: Individual results for method prEN 13396 (% Cl) for the increment nº 2 at the

duration of the tests of 90 and 180 days. Tables 8 to 13 present the cell mean values and the intracell standard deviations, derived from tables 3 to 7, for the different methods, calculated according to procedures given in the International Standard ISO 5725-2:1994 [8]. Table 8: Cell mean values and intracell standard deviations for the method D2 (NT BUILD

443) corresponding to the Dns “effective chloride transport coefficient” and K, “penetration parameter”.

D2-DLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 2,690 17,167 8,100 2,100L2 1,710 17,460 5,363 1,567L3 1,357 15,570 4,233 3,650L4 1,610 7,377 3,190L5 1,017 29,377 8,920 2,493L6 3,717 45,767 2,973L7 1,810 14,013 6,367 2,223L8 2,307 12,160 11,670

Cell means

D2-KLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 24,000 61,333 44,667 22,333L2 17,700 57,243 31,863 17,200L3 15,543 53,023 28,327 25,557L4 16,910 36,703 24,827L5 15,667 76,667 43,667 23,000L6 26,160 237,000 23,587L7 21,500 54,000 37,333 23,667L8 23,000 48,000 52,667

Cell means D2-KLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 3,606 1,155 6,807 1,528L2 1,418 4,633 1,185 2,589L3 2,535 4,751 2,085 2,741L4 3,283 6,369 6,278L5 1,528 12,097 10,693 1,000L6 1,975 47,697 1,237L7 0,707 10,000 1,155 1,155L8 0,000 3,606 9,504

Spread within cells

D2-DLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,915 0,929 2,287 0,361L2 0,265 2,781 0,422 0,431L3 0,490 2,763 0,660 0,775L4 0,624 2,436 1,350L5 0,287 10,028 5,131 0,669L6 0,565 18,294 0,305L7 0,014 5,157 0,692 0,233L8 0,159 2,042 3,610

Spread within cells




Lab 3, test 1 0,148test 2 0,177test 3






prEN 13396 - 180 d - increment 2 (%)MethodLaboratory Mix 1 Mix 2 Mix 3 Mix 4







Lab 7, test 1 0,329 0,218 0,231 0,155test 2 0,3 0,217 0,218test 3


prEN 13396 - 90 d - increment 2 (%)

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 9: Cell mean values and intracell standard deviations for the method M3 (INSA steady-

state-migration).

Table 10: Cell mean values and intracell standard deviations for the method M4 (NT BUILD

492).

Table 11: Cell mean values and intracell standard deviations for the method M6 (Multi-

regime method) corresponding to the Ds (steady state diffusion coefficient) and Dns (non-steady state diffusion coefficient).

Table 12: Cell mean values and intracell standard deviations for the method R1 (Resistivity in

ohm.cm)

M3Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,680 0,923L2 1,593 1,668 1,627 1,623L3 0,115 0,227 0,177 0,080L4 68,437 63,790 49,493 0,431L5 0,867 5,233 1,700 0,667L6 1,333 2,913L7 0,340 0,580 0,290 0,247L8 1,757 2,453 1,407

Cell means M3Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,113 0,159L2 0,042 0,225 0,085 0,067L3 0,030 0,025 0,057 0,020L4 40,205 0,475 15,698 0,058L5 0,115 0,551 0,346 0,252L6 0,217 0,110L7 0,036 0,070 0,026 0,045L8 0,446 0,349 0,492

Spread within cells

M4Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 2,767 15,433 7,733 2,667L2 2,767 16,033 7,500 1,940L3 2,467 15,800 4,800 2,900L4 1,900 15,167 3,700 2,200L5 3,500 16,867 4,133 2,333L6 2,163 4,067 29,033L7 3,033 14,867 5,300 2,667L8 2,300 16,733 4,633

Cell means M4Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,208 1,172 0,751 0,462L2 0,603 0,839 0,800 0,352L3 0,404 1,044 1,114 0,656L4 0,100 3,044 0,529 0,265L5 0,755 1,041 1,617 0,153L6 0,670 0,850 45,095L7 0,058 0,058 0,458 0,404L8 0,300 3,972 1,644

Spread within cells

M6-DnsLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 5,633 16,300 24,367 3,933L2 4,587 15,800 8,875 2,850L3 4,930 21,500 6,790 1,543L4 10,490 22,800 10,240L5 1,275 6,200 2,627 2,877L6L7 6,423 23,333 9,840 2,143L8 54,000 85,000

Cell means M6-DnsLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 1,710 1,153 18,218 0,153L2 0,352 3,111 0,559 0,750L3 0,665 4,764 0,006L4 5,875 0,656 1,782L5 0,403 0,363 0,386 0,480L6L7 1,067 0,751 0,697 1,861L8 49,153

Spread within cells

M6-DsLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,387 1,543 0,340 0,200L2 1,800 2,100 1,108 0,640L3 0,215 0,840 0,540 0,109L4 32,033 5,300 35,223 0,547L5 3,100 5,533 2,600 0,933L6 3,790L7 0,800 1,267 0,467 0,290L8 0,905 3,477 0,687

Cell means M6-DsLaboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,095 0,163 0,044 0,017L2 0,436 0,000 0,517 0,044L3 0,092 0,108 0,041L4 26,992 0,693 4,795 0,055L5 0,566 0,208 0,361 0,115L6 0,726L7 0,173 0,208 0,058 0,017L8 0,021 0,748 0,250

Spread within cells

R1Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 35298 6689 21433 42633L2 26028 5624 19929 36603L3 19167 8317 157600 22267L4 42684 18715 29139 46985L5 38056 5294 39814 37964L6L7 30422 7579 24704 32908L8 40570 8487 28777

Cell means R1Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 2142 146 1890 3686L2 843 390 2705 2319L3 4168 2086 1905 11697L4 2125 2220 461 1725L5 3331 353 3410 3411L6L7 4547 23 1617 2958L8 2437 187 1670

Spread within cells

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 13: Cell mean values and intracell standard deviations for the method prEN 13396 (%

Cl) for the increment nº 2 at the duration of the tests of 28, 90 and 180 days.

prEN-28-2Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,091 0,193 0,152 0,175L2 0,190 0,225 0,104 0,280L3 0,084 0,945 0,180 0,800L4 0,153 0,341 0,147 0,257L5 0,083 0,145 0,108 0,282L6 0,118 0,063 0,322L7 0,089 0,209 0,113 0,088L8 0,132 0,237

Cell means prEN-28-2Laboratory Mix 1 Mix 2 Mix 3 Mix 4

L1 0,007 0,026 0,021 0,035L2 0,035 0,023 0,014 0,014L3 0,036 0,502 0,014 0,141L4 0,054 0,018 0,023 0,029L5 0,029 0,010 0,013 0,005L6 0,032 0,028 0,042L7 0,011 0,077 0,000 0,013L8 0,013 0,000

Spread within cells


L1 0,190 0,170 0,185L2 0,221 0,262 0,163 0,335L3 0,163 0,560 0,255 0,800L4 0,281 0,544 0,282L5 0,144 0,181 0,167 0,339L6 0,155 0,209 0,356L7 0,315 0,218 0,225 0,155L8 0,179 0,313


L1 0,027 0,002 0,013L2 0,014 0,004 0,040 0,007L3 0,021 0,255 0,021 0,000L4 0,012 0,033 0,059L5 0,027 0,006 0,005 0,025L6 0,031 0,016 0,040L7 0,021 0,001 0,009 0,000L8 0,018 0,004

Spread within cells


L1 0,255 0,186L2 0,266 0,275 0,163 0,341L3 0,163L4 0,337 0,617 0,226L5 0,171 0,209L6 0,180 0,231L7 0,451 0,224 0,186L8 0,264 0,201 0,161


L1 0,023 0,023L2 0,037 0,009 0,039 0,008L3 0,021L4 0,000 0,005 0,010L5 0,008 0,049L6 0,028 0,030L7 0,100 0,019 0,007L8 0,020 0,043 0,009

Spread within cells


6 STATISTICAL ANALYSIS OF THE DATA The statistical treatment of the data has been carried out according to the International Standard ISO 5725-2:1994 [8] for the determination of the accuracy (trueness and precision) of measurement methods and results. Part 2: Basic method for the determination of the repeatability and reproducibility of a standard measurement method. These results were double-checked by each laboratory to assure the correctness. According to this standard, the parameters to be calculated are the mean value (m), the repeatability standard deviation (sr), the reproducibility standard deviation (sR), and the relationship between m and (sr), (sR). As a first step in the statistical treatment, it is necessary to critically examine individual values (tables 3 to 7) in order to find entries that are considered irreconcilable with the other data. According to the standard, when several unexplained abnormal test results occur at different levels within the same laboratory, then, that laboratory may be considered to be an outlier. It may then be reasonable to discard some or all of the data from such an outlying laboratory. In addition, the standard points out that, obviously erroneous data should be corrected or discarded. For each method, in this first analysis, only obvious individual discordant data have been taken out because if further treated according to this standard, all the individual data from the lab would have been discarded, even being only one individual data the aberrant one. When all the individual data belonging to one lab are discordant from the rest, they have been examined in the systematic application of statistic tests for outliers included in the standard (step 2). Therefore, in this step, the following data have been discarded:

• M6-Ds-Lab8-Mix1-test3 • M6-Dns-Lab4-Mix1-test2 • M6-Dns-Lab1-Mix3-test2 • M6-Dns-Lab8-Mix3-test3 • R1-Lab3-Mix4-test3 • M4-Lab6-Mix4-test2

It is necessary to point out that for two labs it was obvious that when applying method M6 there was a mistake in the calculations. After getting in contact and given the permission of the responsible people of the laboratories, the data were corrected, being the final values those given in table 5. The second step consists in the application of statistic tests for stragglers and outliers. The three methods recommended by the standard have been used:

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 28 WP2 REPORT – Pre-Evaluation of Test Methods a: Graphical consistency technique: Mandel´s h and k statistics. The Mandel´s h statistic, is the between-laboratory consistency statistic and, for each lab, it is calculated by dividing the cell deviation (cell mean minus the grand mean for the corresponding mix) by the standard deviation among cell means. The Mandel´s k statistic, is the within-laboratory consistency statistic and, for each lab, it is calculated by dividing the intracell standard deviation by the pooled within-cell standard deviation for each mix. The h and k values are plotted for each cell in order of laboratory, in groups for each mix (and separately grouped for the several mixes examined by each laboratory). Examination of these plots may indicate that specific laboratories exhibit patterns of results that are markedly different form the others in the study. The numerical technique has been applied in the following way:

• If the test statistic is less than or equal to its 5% critical value, the item tested is accepted as correct.

• If the test statistic is greater than its 5% critical value and less or equal to its 1% critical value, the item tested is called a straggler, and is marked as it, but it is retained as correct.

• If the test statistic is greater than its 1% critical value, the item tested is called outlier, it is marked as it and in general, it is discarded.

The details of these tests are properly described in the Standard ISO 5725. An example of the application of this test to the D2-D method is given in figures 1 and 2, where the Mandel´s h and statistic are plotted. According to these figures, the laboratory 6 is an straggler for the Mandel’s h statistic for the mixes 1and 2. Concerning the Mandel’s k statistic (figure 2), the laboratory 1 is and straggler and laboratories 6, 5 and 4 are outliers for mixes 2, 3 and 4 respectively. In these pictures, the depicted critical values have been those corresponding to 8 labs participating, as it is the less conservative value, in order to retain as much laboratories as possible. However when making the calculations, the specific number of labs in each level (for each mix) have been considered. The results obtained from these techniques, with detailed indication of straggler and outliers are given in table 14.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Figure 1: Mandel´s h statistic for the method D2 (NT BUILD 443) corresponding to the

diffusion coefficient, D. Figure 2: Mandel´s k statistic for the method D2 (NT BUILD 443) corresponding to the

diffusion coefficient, D.

-2,500

-2,000

-1,500

-1,000

-0,500

0,000

0,500

1,000

1,500

2,000

2,500

1 2 3 4 5 6 7 8

Laboratory i

Man

del´s

sta

tistic

, h

mix 1mix 2mix 3mix 4

0,000

0,500

1,000

1,500

2,000

2,500

1 2 3 4 5 6 7 8

Laboratory i

Man

del´s

sta

tistic

, k

mix 1mix 2mix 3mix 4

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 28 WP2 REPORT – Pre-Evaluation of Test Methods b: Cochran´s test The Standard ISO 5725 assumes that between laboratories only small differences exist in the repeatability variances. However, this is not always the case, so Cochran´s test has been recommended to check the validity of this assumption. The details of this test are properly described in the Standard ISO 5725, but in summary, a parameter, the Cochran´s test statistic, C, is calculated and compared with its 1% and 5% critical values with the same criterion than in the case of Mandel´s statistics. In table 14, straggler and outlier values according to Cochran´s test have also been specified. c: Grubbs´s test The aim of this test is the same than Cochran´s one but regarding the reproducibility variances by testing the significance of the largest observation (sh), of the smallest observation (sl), of the two largest observations (dh) and the two smallest observations (dl). The details of these tests are properly described in the Standard ISO 5725, but in summary, for these four cases the Grubb´s statistics are computed and they are compared with the critical values for Grubbs´s test. In table 14, straggler and outlier values according to Grubbs´s test have also been specified. The third step consists in the calculation of the values of the repeatability standard deviation (sr) and reproducibility standard deviation (sR). To do so, it is necessary to calculate previously the variance of repeatability (sr

2) and the variance of reproducibility (sL2),

according to the equations detailed in the Standard ISO 5725. The statistical values determined according to the standard ISO 5725, as well as the details on the stragglers and outliers according to the different methods applied and then the laboratories discarded and retained for the study are presented in table 14. For a better understanding of table 14, a detailed example of the meaning of the symbols and abbreviations used (for method D2-D and mix 2) is as follows:

• The number of laboratories, n, that finally has been included for the calculations is 6. • The laboratory 6 (L6) is an straggler according to the Mandel´s h statistic • The laboratory 5 (L5) is an outlier according to Cochran´s test and according to

Grubb´s test for the double highest observation. • The laboratory 6 (L6) is an outlier according to Mandel´s k statistic, Cochran´s test

and according to Grubb´s test for the double highest observation. • The laboratory 6 (L6) has been rejected for further analysis • The mean value, m, for the 6 remaining laboratories is : 17.624 • The repeatability standard deviation, sr, is 4.959 • The reproducibility standard deviation, sR, is 7.314 • The repeatability coefficient of variation % sr/m is 28.1 • The reproducibility coefficient of variation % sR/m is 41.5

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 14: Statistical values determined according to the standard ISO 5725. Notes on stragglers and outliers:

M-h : According to Mandel´s h statistics M-k : According to Mandel´s k statistics M-k-h : According to Mandel´s h and k statistics

C: According to Cochran´s test G-sl: According to Grubb´s test for the single lowest observation G-sh: According to Grubb´s test for the single highest observation

G-dl: According to Grubb´s test for the double lowest observation

G-dh: According to Grubb´s test for the double highest observation LX-1: One obvious individual discordant data have been discarded

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 28 WP2 REPORT – Pre-Evaluation of Test Methods The dependence of the repeatability standard deviation (sr) and reproducibility standard deviation (sR) on the mean value, for the different methods is shown in figures 3 to 7. From these figures it can be deduced that in most of the cases, a linear relationship can be found. The equations of the linear regressions for all the cases are given in figures 3 to 7.

Figure 3: Repeatability (sr) and reproducibility (sR) standard deviations in function of the mean value obtained for the method D2 (NT BUILD 443) corresponding to the Dns “effective chloride transport coefficient” and K, “penetration parameter”.

Figure 4: Repeatability (sr) and reproducibility (sR) standard deviations in function of the

mean value obtained for the method R1 (Resistivity) and for the corresponding diffusion coefficients, R1-D.

D2-D

sr = 0,2848 m + 0,1048R2 = 0,975

Sr = 0,4159 m + 0,0383R2 = 0,9968

012345678

0 5 10 15 20m

sr, S

r

srSr

D2-KSr = 0,2094 m + 0,0515

R2 = 0,8996

sr = 0,1308 m + 0,1155R2 = 0,8766

0

2

4

6

8

10

12

14

0 20 40 60 80m

sr, S

r

srSr

R1Sr = 0,2036 m + 715,38R2 = 0,7736

sr = 0,0682 m + 423,36R2 = 0,9304

0

2000

4000

6000

8000

10000

0 10000 20000 30000 40000 50000m

sr, S

r

srSr

R1-DSr = 0,2045 m + 0,0379

R2 = 0,9662

sr = 0,1127 m - 0,0041R2 = 0,9778

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0 1 2 3 4m

sr, S

r

srSr


Figure 5: Repeatability (sr) and reproducibility (sR) standard deviations in function of the mean value obtained for the methods M3 (INSA steady-state-migration) and M4 (NT BUILD 492).


mean value obtained for the method M6 (Multi-regime method) corresponding to the De (steady state diffusion coefficient) and Da (non-steady state diffusion coefficient).

M3

Sr = 1,2127 m - 0,3898R2 = 0,866

sr = 0,1151 m + 0,082R2 = 0,3856

0

0,4

0,8

1,2

1,6

2

0,5 1 1,5 2m

sr, S

r

srSr

M4Sr = 0,1009 m + 0,5946R2 = 0,6537

sr = 0,1165 m + 0,2456R2 = 0,964

0

0,5

1

1,5

2

2,5

0 5 10 15 20m

sr, S

r

srSr

M6-DeSr = 0,6482 m + 0,175R2 = 0,9667

sr = 0,122 m + 0,102R2 = 0,7156

0

0,5

1

1,5

2

2,5

0 1 2 3 4m

sr, S

r

srSr

M6-Da

Sr = 0,3788 m + 0,5347R2 = 0,919

sr = 0,094 m + 0,8202R2 = 0,7593

012345678

0 5 10 15 20m

sr, S

r

srSr



mean value obtained for the method prEN 13396 (% Cl) for the increment nº 2 at the duration of the tests of 28, 90 and 180 days.

7 DISCUSSION In order to make a pre-evaluation of the methods used to identify the best tests methods for final evaluation in WP5, four key-parameters have to be taken into account: 1) The absolute value obtained, 2) The accuracy, 3) The relevance of the information that supply and 4) The appropriateness concerning hours-man and resources. Within each of these four parameters, several indicators can be identified. For each of the key-parameters, those are:

1: The absolute value obtained

o 1-1: Deviation of the mean value with respect to the target value (if there is a target value).

o 1-2: Significant differences with the average (if there is not a target value and the rest gives very similar values).

2: The accuracy

o 2-1: According to the repeatability coefficient of variation o 2-2: According to the reproducibility coefficient of variation

3: The relevance of the information that they supply

pren-28d-pt2

Sr = 0,4934 m - 0,0211R2 = 0,8596

sr = 0,0448 m + 0,0194R2 = 0,0961

0

0,02

0,04

0,06

0,08

0,1

0,05 0,1 0,15 0,2 0,25m

sr, S

r

srSr

pren-90d-pt2

Sr = 0,8438 m - 0,1113R2 = 0,9638

Sr = 0,0279 m + 0,0339R2 = 0,0075

0

0,05

0,1

0,15

0,2

0,25

0,3

0,1 0,2 0,3 0,4 0,5m

sr, S

r

srSr

pren-180d-pt2

Sr = 1,0599 m - 0,175R2 = 0,9072

sr = -0,0342 m + 0,0344R2 = 0,0062

0

0,02

0,04

0,06

0,08

0,1

0,12

0,15 0,2 0,25 0,3m

sr, S

r

srSr


o 3-1: Concerning the diffusion coefficients Ds and Dns o 3-2: Concerning the acceleration during the test

4: The appropriateness concerning hours-man and resources

o 4-1: Handling after the test: Need of milling, grinding or cutting, splitting,... o 4-2: Need of Cl- analysis o 4-3: Difficulty of handling

Each of these indicators have been classified in three categories, divided according to an incremental favourable significance, in order to make the ranking of the methods objective. The key-parameters, indicators and limits for quantification are given in table 15. For each indicator, the worst category has been assigned 1 point, 2 for the medium one and 3 points for the better characteristics of the method. Therefore, for the first indicator, there is a total score of 3 points, as the two indicators exclude mutually, 6 for the second one, 6 for the third one and 12 for the fourth one. Then, it is necessary to average of the weight of each key parameter, in order all of them to have the same weight. For example a test method having 6 points for a total of 6 will have “1” for that parameter. As another example, 3 points on 6 on a parameter will be “0.5” of total score for that parameter. Table 15: Key-parameters, indicators identified and limits for quantification.

Quantification Parameters Indicators 1 point 2 points 3 points total

1-1 Deviation of the mean value with respect to the target value

(%bias)>50 or impossible to calculate

10<(%bias)<50 10>(%bias)1 Absolute value obtained

1-2 Significant differences with the average (if there is not a target value

and the rest gives very similar values)

(%bias)>50 10<(%bias)<50 10>(%bias)

3

2-1 According to the repeatability coefficient of variation

r/m(%)>25 10<r/m(%)<25 10>r/m(%)

2 Accuracy

2-2 According to the reproducibility R/m(%)>60 30<R/m(%)<60 30>r/m (%)

6

3-1 Concerning the diffusion coefficients Ds and Dns

None One of them Both 3 Relevance of the information

supplied 3-2 Acceleration Temperature Electrical field/Concentrati

on

None

6

4-1 Handling after the test: Need of milling, grinding or cutting, splitting

Milling/ grinding

Cutting/splitting None

4-2 Need of Cl- analysis Solid samples Liquid samples Colorimetric

No

4-3 Difficulty of handling Cell and chemical handling

Cell No

4 Appropriateness

4-4 Time consumption > 15 days 8 h > t < 15 d < 8 hours

12

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 22 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Application to these criteria to the methods

Concerning the first parameter (absolute value obtained) there is not a definite reference to be used as a target value, as in the case of steady state tests, no natural tests have been performed. In the case of non-steady-state tests, a natural method have been carried out; however, it is known that in non stationary tests, different experimental conditions, even in natural diffusion can lead to different results. So, the “true value” to be considered in each case is going to be the mean value obtained. In this way, the mean values of the diffusion coefficients obtained for the three mixes obtained by the different methods used are given in figure 8 (a-b) for steady-state and non steady-state coefficients respectively. In the case of M-R, the steady-state coefficient have been calculated from the equation proposed in the recipe M-R by the IETcc and properly described in [2,3].

Figure 8 (a-b): Mean values of the diffusion coefficients obtained for the three mixes

obtained by the different methods The percentages of bias with respect to the average for each mix and method are given in table 16, where, the average (in absolute value) is also given with quantification of the indicator 1-2 purposes. From the results in figure 3 and in table 16 it can be noticed that for both transport regimes, the mean of the coefficients obtained by the three methods are similar enough. All the methods give values in the same range of magnitude and they discriminate well between the different types of concretes, being all of them assigned with 2 points as long as indicator 1-2 is concerned 10<(%bias)<50.

b) non-steady-state

0

5

10

15

20

Mix 1 Mix 2 Mix 3 Mix 4

Dns

(E12

) (m

2 /s)

D2-DM4M6-Dns

a) steady-state

0

1

2

3

4


Ds

(E12

) (m

2 /s)

M3M6-DeM-R1

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 23 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 16: Percentages of bias with respect to the average for each mix and average of bias for

each method.

Concerning the prEN method, the mean of the results obtained are given in figure 9, where it can be seen that, with the exception of mix 1, the values of concentration obtained after 180 days are smaller than those registered after 90 days, which is not the expected behaviour. In this case, it is not possible to calculate the bias, as there is no other method giving those kind of results. Therefore, this is assigned with 1 point as long as indicator 1-2 is concerned.

Figure 9: Mean values of the results obtained using the prEN 13396 standard. Concerning the second key-parameter (values of accuracy), The averaged coefficients of reproducibility and repeatability for every method are given in table 14. For the sake of clarity, the mean values are also depicted in figure 10.

mix 1 mix2 mix 3 mix 4 averagesteady-state average 0,937 2,577 0,942 0,661

M3 6,431 -28,296 10,457 50,213 24M6-Ds 23,838 12,709 7,171 -31,492 19

R1 -30,269 15,587 -17,628 -18,721 21non-steady-state average 3,252 17,077 7,006 2,592

D2-D -37,376 3,204 6,091 0,305 12M4 -19,677 -7,229 -22,924 -3,304 13

M6-Dns 57,053 4,025 16,833 2,999 20

bias (%)

bias (%)

0,0

0,1

0,2

0,3

0,4

0,5


% C

l

prEN-28d-2prEN-90d-2prEN-180-2

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 24 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Figure 10: Coefficients of variations, of reproducibility and repeatability, mean for every

method. From figure 10 it can be deduced than in general, the accuracy of these methods is not very good, mainly as long as reproducibility conditions are concerned. As expected, the values for sr/m (%) are much higher than sR/m (%). Concerning the coefficient of variation for repeatability values, the highest one is found for the diffusion coefficient using the method D2, that reaches almost a 30%, being the lowest one for the measurement of the resistivity of concrete with a value of sr/m smaller than a 10%. For the coefficient of variation for reproducibility values, the highest values can be found when calculating the steady state diffusion coefficients in migration cells (M3 and M6-De). In fact, in table 14 it can be observed that one of the laboratories (L4) has been rejected when applying the M3 method for three mixes. In addition, it has also been rejected for mixes 1 and 3, and was an straggler for mix 2 when determining the steady-state coefficient by M6. This has been attributed to the fact that these two methods (M3 and M6) involve a higher degree of “chemical” handling than the other methods. Therefore, provided that the repeatability values are quite good, it seems to be necessary a more detailed explanation in the recipes of these two methods as long as procedures (taken of samples, calibration of conductivimeter, measurement of effective potential, …) and calculations are concerned. According to the indicators selected and given in table 15, the score assigned to each method is as follows: For the indicator 2-1 based on the repeatability:

• r/m (%) > 25% : D2-D 1 point • 10< r/m (%) < 25% : D2-K, M3, M4, M6, PrEN 2 points • 10> r/m (%): R1 3 points

0,0

20,0

40,0

60,0

80,0

100,0

D2-D D2-K R1 M3 M4 M6-De M6-Da Pren-28-2 Pren-90-2 Pren-180-2

sr/m

, Sr/m

( %

)

% sr/m% sR/m

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 25 of 28 WP2 REPORT – Pre-Evaluation of Test Methods For the indicator 2-1 based on the reproducibility:

• r/m (%) > 60% : M3, M6-Ds 1 point • 30< r/m (%) < 60% : D2-D, M6-Dns, PrEN 2 points • 30> r/m (%): D2-K, R1, M4 3 points

Third key-point: Relevance of the information that each method provides Concerning this point, the first indicator (3-1) refers to the diffusion coefficients Ds and Dns. On one hand, the prEN 13396 standard, only provides specific values of concentration at definite points, not giving any parameter able to be used to make predictions. Therefore, it is given 1 point. On the other hand there are several methods that provides a single diffusion coefficient. They are: D2, R1, M3 and M4. D2 and M4 give the value of the non-steady state and R1 and M3 give the steady state diffusion coefficient. Therefore, they are given 2 points. Lastly, the M6 method provides simultaneously the Ds and Dns coefficients and the ratio between them addresses the binding ability of the matrix, necessary for using these parameters in service life predictions. Therefore, this method has 3 points for indicator 3-1. With respect to the indicator 3-2: Acceleration, prEN accelerates diffusion by temperature, so, it is given 1 point. The rest of them (except R1 that are given 3 points), accelerates by electrical field and/or high increase of concentration. So, they have 2 points. The last point to be taken into account is the convenience concerning hours-man and resources. At this respect, the assignation of methods according to the indicators is as follows: Indicator 4-1: Handling after the test: Need of milling, grinding or cutting, splitting. The most costly methods are those that need milling of the concrete specimens as they need a lot of hours-man. The methods that needs this kind of techniques and therefore they have the score of 1 point are the D2 and the prEN 13396 standard The M4 method needs immediate splitting of the solid sample into two halves: Therefore, it is assigned 2 points. The rest of methods: R1, M3 and M6 are given 3 points. Indicator 4-2: Need of chloride analysis

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 26 of 28 WP2 REPORT – Pre-Evaluation of Test Methods After milling the samples in methods D2 and the prEN 13396 standard, it is necessary to analyse the solid samples, that is quite expensive. Therefore, they have 1 point as long as this indicator is concerned. M3 needs analysis of liquid samples and the M4 method applies a colourimetric technique of analysis. Therefore, they are assigned 2 points. The rest of methods: R1 and M6 do not have analysis, so, they are given 3 points. Indicator 4-3: Difficulty of handling The M3 and M6 method involve mounting and preparing a migration cell and also some chemical handling. So, they are given 1 point. The method M4 involves montage of the cell but not chemical handling. So, 2 points for this indicator. D2, R1 and the prEN 13396 standard are the most easy to handle, therefore, they have 3 points. Indicator 4-4: Time consumption The D2 and the prEN 13396 standard are natural diffusion methods, and therefore time consuming. So, 1 point in this indicator. The M3, M4 and M6 methods are migration methods that involve from more than 1 day to about 10-12 days. Therefore, they are given 2 points. The R1 method is almost instantaneous. Therefore, it is given 3 points. As a summary, the classification of the different methods according to these four different points is given in table 16, where the points per indicator for each method are given. In addition, the sum of the points for indicator have been made, and, as explained previously, in order to balance the weight of the key-points with less indicators, a score per parameter have been given (max. 1 for each key-parameter). The sum of them, for the four parameters give the classification of the methods (last row).

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 27 of 28 WP2 REPORT – Pre-Evaluation of Test Methods Table 16: Application of the analysis of the methods by key-parameters and indicators

according to the criteria of table 15 D2 R1 M3 M4 M6 prENParameters Indicators

1-1 -------- -------- -------- -------- -------- -------- 1 Absolute value obtained 1-2 2 2 2 2 2 1

Sub-total 2 2 2 2 2 1 Score/param 0.66 0.66 0.66 0.66 0.66 0.33

2-1 1 (D) 2 (K)

3 2 2 2 2 2 Accuracy

2-2 2 (D) 3 (K)

3 1 3 1 (Ds) 2 (Dns)

2

Sub-total 3 (D) 5 (K)

6 3 5 3 (Ds) 4 (Dns)

4

Score/param 0.5 (D) 0.83 (K)

1 0.5 0.83 0.5 (Ds) 0.66 (Dns)

0.66

3-1 2 2 2 2 3 1 3 Relevance of the information supplied 3-2 2 3 2 2 2 1

Sub-total 4 5 4 4 5 2 Score/param 0.66 0.83 0.66 0.66 0.83 0.33

4-1 1 3 3 2 3 1 4-2 1 3 2 2 3 1 4-3 3 3 1 2 1 3

4 Convenience

4-4 1 3 2 2 2 1 Sub-total 6 12 8 8 9 6 Score/param 0.5 1 0.66 0.66 0.75 0.5 Total points (max. 4) 2.32 (D)

2.65 (K)

3.49 2.48 2.81 2.74 (Ds) 2.90 (Dns)

1.82

8 CONCLUDING REMARKS AND SUGGESTIONS According to these results, four methods are revealed as being promising to be retained for WP5: R1, that has obtained the highest score, followed by M4 and M6 and D2. Even though it is clear that this classification is not a quantitative one, the obtained results are quite consistent with the ranking that could have been done by making a qualitative assessment of the results. In addition, the methods selected are very complementary, as one of them informs on steady state (R1), two of them on the non-stationary period (M4 and D2) and M6 makes a conjunction of both.


REFERENCES [1] NT BUILD 443: 94, “Concrete, Hardened: Accelerated Chloride Penetration”; [2] ANDRADE C., CASTELLOTE M., ALONSO C., GONZALEZ C., Non-steady-state

chloride diffusion coefficients obtained from migration and natural diffusion tests. Part I: Comparison between several methods of calculation, Materials and structures, vol. 33, jan.-feb. 2000, pp 21-28.

[3] ANDRADE C., ALONSO C., ARTEAGA A., TANNER P., Methodology based on the

electrical resistivity for the calculation of reinforcement service life, Proceedings of the 5th CANMET/ACI International Conference on Durability of Concrete, June 4-9, 2000, Barcelona, Spain, (Ed. by V.M. Malhotra, ACI, 2000), Supplementary paper, pp 899-915.

[4] TRUC, O., OLLIVIER, J.P., CARCASSÈS, M, A new way for determining the

chloride diffusion coefficient in concrete from steady state migration test, Cement and Concrete Research, Volume 30, Issue 2, February 2000, Pages 217-226.

[5] NT BUILD 492: 99, “Concrete, Mortar and Cement Based Repair Materials: Chloride

Migration Coefficient from Non-steady State Migration Experiments”. [6] CASTELLOTE M., ANDRADE C., ALONSO C., Measurement of the steady and non-

steady state chloride diffusion coefficients in a migration test by means of monitoring the conductivity in the anolyte chamber. Comparison with natural diffusion tests, Cement and concrete research, vol. 31, nº 10, 2001, pp 1411-1420.

[7] prEN-13396-1: 2002 “Products and systems for the protection and repair of concrete

structures”. Tests methods. Part I: Measurement of Chloride Ion Ingress by diffusion of repair mortars and concretes”

[8] International Standard ISO 5725-2:1994 (E). Accuracy (trueness and precision) of

measurement methods and results. Part 2: Basic method for the determination of the repeatability and reproducibility of a standard measurement method.

WP3 REPORT – COLLECTION OF IN-FIELD DATA



ACRONYM: CHLORTEST




EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 29 WP3 REPORT – Collection of In-Field Data

















ACKNOWLEDGEMENT: The present document is deliverables of Workpackage 3 – “Collection of In-Field Data”. The consortium members TNO, LNEC, IBRI, EDF, Chalmers, QUB, SP, IETcc, INSA, HSB, ZAG and LCPC were involved in the work of this part of the project. The work was led by TNO, assisted by LNEC, IBRI and EDF.

This document was prepared by Manuela Salta (LNEC) and Rob Polder (TNO)

FURTHER INFORMATION: Regarding this document: Dr Rob Polder Netherlands Organisation for Applied Scientific Research - TNO Schoemakerstraat, 97 NL-2600 AA DELFT, The Netherlands Tel. +31-15 276 3222; Fax: +31 15 276 30 18 e-mail: [email protected]



TABLE OF CONTENTS

1 INTRODUCTION 5


3 DESCRIPTION AND CHARACTERISATION OF DATA COLLECTED 6

4 SELECTION OF DATA FOR MODELS EVALUATION 17

5 FINAL REMARKS 24

6 REFERENCES 24


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 5 of 29 WP3 REPORT – Collection of In-Field Data 1 INTRODUCTION According to the Work-plan of the CHLORTEST project, the objectives to be reached by WP3 – Collection of In-Field Data are:

• Collection of in-field performance data from reinforced concrete structures in marine and road environment and other type of structures

• Characterisation of environmental conditions, • Characterisation of concrete composition

To reach these objectives a previous agreement concerning the interpretation and expression of data was made and it was decided to prepare data sheets to be filled with data by partners. Partners agreed to contribute data from national projects. The work developed in this WP was in interrelation with the WP1 and WP4. 2 PARTICIPATING LABORATORIES The laboratories that have participated in this WP giving data collected from structures or concrete specimens from research studies are listed in Table 1. All of the laboratories are in WP3 according to the consortium agreement, except LCPC, who voluntarily participated in this WP. Table 1- Laboratories that participated in data collection Partner Country EDF France IBRI Iceland IETcc Spain HSB German LCPC France LNEC Portugal CTH Sweden QUB United Kingdom SP Sweden TNO Netherlands UoA Spain ZAG Slovenia

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 29 WP3 REPORT – Collection of In-Field Data 3 DESCRIPTION AND CHARACTERISATION OF DATA

COLLECTED After agreement on the type of information necessary, excel data sheets were formulated upon a proposal by SP and sent by WP3 leader to the participants to be filled. Three types of data sheets were created: one with concrete composition and properties, another with environmental exposure characterisation and the third with chloride profiles and data concerning concrete durability related properties tested in the lab. Annex 1 presents these excel sheets. It was decided to collect three types of data taking into account the environmental conditions and the exposure classes specified in the Standard EN 206. Three task groups were created, one for marine structures, another for road structures and the third for other type of structures. The three task groups have collected data covering almost the full range of marine and road environments in EU. For marine and road environment, some of the data were collected from existing structures (bridges, wharves, …) and other from laboratory specimens exposed in natural environmental conditions, from national research projects carried out by the laboratories involved. Another group of data were collected from other structures, car parks, sewage pipes transporting seawater and precast concrete columns for electricity lines. Some other data were also included concerning results obtained from research studies in simulated standard conditions in the lab with similar concrete compositions that have been exposed in marine conditions. In the third project meeting in May 2003 the attendants agreed that chloride profiles would be considered valid for verification of modelling if they contain the following information:

1- Concrete mixture proportions; 2- Laboratory test methods and results for diffusivity; 3- Exposure conditions; 4- Chloride profiles sampled under the same conditions at preferably three concrete

ages: one or two ages younger than two years (for estimating the age effect) and one or two at ages older than ten years (for verification of the long term prediction).

In the 4th meeting in October 2004 the problem of chloride profiles validation was reanalysed and again it was reported that sufficient information should be available about concrete, environment and at least two profiles for two exposure periods. More than seven hundred profiles were collected, but only few of them fulfilled all the conditions to be considered valid according to the criteria fixed on meeting of May 2003. Only SP's profiles fulfilled all the first requirements. However, having in consideration the less restricted conditions reported on meeting of October 2004, various profiles from TNO, QUB, LNEC and IETcc can be also considered valid. It was decided to consider all the data collected, hoping that some other correlations can be later made, for example from similar concrete conditions exposed in different regions of Europe.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 29 WP3 REPORT – Collection of In-Field Data Table 2 summarises the number of chloride profiles collected by each exposure class according to EN 206. As a survey of the data collected, a summary for the different environments, both natural and in the laboratory, is presented in Tables 3 to 10. In each table the most relevant information is presented concerning the exposure conditions and exposure class, the concrete (type and content of cement, type and content of additions, w/b, compressive strength and the chloride profiles (number of profiles and exposure time). From these tables it is possible to evaluate the data with respect to the validation criteria. Table 2 - Number of chloride profiles collected from structures and laboratory concrete

specimens in several exposure classes. Exposure Class by EN 206

Structures Lab concrete specimens in natural exposure

Lab specimens in simulated conditions in laboratory

XS1 37 105 - XS2 15 58 8 XS3 72 238 - XD3 194 65 - Total 318 466 8

Some of the laboratories gave additional information concerning the data collected in each case, including specific information about the structures or test concrete specimens (geometrical characteristic) and exposure conditions and in a few cases about the corrosion state. This information was given as an additional small report and in some cases, where the data are obtained from national research projects, a reference was given to the published papers or reports. These references are reported in the bibliographic references. In some cases, only internal reports within the national research projects exist; this is indicated by "Rep". In Tables 3 to 5 the most relevant characteristics of the data collected are presented from marine structures, road structures and other structures. Tables 6 to 9 summarise the most relevant information concerning the collected data from lab specimens exposed in several natural environments. In Table 10 data from simulated conditions in the lab are reported. The collected data contain a large number of profiles, with a reasonably to well documented background from rather young to mature concrete exposed under different conditions (with exposure time between 0.5 and 42 years). Performance data for concrete with different compositions, various types of cement, different additions - fly ash, silica fume and blast furnace slag, in different exposure classes are available in the data collected. Chloride profiles have a number of data point variable from 4 to more than 10 and for some cases duplicate profiles exist for the same exposure time. The chloride analyses were made using potentiometric titration or volumetric titration (Volhard Method) In fact, some of the collected data enables to compare the performance of the same concrete in different marine environmental conditions and also in de-icing salt environment. The effect of the presence of additions on the concrete performance can also be evaluated for different exposure conditions.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 29 WP3 REPORT – Collection of In-Field Data Table 3 - Data collected from structures in marine environment [a) and b)] a)

Type of structure Bridge piers (Haringvielt)

Quay Wall(Rotterdam) Bridge Wall breakwater

Class XS3 XS1 XS3/Tidal XS3 Splash Zone

XS1 XS3

Exp

osur

e

Environmental Data

North Sea water Cl-=18 g/L

3 levels, + 9 and 14 m above sea level Cl-=18 g/L

Water Cl-=18 g/L Air T= 1ºC to 6ºC

Air T= 2º to 38ºC

Med. Sea

Type Cem III B 32.5

Cem III/B 32.5 Cem III/B 42.5 Cem IIA-

M 42.5 R - - Cement

Content, (kg/m3) Not def Not def 320 310 Not

def. 349 - -

Type - - - Blast furnace - -

Addition Content,% - - - 40 - -

Admixture - - - - - - Aggregate Type gravel gravel gravel basalt basalt basalt w/b or w/c 0.45 0.45 0.45 0.34 - -

Con

cret

e

Fcm, MPa (age) 60 (42 y) 60 (42 y) 67

(19y) 66 (30y) 71(35y) 80

18 (design value)

18 (design value)

ExposureTime (year) 42 42 30 19 35 5 13 14 Number of profiles 6 18 6 6 6 1 1 1 Lab/Additional Information TNO / Rep [12,13] TNO / Rep [13]

IBRI/Rep UoA

b)

Type of structure Quay wall (Nordland) Storm surge barrier (Scheldt)

Class XS3/Splash XS3/Splash XS2/ Subm XS3/Tidal

Exp

osu

re Environmental Data

Cl-=19 g/L North Sea Cl-=18 g/L

Type Cem III/B 42.5 Cem III/B 42.5

Cement Content (Kg/m3) 300 325 400 350 350

Type - - Addition Content,% - - Admixture - -

Aggregate Type gravel Gravel Light weight gravel

w/b or w/c 0.54 0.45

Con

cret

e

fcm, MPa 75 (8y) 61-75 (18-22y) ExposureTime (year) 8 18 18-22 Number of profiles 4 3 27 21 2 6

Lab/Additional Information

TNO [9,10] /ASTM, Resist.

TNO Rep [11,12,13] / NT Build 492

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 29 WP3 REPORT – Collection of In-Field Data Table 4 - Data collected from road structures

Type of structure

Road (beams, slabs and columns)

Road (beams and columns)

Road Road

Class XD3 XD3 XD3 XD3

Exp

osur

e

Environmental Data -

T= -1.7º to 16 ºC RH=81% Vertical/ Top

NaCl 3000g/L NaCl

Type Cem II 42.5 Cem I Cem I 42.5

Cem I 52.5 Cement Content

(kg/m3) 360 300-360 380 230 410 189 195 325 223

Type - - - - - Fly ash Addition Content

(%) - - - - - 21 20 20 30

Admixture - - - -

Aggregate Type Crushed limestone - basalt Limestone

w/b or w/c 0.55 0.45-0.50 0.43 0.84 0.48 0.67 0.67 0.45 0.52

Con

cret

e

fcm, MPa - - 45 26/30 55 22 23 52 43 ExposureTime,(year) 18 25 to 30 Nd 3 3 3 3 3 3

Number of profiles 28 12 (beams); 50 (columns) 1 2 1 1 1 1 1

Lab/Additional Information ZAG CTH / Rep [2]

IBRI LCPC /[7,8] (Dns Mig. Test)

– Corrosion state, cover, carbonation depth, surface treatment (impregnation); NaCl 12000 kg/km. Table 5 - Data collected from other type of structures

Type of structure Car Park Car Park Pipes (precast)

Columns (precast)

Class XD 3 XD3 XS2 XS1/XD1

Exp

osu

re

Environmental Data T= 7º C to 18º C T= 7ºC to 18º C

Cl-=20 g/L t= 9ºC -18ºC

Distance to Sea > 100 m

Type Cem II/A-M 42.5 R Cem III 32.5 Cem I CPA ES

Cem I CPA Cement

Content(kg/m3) - - 490 400 Type Silica Fume - - - Addition Content (%) 7.5 - - -

Admixture - - - - Aggregate Type basalt basalt - - w/b or w/c - - 0.55 0.47

Con

cret

e

fcm (Mpa) - - - - ExposureTime (year) 27 11 14 27 to 62

Number of profiles 11 5 10 (T+F) 18


IBRI / Rep 3 compositions with silano (3y of exposure)

IBRI/ Rep EDF/Rep EDF/Rep


Table 6 - Data collected from lab concrete specimens in marine environment-Class XS3 [a) to g)]

a)

Type of specimens Lab specimens (0.1x0.1x0.4)m

Lab specimens (0.1x0.1x0.4)m

Lab specimens (1x0.5x0.12)m


Zone Tidal Tidal Splash Tidal

Exp

osur

e

Environmental Data Cl- = 20.3 g/L SO4

=2.8 g/L pH=8.1

Cl- = 20.3 g/L SO4

=2.8 g/L pH=8.1

Cl- = 19 g/L Cl- = 19 g/L

Type Cem IV 32.5 R Cem IV 32.5 R Cem I 32.5 R Cem I 32.5R Cement Content

(kg/m3) 260 340 300 300

Type - - - - Addition Content

(%) - - - -

Admixture - - - - Aggregate Type Limestone Limestone Limestone Limestone w/b or w/c 0.65 0.45 0.50 0.50

Con

cret

e

Fcm, MPa 32 53 34 34 Exposure Time (years) 0.5 to 2 0.5 to 2 1 to 4 1 to 4 Number of profiles 6 6 4 7

Lab/Additional Information LNEC / Rep [5] NT Build 492

LNEC / Rep [5] NT Build 492

LNEC / Rep [6] Diffusion Cell

LNEC / Rep [6] Diffusion Cell

b)

Type of specimens Panels (1.4x1.35x0.15)m Blocks

Zone Tidal Tidal

Exp

osur

e

Environmental Data North Sea Water Cl- = 20 g/L

Type Cem I Cem III/B LH HS OPC SRPC Cement Content

(kg/m3) 350 350 200 and 350 200 and 350

Type - - - Addition Content (%) - - - Admixture Air entrain; plasticizer - - Aggregate Type - - - w/b or w/c 0.48 0.5 0.4

Con

cret

e

fcm (Mpa) 45 35 - -

ExposureTime (year) 10 10 2 periods ( not clear )

2 periods (not clear )

Number of profiles 2 2 55 56 Lab/Additional Information HSB/ Rep HSB/Rep IETcc

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 29 WP3 REPORT – Collection of In-Field Data c)

Type of specimens Cubes (0.1x0.1x0.1)m Not available

Zone Tidal / Splash Tidal

Exp

osur

e

Environmental Data Cl- = 21 g/L T=1 to 20ºC

Atlantic Ocean (La Rochelle)

Type OPC OPC Cem I 52.5 PM ES CP2 Cement Content

(kg/m3) 460 460 230 195 223 325 377 470 410 461 360

Type - - - Flyash Flyash Flyash Sílica Sílica - - Sílica Addition Content

(%) - - 20 30 20 9 11 - - 6

Admixture - - Melamine Aggregate Type Screened Limestone/calcareous w/b or w/c 0.4 0.4 0.84 0.77 0.52 0.45 0.30 0.23 0.48 0.32 0.36

Con

cret

e

fcm (Mpa) 66 30 30 23 52 124 133 0.55 0.80 91 ExposureTime (year) 1 to 7 2 and 5 Number of profiles 38 38 1 1 1 1 1 1 2 6 6 Lab/Additional Information QUB/Rep LCPC/ Rep [7,8]

d)

Type of specimens Lab specimens

Zone Splash

Exp

osur

e

Environmental Data Water Cl-=14g /L

Type CEM I 42.5 N Cement Content

(kg/m3) 240 250.5 309.6 369.3 391.7 421.3 421.9 437.7 452.9 455.8 505.1 503.4

Type - Addition Content

(%) -

Admixture Plasticizer Aggregate Typee Granitic w/b or w/c 0.35

Con

cret

e

fcm (MPa) 21 26 35 41 42 58 54 58 70 60 102 96

ExposureTime (year) 0.5 1 1

0.5; 2 and 5

0.5 0.5 and 10

1,2 and 5

2 and 5 0.5 1 1 1 and

5

Number of profiles 1 1 1 3 1 3 4 2 1 1 1 1 Lab/Additional Information

SP / Rep [1] NT Build 492

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 29 WP3 REPORT – Collection of In-Field Data e) Type of specimens Lab specimens

Zone Splash

Exp

osur

e

Environmental Data

Water Cl-=14g /L

Type CEM I 42.5 N Cement Content (kg/m3)

233.9 352 376 450 400 479 502 530 502 530 348 384

Type Microsilica Flyash Microsilica+ Flyash

Microsilica+ Flyash

Addition

Content (%)

3 5 10 10 5 5 5 5 20 5 4 + 14 5 + 10

Admixture Plasticizer Aggregate Type Granitic w/b or w/c 0.35

Con

cret

e

fcm(MPa) 21 80 65 117 63 112 125 117 98 95 69 84 ExposureTime (year) 0.5 0.5;

2 and 5

0.5 1 and 5

0.5; 2; 5 and 10

1 and 5

1 1 and 5

1 and 5

1 0.5; 2 and 5 0.5; 2 and 5

Number of Profile 1 3 1 2 6 3 1 2 2 1 3 4 Lab/Additional Information


Table 7 - Data collected from lab specimens in marine environment- Class XS2 (a and b) a)

Type of specimens Lab specimens ( 20 x 10) cm

Zone XS2

Exp

osur

e

Environmental Data T= 2ºC - 15ºC

Type CEM II 52.5R Cement Content

(kg/m3) -

Type - Sílica fume Silica fume, blast furnace Blast furnace Sílica fume + Fly ash Addition

Content (%) - 3,7 and 10 7, 30 7 7 + 10,20,40 Admixture - - - - - Aggregate Type basalt w/b or w/c 0.30 – 0.42

Con

cret

e

fcm (MPa) - - - - - Exposure Time (year) 1 Number of profiles 2 2; 6 and 2 2 2 2 Lab/Additional Information IBRI / Rep

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 29 WP3 REPORT – Collection of In-Field Data b)


Class Submerged

Exp

osur

e

Environmental Data Twelve sites Cl-=5 – 23 g /L

Type Cem I 42.5 Cement Content

(kg/m3) 450


(%) -

Aggregate Type - Admixture - w/b or w/c 0.4

Con

cret

e

fcm (MPa) - Exposure Time (year) 1 Profile number 5 Lab/Additional Information

CTH /Rep [3]

c)


Zone Submerged

Exp

osur

e

Environmental Data Water Cl-=14g /L


(kg/m3) 240 250.5 309.6 369.3 391.7 421.3 421.9 437.7 452.9 455.8 505.1 503.4


(%) -


Con

cret

e

fcm (MPa) 21 26 35 41 42 58 54 58 70 60 102 96

ExposureTime (year) 0.5 and 1

1 and 10

1 and 10

0.5, 1,2 and 5

0.5; 1 and 10

0.5 and 10

1;2;5 and 10

1,2 and 5

0.5 and 10

1 and 10

1 and 10

1,2,5 and 10

Number of profiles 2 2 2 4 3 2 4 3 2 2 2 4 Lab/Additional Information


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 29 WP3 REPORT – Collection of In-Field Data d)


Zone Submmerged

Exp

osur

e

Environmental Data Cl-=14g /L


(kg/m3) 233.9 450 376.3 400 479.3 450 530 502 530 348 384


Microsilica+ Flyash Addition Content

(%) 3 10 10 5 5 5 5 20 5 4 + 14 5 + 10


Con

cret

e

fcm (MPa) 21 117 65 63 112 125 117 98 95 69 73

Exposure Time (year) 0.5 and 1

2 and 5

0.5 and 10

0.5;1; 2 ; 5 and 10

0.5;1;2;5 and 10

0.5;2;5 and 10

1; 2; 5 and 10

1; 2; 5 and 10

1 and 10

0.5; 2;5 and 10

0.5;2; 5 and 10

Profile number 2 2 2 10 6 7 4 4 2 4 6 Lab/Additional Information


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 29 WP3 REPORT – Collection of In-Field Data Table 8 - Data collected from lab concrete specimens in marine environment- Class XS1 [a) to c)]

a)


Zone Atmospheric

Exp

osur

e



(kg/m3) 240 250.5 309.6 369.3 391.7 421.3 421.9 437.7 452.9 455.8 505.1 503.4


(%) -

Admixture Plasticizer Aggregate Time Granitic w/b or w/c 0.35

Con

cret

e

fcm (MPa) 21 26 35 41 42 58 54 58 70 60 102 96

Exposure Time (year) 0.5 1 and 10

1 and 10

0.5, 2 and 5

0.5 and 10

0.5; 1; 5 and 10

2 and 10

2 , 5 and 10

0.5 and 10

1 and 10

1 and 10

1,5 and 10

Number of profiles 1 2 2 3 4 4 2 3 5 5 2 3



b)


Zone Atmospheric

Exp

osur

e



(kg/m3) 233.9 286 376.3 400 479.3 450 450 502 530 348 384


Microsilica+ Flyash Addition Content

(%) 3 5 10 5 5 5 10 20 5 4 + 14 5 + 10


Con

cret

e

fcm (MPa) 21 45 65 63 112 125 117 98 95 69 84

Exposure Time (year) 0.5

0.5, 2 and 5

0.5 , 2 and 10

0.5, 2, 5 and 10

1 , 5 and 10

1,5 and 10

1,5 and 10

1,5 and 10

1 and 10

0.5, 2,5 and 10

0.5; 2; 5 and 10

Number of profiles 1 3 6 10 9 3 10 3 2 6 6 Lab/Additional Information


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 29 WP3 REPORT – Collection of In-Field Data c)

Type of specimens Lab specimens (0.1x0.1x0.4)m

Lab specimens (0.1x0.1x0.4)m


Zone Atmospheric 100 m sea

Atmospheric 100 m sea

Atmospheric 200 m sea

Exp

osur

e

Environmental Data Air 1215 mgCl-/m2d 4-20ºC 77% HR

Air 1215 mgCl-/m2d 4-20ºC 77% HR

Air 68 mgCl-/m2d 12-40ºC 78% HR

Type Cem IV 32.5 R Cem IV 32.5 R Cem I 32.5 R Cement Content

(kg/m3) 260 260 300

Type - - - Addition Content (%) - - - Admixture - - - Aggregate Time Limestone Limestone Limestone w/b or w/c 0.65 0.45 0.50

Con

cret

e

fcm (MPa) 32 52.5 34 Exposure Time (year) 0.5 to 2 0.5 to 2 1 to 4 Number of profiles 12 12 5

Lab/Additional Information LNEC/ Rep[5] NT Built 492 carb. depth(1 y)

LNEC/ Rep[5] NT Built 492 carb. depth(1 y)

LNEC/ Rep[6] Diffusion Cell carb. depth

Table 9 - Data collected from lab concrete specimens in de-icing salt environment- class XD3


Class Vertical or Top Surfaces

Exp

osur

e

Environmental Data NaCl 10000 Kg/Km

Type Cem I 42.5 Cement Content

(kg/m3) 420 399 260

Type - Sílica fume - Addition Content (%) - 5 - Admixture - - - Aggregate type w/b or w/c 0.40 0.40 0.75

Con

cret

e

fcm (Mpa) 65 78 30 ExposureTime 0.5 , 1 , 2 to 4 number of profiles 65 Lab/Additional Information CTH/ Rep[2]

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 29 WP3 REPORT – Collection of In-Field Data Table 10 - Data collected from lab concrete specimens in simulated accelerated conditions in

the laboratory

Type of specimens Cylinders (0.1x0.5m) Lab specimens

Class NaCl

Exp

osur

e

Environmental Data

NaCl solution Cl- = 10 x sea water 20g/L

7ºC 20ºC

5g/L 7ºC 20ºC

Type Cem II AM 42.5R Cem III 32.5 Cem I 42.5 R

Cement Content (kg/m3) 411 385 411 385 450

Type Sílica fume Blast furnace - Addition Content (%) 7.5 40 - Admixture Plasticizer - Aggregate type Basalt - w/b or w/c 0.40 0.45 0.40 0.45 0.40

Con

cret

e

fcm (Mpa) 57 53 48 46 - ExposureTime (year) 2.5 1 Number of profiles 2 2 4 Lab/Additional Information IBRI/ Rep - CTH/ Rep

4 SELECTION OF DATA FOR MODELS EVALUATION Data collected from tables 3 to 10 (grey cells) were assorted and some of them assembled in three series by WP4 for model evaluation and benchmarking. Series 1 – Thirteen chloride profiles, from tables 6 a), c) , d), 7 c) , d) , 8a), b) and 9, were selected and assembled in this series. All the cases have the same type of cement (CEM I). Different sub-groups can be organized assembling the same concrete composition and different exposure conditions. For each case one chloride profile is given and data from short term tests. Series 2 - This contains four cases selected from cases in Table 3a). The data selected correspond to the same type of concrete in two exposure conditions - XS3 and three levels in XS1. For all 4 cases one chloride profile with exposure time between 6 and 8 years and NT Build 492 at the same age is given. Series 3 - This contains five cases with four types of concrete, selected from Table 3b) and Table 6b). These four cases correspond to similar exposure conditions (XS3) and one case in XS2. For three cases the chloride profiles obtained after a long exposure time (42 year) is given and also results from additional tests. For the other two cases only concrete resistivity data are given. In Figures 1 to 9 the different chloride profiles selected as Series 1 with similar periods of exposure are represented. The Figures correspond to the profiles assembled in different sub-

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 29 WP3 REPORT – Collection of In-Field Data groups to show the influence of concrete composition and of the exposure conditions on the concrete performance under chloride load. Figures 1 and 2 present two profiles selected in this series 1 showing the efect of exposure conditions with the same concrete composition with and without silica fume, respectively. Figure 3 shows two profiles from a concrete mix (CEM I) in XS3 exposure conditions (splash and tidal) Figure 6 show also two profiles with the same concrete mix (CEM I with silica fume or fly ash) in the same exposure XS2. Figures 5, 6 and 9 each represent two profiles with the same concrete mix (CEM I with and without addition of silica fume) in XS2, XS1 and XS3 conditions, respectively. From this Figure it is possible to see the increase of performance due to the presence of silica fume in the three environmental conditions. Figures 7 and 8 present two chloride profiles with the same concrete composition in each one with (CEM I with silica fume) and (CEM I), respectivly, in XD3 exposure conditions for a vertical face and horizontal top face. From these two Figures it is possible to see that the influence of the concrete surface orientation (vertical or horizontal face) is not relevant for concrete with silica fume, at least for these short periods of exposure. Figure 10 presents the four chloride profiles selected in Series 2 and shows the influence of exposure class and location conditions in XS1 for the same type of concrete (CEM III). Figure 11 presents the profiles selected in Series 3 and shows the influence of the type of cement on concrete performance under the same exposure conditions XS3.

0

0,5

1

1,5

2

2,5

0 10 20 30 40

x, mm

C, C

l % b

inde

r

XS2 XS3 XS1

Figure 1 – Three chloride profiles with the same concrete composition (CEM I at the same

content) after the same exposure time in different conditions XS3, XS2 and XS1 (series 1).


00,20,40,60,8

11,21,41,61,8

0 10 20 30 40

x, mm

C, C

l % b

inde

r

XS1 XS2

Figure 2 – Two chloride profiles for the same concrete (with microsilica) in different

exposure conditions (series 1)

0

0,2

0,4

0,6

0,8

1

1,2

1,4

1,6

0 10 20 30 40

x, mm

C, C

l % b

inde

r

XS3/Splash XS3/Tidal

Figure 3 – Two chloride profiles for same concrete mix (CEM I) and the same exposure time

in XS3 splash and tidal locations (series 1).


0

0,5

1

1,5

2

2,5

0 10 20 30 40

x, mm

C, C

l % b

inde

r

Microsilica Fly ash

Figure 4 - Two chloride profiles for two concrete mixes (CEM I with silica fume and with

flyash) after the same exposure time in XS2 conditions (series 1).

0

0,5

1

1,5

2

2,5

0 10 20 30 40

x, mm

C, C

l % b

inde

r

CEM I_Microsilica CEM I

Figure 5 - Two chloride profiles for two concrete mixes (CEM I with and without addition of

silica fume) after the same exposure time in XS2 conditions (series 1).


00,10,20,30,40,50,60,70,80,9

1

0 5 10 15 20

x, mm

C, C

l % b

inde

r

CEM I_Microsilica CEM I

Figure 6 - Two chloride profiles for concrete mixes (CEM I with and without addition of

silica fume) after the same exposure time in XS1 conditions (series 1).

0

0,2

0,4

0,6

0,8

1

1,2

1,4

0 10 20 30 40

x, mm

C, C

l % b

inde

r

Vertical Face Horizontal Top

Figure 7 - Two chloride profiles from the same concrete composition (CEM I with silica

fume), after the same time in XD3 exposure conditions. Comparison between vertical face and horizontal top face (series 1).


00,10,20,30,40,50,60,70,80,9

1

0 10 20 30 40

x, mm

C, C

l % b

inde

r

Vertical Face Horizontal Top

Figure 8 - Two chloride profiles from the same concrete composition (CEM I without

additions), and the same exposure time in XD3 conditions. Results on vertical face and horizontal top face (series 1).

0

0,2

0,4

0,6

0,8

1

1,2

1,4

0 10 20 30 40

x, mm

C, C

l % b

inde

r

CEM I CEM I_Microsilica

Figure 9 - Two chloride profiles with different concrete compositions (with and without silica

fume) and the same exposure time in XS3 conditions (series 1).


.

0

0,4

0,8

1,2

1,6

2

0 10 20 30 40 50 60

x, mm

C, C

l % b

inde

r

XS3/Tidal XS1/9m XS1/14m XS1/14m

Figure 10 - Four chloride profiles with the same type of concrete and the same exposure time

in different exposure conditions (XS3 and XS1 at three levels) (series 2).

0,00

0,20

0,40

0,60

0,801,00

1,20

1,40

1,60

1,80

0 50 100 150 200

x,mm

C, C

l %bi

nder

OPC_350 SRPC_350 CEM III_300

Figure 11 - Chloride profiles from three concrete compositions, during the same exposure

time in similar XS3 conditions (series 3).

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 24 of 29 WP3 REPORT – Collection of In-Field Data 5 FINAL REMARKS Although not all the data collected fulfilled all the required criteria for validation in general, they include information from the performance of different concrete compositions (with the most important characterization parameters known) exposed in different environmental conditions representative for the different aggressiveness classes due to chlorides in different locations in Europe. Most of the cases include information about concrete performance during early periods of exposure, but some data also exist where information after very long exposure is available. Few data have information correlating the concrete performance under chloride load with the corrosion state of the reinforcement. Nevertheless, some of the data collected represent a very useful tool for model validation and other for correlation with new similar cases. The authors consider that it is important to spread this information into the scientific community interested in this field. As it was previously planned, a CD was made with all the information concerning the data collected. We recognize the relevance of having this data in a more easy and friendly way, for example creating an access database, to be used by all the scientific community by internet access. As a last comment it is also important to stimulate the different members of scientific community dealing with questions about service life to collect useful information related to the performance of concrete structures. It is important to collect not only the relevant data concerning the initiation period but also data about the corrosion state and corrosion propagation (corrosion rate) in different environmental conditions relevant for the propagation period. 6 REFERENCES [1] Tang, L. (2003), "Chloride Ingress in Concrete Exposed to Marine Environment – Field

data up to 10 years’ exposure", SP Report 2003:16, SP Swedish National Testing and Research Institute, Borås, Sweden

[2] Lindvall, A. (2001), Environmental Actions and Response – Reinforced Concrete Structures exposed in Road and Marine Environments, Publication P-01:3, Department of Building Materials, Chalmers University of Technology, Göteborg, 2001, 320 pp.

[3] Lindvall, A. (2003), Chloride ingress data from field exposure, at twelve different marine exposure locations, and laboratory exposure, Publication P-03:1, Department of Building Materials, Chalmers University of Technology, Göteborg, 2003, 53 pp.

[4] Lindvall, A. (2002), Chloride ingress in a Swedish road environment – Five years exposure for three concrete compositions, Publication P-02:4, Department of Building Materials, Chalmers University of Technology, Göteborg, 2002, 42 pp.

[5] Salta,M.; Melo, A.; Estudo da corrosão do betão armado em exposição marítima e urbana – Relatório de progresso aos 2 anos, Rel 129/04, Departamento de Materias, LNEC, Lisboa, 2004

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 25 of 29 WP3 REPORT – Collection of In-Field Data [6] Costa, A.; Durabilidade de estruturas de betão armado em ambiente marítimo, Tese de

douturamento, Universidade Técnica de Lisboa, IST, Lisboa, 1997

[7] Baroghel-Bouny V., Arnaud S., Henry D., Carcassès M., Quénard D., Ageing of concretes in natural environments: an experiment for the 21st century. iii - durability properties of concretes measured on laboratory specimens, Bulletin des Laboratoires des Ponts et Chaussées, n° 241, nov.-dec. 2002, pp 13-59 (available in French and in English)

[8] Baroghel-Bouny V., Gawsewitch J., Belin P., Ounoughi K., Arnaud S., Olivier G., Bissonnette B. Ageing of concretes in natural environments: an experiment for the 21st century. IV - Results on cores extracted from field-exposed test specimens of various sites at the first times of measurement, Bulletin des Laboratoires des Ponts et Chaussées, n° 249, march-april 2004, pp 49-100 (available in French and in English),

[9] Polder, R.B., Stoop, B.Th.J., Visser, J., 1997, Chloride profiles in cores from a quay wall, TNO Building and Construction report 97-BT-R0571/01, DuraCrete

[10] Polder, R.B., Walker, R., Page, C.L., 1995, Electrochemical Desalination of Cores from a Reinforced Concrete Coastal Structure, Magazine of Concrete Research, vol. 47, no. 173, 321-327

[11] Polder, R.B., Rooij, M. de, Vries, J. de, Gulikers, J., 2003, Observed Chloride Penetration in a marine concrete structure after 20 years in North Sea Environment, Workshop "Risk based maintenance of Structures", TU Delft, 21 January

[12] Rooij, M.R. de, Polder, R.B., 2003, Accuracy of chloride penetration predictions based on chloride profile analysis, Advances in Cement and Concrete, Proceedings of a conference held at Copper Mountain, Colorado, August 10-14, Engineering Conferences International, 79-88

[13] Polder, R.B., Rooij, M.R. de, 2005, Durability of marine concrete structures – field investigations and modelling, HERON, Vol. 50 (3), 133-143


ANNEX

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 27 of 29 WP3 REPORT – Collection of In-Field Data Sheet 1- CONCRETE CHARACTERISATION Sample_ID Mix_ID Designation_No Date_prod, yyyy-mm-dd Cement_type Cement_class Cement_content CEM_Blaine CEM_Loi CEM_SO3 CEM_Cl- CEM_C3S CEM_C2S CEM_C3A CEM_C4AF Watercontent w_c w_b k Agg_type Agg_origine D_max Agg_content Grading curve air entr_type air entr_content air_nature plast_type1 plast_nature1 plast_content1 plast_type2 plast_content2 plast_nature2 Entrapped air Entrained air target_air Gross Density flow_value slump_value flowdin_value Concrete Temp N_spec_c c_spec_geom c_gross_dens fcm_2 fcm_7 fcm_28 c_age_cur c_RH_cur

c_Temp_cur pre_cond_c design_c required_c day_c fcm_x N_spec_f f_spec_geom f_gross_dens fct_2 fct_7 fct_28 f_age_cur f_RH_cur f_temp_cur pre_cond_f design_f required_f day_f fct_x N_spec_Ym

Ym_spec_geom Stat/Dyn

Ym_2 Ym_7

Ym_28 Ym_age_cur Ym_RH_cur

Ym_temp_cur pre_cond_y

design_y required_y

day_y ym_x

Type I add_nature_1 Type I add_cont_1

Type I add_cont_1% Type I add_nature_2

Type I add_cont_2 Type I add_cont_2%

Type II add_nature_1 Type II add_cont_1

Type II add_cont_1% k1

Type II add_nature_2 Type II add_cont_2

Type II add_cont_2% k2

Na2Oeq_%Binder Ini_Cl_%Binder

SHEET 2 – EXPOSURE ENVIRONMENT

Exposure Environment of Concrete

Sample_ID Mix_ID

Designation_No Date_Exp, yyyy-mm-dd Exp class acc to EN 206

Type of exposure Marine environment Road environment

Exposure zone Type of road Extension of tidal actions, m Number of lanes Extension of splash zone, m Wideness of each lane, m

Elevation to seawater, m Speed limit, km/h Distance to sea, m Traffic density, vehicles/day

Mean Cl concentration, gCl/litre Percentage of heavy vehicles, % Std dev of Cl concentration Distance to the edge of lane, m

Elevation to seawater, m Metrological data Type of de-icer

Annual min temp, °C Annual Cl from deicer, kgCl/km Annual max temp, °C De-icing period, month

Max temp in month No. Max de-icing in month No. Annual precipitation, mm

Annual mean RH, % Other information Std dev of RH Attach 1

Exposure to rain Attach 2 Exposure to sun Attach 3

Effect of wind Attach 4 Typical speed of wind, m/s Attach 5


SHEET 3- CHLORIDE PROFILES Chloride Ingress in Concrete

Sample_ID Mix_ID

Designation_No Date_prod, yyyy-mm-dd Date_Exp, yyyy-mm-dd

Date_Samp, yyyy-mm-dd

Sample characteristics Structural part examined

Position of exposure surface Surface treatment Surface condition

Samp_size, mm Cover_thick, mm Crack_width, mm

Type_steel Corrosion_status Carb_depth, mm

Test methods

Profiling method Cl_analysis method

Binder_analysis method Lab_test method

Test surface Free Cl in test, gCl/litre

Laboratory test results Chloride profile

Test age, days

D x10-12 m2/s

Cs Cl% samp x, mm

C, Cl% binder

Other information Attach 1 Attach 2 Attach 3 Attach 4 Attach 5

WP4 REPORT – MODELLING OF CHLORIDE INGRESS



ACRONYM: CHLORTEST




EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 112 WP4 REPORT – Modelling of Chloride Ingress

















ACKNOWLEDGEMENT: The present document is deliverables of Workpackage 4 – “Modelling of Chloride Ingress”. The consortium members LTH, Chalmers, INSA, IETcc, SP and EDF were involved in the work of this part of the project. The work was led by LTH, assisted by INSA, IETcc, SP and EDF.

This document was prepared by Lars-Olof Nilsson (LTH)

FURTHER INFORMATION: Regarding this document: Prof Lars-Olof Nilsson Lund Institute of Technology Division of Building Materials S-221 00 LUND, Sweden Tel. +46-46 222 7408; Fax: +46-46 222 4427 e-mail: [email protected]



TABLE OF CONTENTS

Page

1 INTRODUCTION 5 1.1 Background 6 1.2 Objectives of WP4 and the work performed 7 1.3 Assessment criteria 8 1.4 The structure of the report 8

2 PHENOMENOLOGICAL DESCRIPTION OF THE CHLORIDE INGRESS PROCESS 9

2.1 Concrete, as an ingress media 9 2.2 The concrete pore solution 10 2.3 The exposure conditions 10 2.4 The ingress process 10

3 SOME FUNDAMENTAL PARTS OF INGRESS MODELS 12 3.1 Chloride concentrations 12 3.2 Fick’s 1st law of diffusion 14 3.3 Chloride binding/interaction, binding capacity 14 3.4 Mass balance equations 17 3.5 Fick’s 2nd law of diffusion 18 3.6 The error function, and its compliment 19 3.7 The apparent diffusion coefficient Da and apparent surface chloride content Csa 20 3.8 Time-dependency of diffusion coefficients 20 3.9 The Nernst-Planck equation for ion flux 27

4 TYPE OF CHLORIDE INGRESS MODELS 28 A. MODELS BASED ON FICK’S 2ND LAW. 28 B. MODELS BASED ON FLUX EQUATIONS 29

5 CHLORIDE INGRESS MODELS BASED ON FICK’S 2ND LAW 30 5.1 General 30 5.2 Boundary conditions concepts in empirical models 31 5.3 Models A1a: ERFC-model, constant Da & Csa 33 5.4 Models A1b: ERFC-model, Da(t) & constant Csa 35 5.5 Models A1c: ERFC-model, D(t) & Csa 38 5.6 Models A1d: ERFC-model, Da(t) & Csa(t) 41 5.7 Models A2: Analytical solutions, DF2(t) & Csa(t) 43 5.8 Models A3a and A3b: Numerical, DF2(t) & Csa(t) 46 5.9 Conclusions on models based on Fick’s 2nd law 48

6 CHLORIDE INGRESS MODELS BASED ON FLUX EQUATIONS 49 6.1 General 49 6.3 Boundary conditions concepts in physical models 50 6.3 Models B1a: Models based on Fick’s 1st law of diffusion, no convection 53


6.4 Models B1b: Models based on Fick’s 1st law of diffusion, with convection 56 6.5 Models B2a, b and c: Physical models based on the Nernst-Planck flux equation 59 6.6 Conclusions, models based on flux equations 63

7 SHORT TERM SENSITIVITY ANALYSIS 64 7.1 Introduction 64 7.2 Methodology for sensitivity analysis 64 7.3 Application of the methodology for ERF 69 7.4 Application of the methodology for LEO 72 7.5 Application of the methodology for MS-Diff 75 7.6 Conclusions 78

8 LONG-TERM SENSITIVITY OF ERFC-MODELS 79 8.1 Mathematical expressions 79 8.2 Sensitivity of various parameters for prediction of chloride concentration 80 8.3 Effect of different parameters on the sensitivity 84 8.4 Discussions 84 8.5 Combined uncertainty of models for prediction of chloride concentration 85 8.6 Conclusions 86

9 BENCHMARKING OF MODELS 87 9.1 Objectives and over-view of work performed 87 9.2 Establishment of criteria for benchmarking 87 9.3 Selection of profiles and documentation prepared for the benchmarking 89 9.4 Selection of models 92 9.5 Responses obtained 92 9.6 Comparison of results 94 9.7 Some examples of results 94 9.8 Analysis of all predictions 99 9.9 Final comments on the benchmarking of models 103

10 CONCLUSIONS 104

REFERENCES 107


1 INTRODUCTION Chloride ingress into reinforced concrete is one major process that is limiting the service-life of the structure. The ingress process, however, is not the only important part of service-life predictions. For instance, without any accurate chloride threshold levels for initiation of corrosion, chloride ingress is of little significance. In spite of this, this report only deals with the ingress part of the process. The ”state-of-the-art” in modelling chloride ingress into concrete was tested at a Nordic (& fib TG) seminar in May 2001, Nilsson (2001). A test case was defined with a given concrete, with given properties in a given environment. Exposure data up to 2 years in that environment was given. From that data it was easy to determine a suitable surface chloride content in that particular environment for the submerged zone. Significant scatter was found in predictions of further ingress, by the ”best” prediction models available, cf. Figure 1.1, even for the submerged zone. For the splash zone the scatter was fantastic. These results clearly show that we still have a long way to go before chloride ingress modelling is accurate enough for design applications.

H4

0

1

2

3

4

5

0 20 40 60 80 100Depth [mm]

Chl

orid

e [w

t-% o

f bin

der]

given0.6-2 y

SUB 100 years

H4

0

1

2

3

4

5

0 20 40 60 80 100

Depth [mm]

Chl

orid

e [w

t-% o

f bin

der] SPL 100 years

given0.6-2 y

Figure 1.1 Predicted chloride profiles from various prediction models for a given concrete in a well-defined environment and with profiles up to two years given. Prediction results for 100 years in the submerged (left) and splash zone, Nilsson (2001)

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 112 WP4 REPORT – Modelling of Chloride Ingress An analysis of the predictions shows that the differences may very well be explained by the background exposure data for each prediction model. That data is more or less individual for each model. Better data and better exchange of data would probably reduce the scatter significantly. That exercise remains to be done! A model for predicting chloride ingress into concrete always aims at predicting the chloride profile C(x,t) or at least the chloride content at the depth of the reinforcement. The output is always meant to be compared to a “critical chloride content that is relevant for reinforcement corrosion, see Figure 1.2. In recent models the prediction and the comparison should include some measure of the uncertainty in the prediction.

C(xc,t) ≥ Ccr ?

Cenv

xCl-

xc

C(xc,t) ≥ Ccr ?

Cenv

xCl-

xc

Figure 1.2 The comparison between a predicted chloride content at the depth of the reinforcement and the chloride threshold level for corrosion

What is a ”model”? A model has some input data that include some information on concrete and some information on the environment. A model is the process of how to arrive into a predicted chloride profile or chloride content, from that input data. The output should then fit or explain data from field performance in various environments such as submerged, tidal/splash, atmospheric, wicking (tunnels, caissons), de-icing (roads, road bridges, parking decks, stairs), pools etc.

1.1 Background This report is the end result of WP4 of EU-project G6RD-CT-2002-00855, ChlorTest. The over-all objectives of the project include: • to evaluate different laboratory performance test methods in terms of the theoretical basis,

technical feasibility, measurement precision and applicability in practical construction design and quality assessment.

• to recommend two reliable methods, one reference method and another rapid method, for testing the resistance of concrete to chloride ingress.

• to collect in-field performance data of chloride ingress and reinforcement corrosion.

• to verify the laboratory performance tests with the collected in-field performance data using different models, including scientific, empirical and probabilistic approaches.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 112 WP4 REPORT – Modelling of Chloride Ingress • to recommend the practical use of the laboratory performance tests and the interpretation

of the test results, including a proposal for acceptance criteria.

The project has 17 partners from all of Europe, including universities, research institutes, owners, a material supplier and a contractor. The project period is three years, starting from February 2003.

1.2 Objectives of WP4 and the work performed The Work Package 4 of ChlorTest has three objectives:

• Analysis of models, • Correlation of lab tests to in-field performance, • Bench marking of models (regarding EU standardization needs as in e.g. EN206,

EC2) In reality these objectives may be summarized in one sentence: Make sure/exemplify that test results can be used (in practice)! The work package has been divided into three tasks, 4.1 - 4.3, with the objectives “Critical evaluation of models” (4.1), “Sensitivity analysis of models” (4.2) and “Benchmarking of models” (4.3). The first chapters in this report complete Task 4.1 of WP4. The task 4.1 has been carried out by partner LTH with the aid of partner INSA. The objective of this part is limited to:

• A critical evaluation; collect and critically analyse existing models

The task 4.1 has been fulfilled by performing three sub-tasks: 4.1.a Pre-select key issues that a model should take into account (regarding demands from

EN206 & EC-2). 4.1.b Collect models from partners & experts; post list on web-page. 4.1.c Theoretical analysis of models with respect to requirements. Sub-task 4.1.a has been completed at the 2nd meeting of Chlortest in Bremen by a thorough discussion of a proposal by the working group. The final result is shown in the next section, as ”assessment criteria” for models. Sub-tasks 4.1.b and 4.1.c are completed by presenting these chapters to the partners and a number of invited to a workshop in May 2004 in Lund, Sweden. At the 3rd meeting of the ChlorTest partners, following the workshop, a limited number of models was to be selected for further treatment in tasks 4.2 and 4.3.

Task 4.2 Sensitivity analysis (by partners INSA (task leader), SP, EdF) on how each model appraises

• Environmental differences • Concrete characteristics • Corrosion onset

Task 4.2 has been dealt with theoretically, first with a probabilistic approach on the first 10 years ingress for a selection of models and then in a general way on the ERFC-models

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 112 WP4 REPORT – Modelling of Chloride Ingress Task 4.3 Bench Marking (by partners IETcc (task leader), SP, LTH, EdF) by studying the model’s abilities to reproduce reality

• Laboratory test results as input • In-field performance data as comparison

This was done by selecting some field data and ask model developers to make predictions with their models on these field cases. The prediction results were then to be analyzed by using a probabilistic approach.

1.3 Assessment criteria The basis for the analysis of the chloride ingress models are the assessment criteria. After proposals that have been presented and discussed at the two first project meetings, a final decision was taken on the assessment criteria: • Satisfactory interpretation of real behaviour by the models, • Practicality of models, • Ability to relate short-term laboratory test results to long-term performance i.e. test

results as input data, • Models must be “open”: available, transparent (not black box), input data must be

possible to quantify by the user, • Ability to take into account different environmental conditions

1.4 The structure of the report First a qualitative description of the ingress process is given (chapter 2). Then some of the fundamental parts of various chloride ingress models are explained (chapter 3). The available models are structured with regard to their physical, chemical and mathematical contents in chapter 4. Then the different types of models are theoretically, critically analysed in chapters 5 and 6. Separate chapters describe the sensitivity analysis (chapters 7 and 8) and the final benchmarking of models (chapter 9). In chapter 10 the final conclusions are drawn.


2 PHENOMENOLOGICAL DESCRIPTION OF THE CHLORIDE INGRESS PROCESS

In this chapter a brief, qualitative description is made of the most essential parts of the chloride ingress process, and the interaction between these parts. The description is intentionally made in a very brief way just to point out what kind of “reality” ingress models are expected to quantify. The description starts with the ingress media, the concrete!

2.1 Concrete, as an ingress media Chloride ingress models are generally applied to homogeneous, crack-free concrete with a one-dimensional ingress at a macro-scale. Some exceptions are available, but the main difference is pure geometrical. Since a 1D model might be applied to 2D or 3D cases, possibly with cracks, with pure mathematics and geometry the analysis in this work is limited to 1D ingress models for a crack-free concrete. The assumption of homogeneity, however, might be too simple, even if the concrete is very well mixed and compacted. The wall effect will certainly create a binder content profile close to a cast concrete surface. This will directly influence the shape of the chloride ingress profile and the total chloride contents. This is especially significant if the depth of penetration is small, since the influencing depth of a large wall effect is at least half of the maximum size of stones. Also at greater depths a certain wall effect is influencing the aggregate content and, consequently, the binder content. Vertical separation may give differences in w/c and binder content at different vertical levels. Concrete will change with time and those changes are somewhat different with depth. The continuous binder reactions will densify concrete with time and, consequently, change the pore system with time. Changes with depth will depend on the initial curing and the moisture conditions of the concrete created by the exposure conditions during its service-life. Concrete will also interact with the surrounding exposure conditions in other ways that will change the material with time and in different ways at different depths. Example of mechanisms that will have such an effect, and influence the chloride ingress process, are drying and wetting causing shrinkage and swelling, carbonation etc.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 112 WP4 REPORT – Modelling of Chloride Ingress 2.2 The concrete pore solution All concretes have a “self-desiccation” due to the contraction of the chemically bound water. This means that the pore system is not saturated. For low w/c concrete this also means that the initial pore water has an activity, an RH (relative humidity) much lower than 100 %. Concrete for chloride environments may have a self-desiccation down to some 85-90 % RH, corresponding to a degree of capillary saturation of the pores, excluding the air pores, of some 0.8-0.9. The pore liquid is not pure water, but a strong solution of ions, mainly sodium, potassium and hydroxides, besides the chlorides, different for different types of binders and concrete compositions. The self-desiccation will further increase the concentrations of these species.

2.3 The exposure conditions The boundary conditions for the chloride ingress process are very different in various environments. The most simple environment, the submerged zone of marine concrete structures, the main components of the sea water are chloride and sodium, but a number of other ions are present in small quantities. Depending on the local conditions, the salinity and the sea water temperature may have annual variations. Consequently the exposure solution is quite different from the pore solution in the concrete, not only concerning the chlorides. In other environments, the exposure conditions are much more complicated. A “constant” exposure solution may not be present all the time, but occasionally replaced by pure water from driving rain, or drying conditions, as in the splash zone of marine structures. In environments with de-icing salts the exposure solution is very strong in sodium chloride for a short while, then diluted and sometimes replaced by pure water from rain. Most of the time the concrete surface is exposed to drying conditions.

2.4 The ingress process For a saturated concrete constantly exposed to a solution of sodium chloride, the difference in ionic strength between the pore solution and the exposure solution will initiate a transport of ions. Chloride ions will move into the concrete pore solution and some of the alkalis and hydroxides in the pore solution will start to leach out. This is popularly expressed as “diffusion” of ions, but the various ions will interact with each other in such a way that charge balance is maintained. Flux of one ion is affected by the fluxes of other ions because of the electrical field created by all the ions. In a real concrete, the pore system is not saturated with water. Consequently, there will be a moisture flow into the pore system. Additionally, depending on the environmental conditions, that moisture flow can vary with time, in magnitude and direction, causing convection of ions in the pore system and in and out of the concrete. Below a certain moisture content, the moisture flow cannot carry any ions, possibly causing an accumulation of ions in certain areas. The movement of ions in the pore system is accompanied by a significant interaction with the matrix, sometimes called “binding”, especially for chloride and hydroxide. Because of this interaction penetration of chloride is significantly delayed. This interaction depends mainly of

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 112 WP4 REPORT – Modelling of Chloride Ingress the type and amount of binder in the concrete, but is affected by for instance the temperature and the concentration of alkalis. The chloride interaction/binding seems to be reversible to a large extent. That means that if the concentration of chloride decreases, e.g. if the concrete surface temporary is exposed to rain or if the concrete temperature increases, part of the binding is lost. Carbonation will also cause a loss of most of the binding capacity. In conclusion, chloride ingress is never a simple “diffusion” process due to concentration differences only and the “chloride binding capacity” is not a simple, constant property. Chloride ingress models must include a number of assumptions and simplifications and these must be stated. The problem is to realize the significance of these simplifications! This is further discussed in later chapters.


3 SOME FUNDAMENTAL PARTS OF INGRESS MODELS

In this chapter a number of “fundamentals” for chloride ingress models are defined. Some fundamental parts have been chosen that are frequently used in various models. The objective is to state what fundamentals could be agreed upon, without discussing specific models and have this as a background when the different models are presented, analysed and discussed.

3.1 Chloride concentrations The content of chloride ions in concrete may be expressed in a number of ways. Nilsson et al (1996) gave an overview that is summarized here. The most frequently used expressions are: cI [kg Cl/m3 of solution] [3.1:1] cII [kg Cl/m3 of pore solution] [3.1:2] cIII [kg Cl/m3 of material] [3.1:3] cIV [kg Cl/m3 of solid material] [3.1:4] cV [kg Cl/kg of cement or binder] [3.1:5] cVI [kg Cl/kg of gel] (“gel”=aC+wn) [3.1:6] cVII [kg Cl/kg of concrete] [3.1:7] cVIII [kg Cl/kg of sample] [3.1:8] Consequently, giving the amount of chlorides in only [kg/m3], [kg/kg] or [%] might cause confusion and large errors in trying to use the data. The mass in kg is sometimes replaced by moles. It does not change the relations and the relation between kg Cl and moles Cl is obvious. Concentrations in g/l are identical to concentrations in kg Cl/m3 of solution. Most of the relations between the different concentrations in [3.1:1] to [3.1:8] are simple. The concentrations cIII to cVIII are related to each other by the dry density of the material γ, dry density of the solid material ρ, cement content C, degree of hydration α, the moisture content w and the molar weight of chloride MCl. The relation between cII and cIII to cVIII is more difficult, however:

cII = cIII/psol [3.1:9]

where the problem is to define the porosity psol. It includes only that part of the porosity that contains a liquid which acts as a solvent. One might question whether all the pore water

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 112 WP4 REPORT – Modelling of Chloride Ingress really acts as a solvent, i.e. psol = total porosity. Recent measurements by Mangat & Molloy [1995] give some information. They measured the concentration of free chlorides. When multiplied by the total amount of pore water, this gave more free chlorides than the total actually present in the samples! If the measured concentrations are correct, it is obvious that not all the pore water acts as a solvent. The capillary porosity or the empty porosity at 11 % RH or 45 % RH might be better alternatives until further research shows the relationship. Chlorides occur in different forms in concrete. Some of the chlorides are free ions dissolved in the pore solution. The rest of the chlorides are chemically and physically bound to the reaction products and their surfaces C = c + cb [3.1:10]

where c and C are concentrations in [kg/m3], [moles/volume] or [kg/kg]. The free chlorides are usually given per volume of solute, i.e. [kg/m3 solution]. The bound, and total, amount of chlorides, however, are given per weight of binder or concrete. Consequently, to distinguish between free and bound chlorides as in [3.1:10] the porosity psol has to be used. The question is of course how to define the concept “free” chloride! Several options are possible:

• Free to move? • Free to be leached out? • Free to corrode steel? • Free to be liberated?

The answer to this question is absolutely vital for physical models, cf. Figure 3.1. Without consensus about this vital question, it is hardly possible to agree upon physical, or any, models at all.

Total Cl

Bound?

DepthFree?

Total Cl

Bound?

DepthFree?

Total ClBound?

DepthFree?

?)0(

)0(

==

=<

xc

xc

free

free

Total ClBound?

DepthFree?

?)0(

)0(

==

=<

xc

xc

free

free

Fig. 3.1 Alternative divisions of the total amount of chloride into “free” and “bound” in a chloride ingress profile

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 112 WP4 REPORT – Modelling of Chloride Ingress 3.2 Fick’s 1st law of diffusion Traditionally, Fick’s 1st law has been used to describe the chloride flux as a result of pure diffusion in liquid water, with the free concentration c of dissolved chloride as the transport potential.

xcDq FCl ∂

∂−= 1 [3.2]

The “diffusion coefficient” DF1 is then defined by equation [3.2] and by the test set-up used to measure it.

3.3 Chloride binding/interaction, binding capacity The interaction between chloride and the matrix of cement-based materials is still not very well understood. Instead, for most applications a “binding isotherm” is used to give the relation between the free and bound chloride. Here a number of questions are still disputed, besides the approach by a binding isotherm at all. Since the binding isotherm is a pure empirical “property”, it must be measured and the effect of a number of parameters must be quantified. The shape of the binding isotherm is one such question, cf. Figure 3.3.1. Does most of the binding really occur at concentrations close to zero, or is the chloride binding significantly concentration-dependent?

[free chloride], cf

[Bound or total chloride]

cb or ctot Or?

Or?

[free chloride], cf

[Bound or total chloride]

cb or ctot Or?

Or?

Or?

Or?

Fig. 3.3.1 Alternative shapes of a possible “binding isotherm”.

Another important question is the relation between “free” and “bound” chloride, cf. Figure 3.3.2.


[Free Chloride], cf [g/l]

Total

Bound

Ctot

Free

[Free Chloride], cf [g/l]

Total

Bound

CtotCtot

Free

Fig. 3.3.2 The division of the total amount of chloride into “free” and “bound”

The binding capacity is the capacity of a material to bind chlorides when the ion concentration changes

cc

capacitybinding b

∂∂

= [3.1:13]

This is the slope of a “binding isotherm” with linear scales as in figure 3.3.1. The dimension of the binding capacity depends on the units chosen for cb and c, respectively. Obviously the binding capacity depends on the concentration, as seen from figure 3.3.1, but is sometimes assumed to be a constant. As seen the chloride binding capacity is very high at low chloride concentrations. One important, but not obvious, question is whether the concentration of chloride in the pore water is equal to the concentration in the surrounding sea water or exposure solution. Some recent ideas on determining “free” chloride by leaching in water cause confusion about “free” and “water soluble” chloride that really question such a statement.

Chloride binding is, from measurements, a function of several parameters: concentration, pH, temperature, moisture content, gel content, type of binder, binder content, water-cement ratio, degree of hydration, time etc. The mechanisms behind all these observations are not fully understood and much more research is needed and measurements must be much better performed. See also Larsen (1998). Recently some ”strange” time effects have been observed on chloride binding in submerged concrete. The surface chloride content has been found to be time-dependent Cs(t)! A longer exposure time does not only give deeper chloride ingress but also higher chloride content, cf. Figure 8.


H4

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 10 20 30 40 50 60

Depth [mm ]

Chl

orid

e [w

t-%

of b

inde

r]

5y

1y.

0.6y.

2y0-50TL

2y50-100

Fig. 3.3.3 Chloride profiles after several exposure times showing higher chloride concentration with time. Submerged SRPC concrete with w/b=0.40.

These observations cannot be explained by today’s knowledge. The consequences of alternative causes and how these findings are utilized in chloride ingress models are quite contradictory, cf. Figure 3.3.4.

0

1

2

3

4

5

6

7

8

0 20 40 60 80 100

Depth [mm]

C(x

,t) [w

t-% o

f bin

der]

Cs(t)

Cs=const0

1

2

3

4

5

6

7

8

0 20 40 60 80 100

Depth [mm]

C(x

,t) [w

t-% o

f bin

der]

Cb(t)

Cb=const

Fig. 3.3.4 The consequences of alternative causes for the findings in Figure 3.3.3: time-dependent surface chloride content Cs(t) (left) or time-dependent chloride binding Cb(t) (right).

A time effect on the ”surface chloride content” Cs in an empirical ingress model, that has no simple explanation, has a significant consequence: a larger depth of penetration than with a constant Cs. On the contrary, a time effect on chloride binding will give a smaller predicted depth of penetration, simply because the larger chloride binding capacity with time will retard the ingress even more!

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 112 WP4 REPORT – Modelling of Chloride Ingress 3.4 Mass balance equations The mass balance equation for chloride can be expressed in different ways, depending on whether separate terms are used for diffusion and binding. The most simple form gives the balance of the total amount of chloride per unit volume.

Ctot

dx

qCl qCl +dqCl

x

Ctot

dx

qCl qCl +dqCl

x

Figure 3.4.1. A visual description of the mass balance equation

The flux of chloride qCl is different at different positions in the concrete. The difference in chloride flux to and from an infinitesimal slice of concrete with a thickness dx will change the total amount of chloride C in such a slice. The change in total chloride content per unit of time will be equal to the difference in chloride flux to and from the slice, divided by the thickness of the slice. Consequently, the mass balance equation will be

xq

tC Cl

∂∂

−=∂∂ [3.4.1]

To get the dimensions correct, the chloride content here is the content per unit volume of concrete, not the pore volume, and the flux is per unit area of concrete. The negative sign simply says that the chloride content will decrease if dqCl is positive. For chloride ingress dqCl is negative and the chloride content will increase with time. The change in total chloride content can also be split into a change in free chloride dissolved in the pore water and a change in bound chloride in such a way that equilibrium will be maintained between free and bound chloride. Since the chloride flux occurs in the pore water, the change in chloride content will first occur as change in free chloride. An almost “instant” change in bound chloride will follow, since the rate of chloride binding is fairly high. The mass balance equation will then be

dtc

xq

dtc

orx

qdtc

dtc

dtC

bCl

Clb

∂−

∂∂

−=∂

∂∂

−=∂

+∂

=∂

[3.4.2]

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 112 WP4 REPORT – Modelling of Chloride Ingress 3.5 Fick’s 2nd law of diffusion What usually is called Fick’s 2nd law is in reality a mass balance equation. Originally it was meant for mass balance in solutions, where no binding exists, but it has been widely applied to mass transport in porous systems, with and without binding. 3.5.1 Fick’s 2nd law of diffusion for a solution The law of mass conservation in a small volume of a solution gives the changes with time of the (free) chloride concentration in that unit volume

xq

tc Cl

∂∂

∂∂

−= [3.5:1]

This is the same equation as Equation [3.4.1], with the free chloride content equal to the total content, and Equation [3.4.2], with the binding term equal to zero. With the flow description as Fick´s 1st law for diffusion in a solution, [3.2] and with the diffusion coefficient DF1 as a constant, one gets

2

2

1 xcD

tc

F ∂∂

∂∂

= [3.5.2]

This looks like the “Fick’s 2nd law”, but it should be noticed that it is applicable for solutions only!

3.5.2 Fick’s 2nd law of diffusion for a porous material For a porous material, with binding, Equation [3.4.2] is applicable. By assuming that the flux can be described by Fick’s 1st law, Equation [3.2], and that D is a constant, one gets

2

2

11 xcD

xcD

xtC

FF ∂∂

=⎟⎠⎞

⎜⎝⎛

∂∂

−∂∂

−=∂∂ [3.5.3]

This equation is not identical to “Fick’s 2nd law”, since the concentrations are different in the two terms. C is expressed per volume of concrete but c is per volume of pore solution. A factor equal to psol, according to Equation [3.1:9], must be inserted. Additionally, C is the total content but c is the free content. A factor of

cc

cc

cc

cC bb

∂∂

+=∂∂

+∂∂

=∂∂ 1 [3.5.4]

Then the mass balance equation will be

xC

cc

p

DxC

xcc

p

Dxc

xD

tC

bsol

F

bsol

FF 2

211

1

11 ∂∂

⎟⎠⎞

⎜⎝⎛

∂∂

+=

∂∂

∂∂

⎟⎠⎞

⎜⎝⎛

∂∂

+=

∂∂

∂∂

=∂∂ [3.5.5]

Comparing this equation to the traditional Fick’s 2nd law


xCD

tC

F 2

2

2 ∂∂

=∂∂ [3.5.6]

one realize that the two equations are equal, but that the “diffusion coefficient” DF2 in Fick’s 2nd law is different from the diffusion coefficient DF1 in Fick’s 1st law, for a porous material with binding. In fact, the total concentrations C in Fick’s 2nd law could be replaced by the free concentrations c, but the diffusion coefficient still is the same, DF2

xcD

tc

F 2

2

2 ∂∂

=∂∂ [3.5.7]

3.5.7 The relation between the D:s in Fick’s laws From Equations [3.5.5] and [3.5.6] the relationship between the two coefficients in Fick’s laws are clear

⎟⎠⎞

⎜⎝⎛

∂∂

+=

cc

p

DD

bsol

FF

1

12 [3.5.8]

The magnitude of the relationship depends on the porosity and the binding capacity of the concrete.

3.6 The error function, and its compliment The mathematical solution to Fick’s 2nd law, Equation [3.5.6], is well known for a semi-infinite medium, with a constant diffusion coefficient Da=DF2 and a constant surface chloride content C(x=0, t)=Csa. If the initial chloride content Ci is negligible the solution is the compliment to the error-function, Crank (1976)

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅⋅=

tDxCtxC

a

sa 2erfc),( [3.6]

where t is the exposure time. This equation gives a “chloride profile” C(x,t)/Cs=erfc(z)


0

0.2

0.4

0.6

0.8

1

0 0.5 1 1.5 2Z=x/sqrt(4Dt)

C(x

,t)/C

s

Figure 3.6 A chloride profile, chloride concentration with “normalized” depth, from the erfc-solution to Fick’s 2nd law

3.7 The apparent diffusion coefficient Da and apparent surface chloride content Csa

Frequently, the erfc-solution to Fick’s 2nd law, Equation [3.6], is fitted to measured chloride profiles, from structures or specimens. In such a curve-fitting, if the exposure time t is inserted, the best curve-fitting gives two regression parameters, Da and Csa. The index a means “achieved” or “apparent”. The definitions are

- Da is the “apparent diffusion coefficient” achieved after an exposure time t, assuming that the diffusion coefficient DF2 in Fick’s 2nd law was constant during the whole exposure. Then Da equals DF2.

- Csa is the “apparent surface chloride content” achieved after an exposure time t, assuming that the boundary condition was constant during the exposure. Then Csa =C(x=0,t).

3.8 Time-dependency of diffusion coefficients Nowadays, in most chloride ingress models the diffusion coefficient is treated as time-dependent. The time-dependency is very much different for different types of diffusion coefficients. Empirical models in the beginning of the 1990’s were developed to include the effect that was obvious in data from different exposure times: a time-dependent Da. Several researchers started to use expressions for the time-dependency like equation [3.8.1].

α

⎟⎠⎞

⎜⎝⎛=

tt

DD exaexa

[3.8.1] One example of data for 0.6 to 5 years of exposure is shown in Figure 3.8.1.


Achieved diffusion coefficient, 0.6-5 years

y = 1E-12t-0.4162

y = 1E-12x-0.523

1.E-13

1.E-12

1.E-11

0.1 1 10 100Age, t [years]

Da(

t)

BinderConcreteFit ConcreteFit Binder

Fig. 3.8.1 Apparent “diffusion coefficients” plotted against the length of the exposure time

The time-dependency was found to be different in different exposure zones. Part of it could be explained by laboratory measurements showing a normal densification of concrete, somewhat different for different binders. The effect has not yet been fully explained. Can we use the observations for long-term predictions if we cannot understand it? One possible explanation is the physical change in the surface layer. This effect and its consequence for the Da(t) are for instance given by Maage & Helland (1991) and Mohammed et al (2002). Significant parameter studies are reported by for instance Maage et al (1999). This research demonstrates that the contribution from ongoing hydration to the aging exponent is in the range of 0.1 – 0.15 for a number of binders (hydration in fresh water for periods up to one year) while concrete exposed for seawater experienced an aging exponent in the range of 0.50 to 0.80 (average 0.60). Changes of the surface layer as a densification will, however, only change the diffusion coefficient in the surface layer. It should then be smaller in the outer layer than in the bulk. In all models, however, the time-dependency of the diffusion coefficient is treated as a change with time of a diffusion coefficient that is constant with depth. A lower D close to the surface should change the shape of the chloride profile in such a way that it should be steeper with time in the surface layer. This is not we observe. An explanation as a “densification” due to contact with sea water, consequently, must be more or less instantaneous in the whole part of the concrete that has any chloride ingress. Further research is needed to confirm that. A time-dependent D causes a lot of confusion. On one hand there are observations showing the apparent Da to be time-dependent. That D is taken from curve fitting data after a certain length of exposure to the error-function. That means that the DF2 is regarded as constant through the whole exposure and the D is a kind of an “average D” during that period, cf. Figure 3.8.2.


Achieved diffusion coefficient, 0.6-5 years

y = 1E-12t-0.4162

y = 1E-12x-0.523

1.E-13

1.E-12

1.E-11

0.1 1 10 100Age, t [years]

Da(

t)


texp

),0()( 22 tDtDapp =

),0()( 11 tDtDapp =Achieved diffusion coefficient, 0.6-5 years

y = 1E-12t-0.4162

y = 1E-12x-0.523

1.E-13

1.E-12

1.E-11

0.1 1 10 100Age, t [years]

Da(

t)


texp

),0()( 22 tDtDapp =

),0()( 11 tDtDapp =

Fig. 3.8.2 The time-dependent apparent D after several lengths of exposure during which the D is assumed constant!

The time-dependency of the constant, apparent D during an exposure from 0 to t must not be confused with the diffusion coefficient at a certain age t.

)(),0( tDtD testapp ≠agetexp [3.8.2]

The time-dependency of the apparent diffusion coefficient must be explained before it can be used with confidence in predictions! Additionally, the age-dependency of a D being a material property must be determined. We need test methods for that! An attempt to clarify these things is done in the next section. 3.8.1 Time-dependency of instantaneous diffusion coefficients A number of measurements show that a chloride diffusion coefficient decreases with time. A reasonable explanation is of course the continuous cement hydration and binder reactions that densify the concrete with time. A common equation to describe this time-dependency of the diffusion coefficient D(t) is

n

testtest t

tDtD ⎟

⎠⎞

⎜⎝⎛=)( [3.8.1a]

where t is the age, ttest is the age at testing, Dtest is the test result at that age. An example is shown in Figure 3.8.3, with Dtest=10-11

m2/s at ttest=30 days and n=0.50.


1.E-13

1.E-12

1.E-11

1.E-10

0.00 0.01 0.10 1.00 10.00 100.00

Age [years]

D(t)

[m2 /s

]

D(t)

Fig. 3.8.3 An example of the time-dependency of the chloride diffusion coefficient. The parameter t is the age of the concrete

Such a time-dependency of the diffusion coefficient means that the diffusion coefficient continuously changes with time or actually with age. It should be noticed, however, that the decrease with time is shown for a limited age only, sometimes only for six months and a few years for certain binders.

3.8.2 Time-dependency of apparent diffusion coefficients A lot of field and laboratory data also show a clear time-dependency of the apparent diffusion coefficient Da(t-texp) That time-dependency is usually expressed in terms of the exposure time t-texp, where texp is the concrete age at exposure.

m

a ttt

DttD ⎟⎟⎠

⎞⎜⎜⎝

⎛

−=

exp

00exp ),( [3.8.2a]

where t is the age, texp is the age at exposure, D0 is the diffusion coefficient at a reference age t0. The exponent m is different from n in equation [3.8.1]. One obvious peculiarity with this expression is that Da(t-texp) is not defined at the time of exposure! It is obvious that the Da(t-texp) is, and must be, different from D(t), but the relation between them is not obvious and among other things dependent on the age at exposure. Additionally, the explanation of the time-dependency of the apparent diffusion coefficient is not only the continuous densification of the concrete. The apparent diffusion coefficient is time-dependent, but it is assumed to be constant during the complete exposure time, as a consequence of its definition! An example is shown in Figure 3.8.4a, where the time-dependency of the instantaneous diffusion coefficient in Figure 3.8.3 is used to determine a numerical solution to Fick’s 2nd law.


1.E-13

1.E-12

1.E-11

1.E-10

0.00 0.01 0.10 1.00 10.00 100.00

Age [years]

D(t)

[m2 /s

]

D(t)Da(texp, t)

Fig. 3.8.4a An example of the time-dependency of the apparent chloride diffusion coefficient. The parameter t is the age of the concrete and texp is the age at exposure, 30 days or 0.082 years.

From Figure 3.8.4a it is obvious that the time-dependency of the apparent diffusion coefficient does not exactly follow an equation like [3.8.2a]. However, such an equation gives an approximate description, with a different exponent, of course. An m-value of 0.39 gives a fairly good fit for the example in Figure 3.8.4a, cf. Figure 3.8.4b.

1.E-13

1.E-12

1.E-11

1.E-10

0.00 0.01 0.10 1.00 10.00 100.00

Age [years]

D(t)

[m2 /s

]

D(t)Da(texp, t)Da(t)

Fig. 3.8.4b An example of an approximate time-dependency of the apparent chloride diffusion coefficient according to an equation like [3.8.2a], for m=0.39.


0.E+00

2.E-12

4.E-12

6.E-12

8.E-12

1.E-11

0.01 0.10 1.00 10.00 100.00

Age [years]

D(t)

[m2 /s

]

D(t)Da(texp, t)Da(t)

Fig. 3.8.4c An example of an approximate time-dependency of the apparent chloride diffusion coefficient according to an equation like [3.8.2a], for m=0.39. Figure 3.8.4.b, but with a linear scale on the vertical axis.

Some researchers quantified theoretically the relationship between the apparent diffusion coefficient and the instantaneous one. Visser et al [2002] for example pointed out the theoretical relationship if they are described by Equations [3.8.1] and [3.8.2a] respectively.

n

oa tt

tn

DttD ⎟

⎟⎠

⎞⎜⎜⎝

⎛

−−=

exp

0exp 1

),( [3.8.2b]

This equation, however, is not quite realistic. It is based on another assumption, that the diffusion coefficient is not defined at the time of exposure. It is said to be valid for large values of t, only. A correct description of Da(t-texp) can be derived from Crank (1976). For a time-dependent D(t), the Da(t) times the exposure time t-texp in the erfc-solution is found by integrating D(t) over time, from the start of exposure

∫=

⋅=−⋅t

tta dttDttD

exp

')'()( exp [3.8.2c]

With Equation [3.8.1] inserted, the resulting apparent diffusion coefficient will be

dtt

tD

ttttD

ntest

t

ttesta ⎟

⎠⎞

⎜⎝⎛

−= ∫

expexpexp

1),( [3.8.2d]

and the expression will be


expexp

expexp

exp

1exp

exp

11

),(

1),(

ttt

ttt

tt

nD

ttD

or

ttttt

tt

nD

ttD

nntesttest

a

nnntesttest

a

−⋅⎟

⎟

⎠

⎞

⎜⎜

⎝

⎛⋅⎟

⎟⎠

⎞⎜⎜⎝

⎛−⋅⎟

⎠⎞

⎜⎝⎛

−=

−

⋅−⋅⎟

⎠⎞

⎜⎝⎛

−=

−

[3.8.2e]

A similar expression was shown by Gulikers (2004).This expression is shown in Figure 3.8.4d to coincide with the result from curve-fitting the numerical solution in Figure 3.8.4a.

1.E-13

1.E-12

1.E-11

1.E-10

0.00 0.01 0.10 1.00 10.00 100.00

Age [years]

D(t)

[m2 /s

]

D(t)Da(texp, t) numDa(t-texp) eq

Fig. 3.8.4d An example of a time-dependent apparent chloride diffusion coefficient according to equation [3.8.2e], compared to the numerical solution and curve-fitting

3.8.5 Conclusions on time-dependent diffusion coefficients The time-dependency of diffusion coefficients is certainly a possible source of misunderstanding and even mathematical errors. In any model where such a time-dependency is used, the time-dependency must be very well defined. Here only the continuous densification of concrete was taken as a cause of the time-dependency of diffusion coefficients. Other reasons are possible, and likely, especially under field conditions. Those reasons, however, are not yet fully explained. Since the aging factor gets a contribution both from continued hydration and the beneficial, but not understood, contact with seawater (exponents β and γ, respectively). Thus the correct argument should be the concrete’s total age for the hydration effect and the time of exposure for the extra effect of being in contact with seawater. For many practical applications (assessment of existing structures), these two periods will be more or less the same. However for parameter studies of the effect of curing conditions etc., we have to split up these two effects. A possible way of getting a consistent model is therefore to describe the two components of the aging factor and let them be functions of total age and period of exposure respectively.


3.9 The Nernst-Planck equation for ion flux Nowadays it is widely recognized that Fick’s 1st law is a very strong over-simplification of chloride transport. Instead, the effect of the electric potential field Φ, from other ions and from an applied potential difference in migration tests, is included in the flux equation. The flux is then described by the Nernst-Planck equation.

⎟⎠⎞

⎜⎝⎛

∂Φ∂

+∂

∂+

∂∂

−=x

cRT

Fzxa

cxc

Dq iii

ii

iClln

[3.9]

The “diffusion coefficient” in Fick’s 1st law DF1 is obviously not a material property but depends on the conditions. Consequently, it cannot be determined by a simple “diffusion” test. Any attempt with straight forward diffusion cells will determine something else since equation [3:2] is not a correct description of what happens in such a test. Instead it must be acknowledged that the flux of chloride is influenced by the other ions. In any test set-up for determining the “diffusion coefficient” Di for chloride, using equation [3:8], the result will depend on the diffusion coefficients for all other ions. Consequently, it is not possible to determine the chloride diffusion coefficient directly in one simple test. Different approaches to solve this problem have been used. Truc (2000) made estimations from chloride migration tests where the diffusion coefficients for sodium, hydroxyl and potassium ions were adjusted to fit predictions from a multi-species model. Samson (1999) instead utilized the diffusion coefficients for several ions from data for diffusion in a solution. By determining the “formation factor”, i.e. the effect of the tortuosity and the restrictivity of the pore system, he could describe the flux of all ions by applying the same formation factor to the flux of all ions. The main difficulty in applying Equation [3.9] is to quantify the potential gradient ∂Φ/∂x since it is a function of the fluxes of all ions and consequently changes with time. In a migration test simple versions of Equation [3.9] are used since the applied electrical field may dominate the potential field.


4 TYPE OF CHLORIDE INGRESS MODELS Different prediction models have a number of similarities and differences. They require different input data that look alike, but sometimes are quite different. A structure of the available methods is shown in this chapter. The various models are described in more details in the next chapter.

A. MODELS BASED ON FICK’S 2ND LAW. “Empirical methods” are empirical or semi-empirical methods where chloride profiles are predicted from analytical or numerical solutions (true or false!) to Fick’s 2nd law of diffusion or other models where the diffusion coefficient DF2 in Fick’s 2nd law is used, sometimes as an “apparent diffusion coefficient” Da. ERFC-methods

A1a Constant Da & Csa (Traditional, Collepardi, Japan, Tuutti) A1b Da(t) & constant Csa (Selmer-Poulsen, Bamforth, Allied, Firth) A1c D(t) & constant Csa (DuraCrete) A1d Da(t) & Csa(t) (Nilsson)

Analytical solutions to Fick’s 2nd law (other than erfc)

A2 DF2(t) & Cs(t) (Mejlbro-Poulsen, Hetek)

Numerical solutions to Fick’s 2nd law A3a DF2(t) & Csf(t), binding isotherm (NIST, LEO) A3b DF2(t) & Cs(t) (Life-365)


B. MODELS BASED ON FLUX EQUATIONS “Physical methods” are methods where chloride transport and chloride binding are described with separate, physical expressions.

Models based on Fick’s 1st law B1a Binding isotherm (ClinConc) B1b Convection models (HetekConv, Imperial, Toronto, THI EdF, LERM,

Meijers)

Models based on the Nernst-Planck flow equation B2a Binding isotherm (MS-Diff, Li&Page) B2b Ion-solid equilibrium (Johannesson) B2c Chemically bound Cl (Stadium)


5 CHLORIDE INGRESS MODELS BASED ON FICK’S 2ND LAW

5.1 General Empirical prediction models for chloride ingress into concrete utilize experience to fit data to mathematical models. If the fit is not good enough, more regression parameters are added. New experience is required for a new concrete or a new environment. Predictions by empirical models should be done during the observation period only. Verification is then achieved from a comparison between data and model, cf. Fig. 5.1.1. Scatter between observations and predictions give the model uncertainty.

Depth [mm] of Penetration

0

5

10

15

20

25

30

35

40

0 1 2 3 4 5 6

Exposure Time [years]

Fig. 5.1.1 Predicted chloride ingress by an empirical model in the time interval where the model is valid: the observation period.

Predictions beyond exposure times where the model has been calibrated the data are highly questionable, since an empirical model by definition is only a mathematical fit to observations. Since it really has no clear physical meaning, it is dangerous to use it for extrapolation outside the borders of the observation data. Figure 5.1.2 shows this danger very clear. It is an alternative way of showing the predictions in the left part of Figure 1.1.


0

20

40

60

80

100

120

0.1 1 10 100

Exposure time [years]

Pen

etra

tion

dept

h (fo

r C=1

.0 %

of b

inde

r) [m

m

Data0.6-2 y

Alternativepredictions:

0

20

40

60

80

100

120

0.1 1 10 100


Pen

etra

tion

dept

h (fo

r C=1

.0 %

of b

inde

r) [m

m

Data0.6-2 y

Alternativepredictions:

Fig. 5.1.2 The alternative predictions for 30-100 years in Figure 1.1 compared to the observation data between 0.6 and 2 years. The circle at 5 years represents later measurements that were not available for the predictions.

5.2 Boundary conditions concepts in empirical models In empirical models the boundary condition is expressed as a surface chloride content Cs, that really is the response by a particular concrete to particular environmental actions. That response does not only depend on the environment but also on the concrete! Consequently, a Cs cannot be taken as boundary conditions for a particular environment. The effect of the concrete must be considered as well. The Cs is the total amount of chloride in the surface-near region of the concrete. Consequently, it depends on the porosity of the concrete and on the binding capacity of the concrete. That explains why Cs is a function of the concrete mix, especially the type of binder, the binder content, and the water-binder ratio. It also explains why Cs is a function of temperature, pH, carbonation etc.

13 concretes submerged for 5 years

0

0.2

0.4

0.6

0.8

1

1.2

0 10 20 30 40 50 60

Depth(mm)

Cl %

per

sam

ple

Cs = 0.5 to 1.2( wt-% of concrete)

Fig. 5.2.1 Several examples of Csa = f( concrete mix) in one marine, submerged environment

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 32 of 112 WP4 REPORT – Modelling of Chloride Ingress Since the apparent diffusion coefficient obviously depends on the binding capacity, cf. Equation [3.5.8], there is correlation between the Da and Csa for a particular concrete. They cannot be chosen independently! In the 1990’s the Csa has been regarded as time-dependent by a number of researchers. The explanation for this is not yet given, but in some environments it is obvious that the chloride content could increase with time, e.g. the splash zone in road environments. Drying and wetting in the splash zone of marine structures could possible explain a time-dependent Csa in that zone. However, recently it was observed that Csa(t) also is time-dependent in the submerged zone of marine structures, cf. Figure 3.3.3. The effect was not visible if the data is expressed as chloride by weight of sample simply because of the large scatter, cf. Figure 5.2.2. However, the time-dependency is clearly visible if Csa is expressed by weight of binder, see Figure 5.2.3. We cannot explain why, however! This is vital, both for empirical models and for physical models. We urgently need better, long-term data.

Achieved Surface Chloride Content

y = 0.0222Ln(x) + 0.6542

0.0

0.2

0.4

0.6

0.8

1.0

0.1 1 10 100Age, t [years]

Csa

(t)

Figure 5.2.2 Surface chloride contents Csa as a function of exposure times 0.6 to 5 years for the concrete in Figure 1.1. Csa expressed by weight of sample.

Achieved Surface Chloride Content

y = 0.5676Ln(x) + 3.5998

0

1

2

3

4

5

6

7

0.1 1 10 100Exposure time, texp [years]

Csa

(t) [w

t-% o

f bin

der]

Figure 5.2.3 Surface chloride contents Csa as a function of exposure times 0.6 to 5 years for the concrete in Figure 1.1. Csa expressed by weight of binder.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 33 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.3 Models A1a: ERFC-model, constant Da & Csa

(Traditional, Collepardi model) The most frequently used empirical model for chloride ingress is the error-function solution to Fick’s 2nd law. This model was almost the only one for twenty years, between 1970, when Collepardi [1970] proposed it, up to around 1990. It is still widely used. 5.3.1 Principle The principle of this model is: 1) Estimate the “apparent diffusion coefficient” Da and the “apparent surface chloride

content” Csa, from field exposure or laboratory tests. 2) Predict the chloride ingress, the chloride profile, from the erfc-solution to Fick’s 2nd law,

with constant parameters.

5.3.2 Mathematics In full the prediction equation is

⎟⎟⎠

⎞⎜⎜⎝

⎛

−⋅⋅−+=

)(2erfc)(),(

exa

isai ttDxCCCtxC [5.3.2a]

but it can be shortened, if the initial chloride content Ci is negligible, to

⎟⎟⎠

⎞⎜⎜⎝

⎛

−⋅⋅=

)(2erfc),(

exa

sa ttDxCtxC [5.3.2b]

An alternative empirical model that has been proposed is, i.e. by Andrade (2004) : tDkx aCc cr

⋅⋅== [5.3.2c] It looks much simpler, but it is not! The parameter k can easily be found from equation [5.3.2a], where it is obvious that the two models really are identical! The chloride content C is the total chloride content of the particular concrete. However, alternatively the free chloride concentration in the pore system could be used in the same set of equation, of course with the Csa replaced by csa of the surrounding environment, Tuutti [1983].

5.3.3 Required Input data The model requires only two parameters to be quantified, Da and Csa.

5.3.4 Input relation(s) to test methods Da can be derived from a short term immersion test, in the laboratory or under field conditions.

5.3.5 Required boundary conditions The required boundary conditions are simple, the future achieved response by the particular concrete in the particular environment.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 34 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.3.6 Model(s) for environmental actions Environmental actions are translated into a surface chloride content Csa mainly from exposure tests in the true environment, for the concrete that is going to be used. Some models for Csa are available, i.e. Lindvall [2003], where Csa can be estimated from the concrete composition (type and amount of binder, w/c etc.), and various environmental parameters, such as sea water temperature and salinity, distance to sea level, distance to road surfaces etc.

5.3.7 Output data The prediction gives a chloride profile like the erfc-solution for each exposure time.

5.3.8 Openness/Practicality The model is very simple and fully transparent for any user.

5.3.9 Interpretation of real behaviour Today it is obvious from numerous field observations that the model tremendously overestimates the chloride ingress rate. Expressed in terms of achieved diffusion coefficient Da it is quite clear that Da could not be treated as a constant in time. The model is today almost always replaced by one of the models where Da is time-dependent.

5.3.10 Advantages The main advantage is the simplicity in performing the predictions.

5.3.11 Limitations & Drawbacks Besides the general limitations and drawback for empirical models, see section 5.9, the main limitation and drawback is the great lack of coincidence between predicted and measured profiles, also after a few years of exposure, if more than one exposure time is used in the observations.

5.3.12 References Collepardi [1970], Tuutti [1983], Poulsen [1990], Andrade [2004]


5.4 Models A1b: ERFC-model, Da(t) & constant Csa (The Selmer-Poulsen model etc.)

In the late 1980’s a number of researchers started to apply time-dependent apparent diffusion coefficients to chloride ingress models. The most elaborated model is the Selmer-Poulsen model, based on the research of a number of other researchers. 5.4.1 Principle The principle of this model is: 1) Estimate the time-dependency of the “apparent diffusion coefficient” Da(t) and the

(constant) “apparent surface chloride content” Csa, from field exposure or laboratory tests. 2) Predict the chloride ingress after a certain exposure time, the chloride profile, from the

erfc-solution to Fick’s 2nd law, assuming constant parameters DF2 and Csa during exposure.

5.4.2 Mathematics The mathematical model simply is the same as the previous one

⎟⎟⎠

⎞⎜⎜⎝

⎛

−⋅⋅−+=

)()(2erfc)(),(

exa

isai tttDxCCCtxC [5.3.2a]

where t is the age of the concrete and tex is the age at exposure, tex being negligible at long exposure times. The time-dependency of the apparent diffusion coefficient Da(t) is found from empirical equations like

α

⎟⎠

⎞⎜⎝

⎛=t

tDD ex

aexa [3.8.1] & [5.3.2b]

where t is the age of the concrete and tex is the age at exposure. Daex is the (undefined) apparent diffusion coefficient at the time of exposure. The exponent α depends on the environmental conditions and the concrete (type of binder) for a particular concrete. α varies between 0.5 and 0.85. In some models, cf. Maage & Helland (1991) and Poulsen (1995), a separate expression is used to estimate the Daex at the time of exposure from laboratory test data

β

⎟⎠

⎞⎜⎝

⎛=

tt

DD expaex [5.3.2c]

where Dp is a “potential diffusion coefficient” determined from a test method at different ages. The “aging exponent” α is then the sum of two parts β and γ. Τhe aging exponent β is due to concrete aging and the aging exponent γ when exposed to sea water is possibly

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 36 of 112 WP4 REPORT – Modelling of Chloride Ingress explained by ion exchange blocking the concrete surface, Maage & Helland (1991), Mohammed et al (2002).

5.4.3 Required Input data Required input data is the apparent diffusion coefficient with its time-dependency Da(t) and the constant surface chloride content Csa.

5.4.4 Input relation(s) to test methods The apparent diffusion coefficient Daex at the time of exposure can be estimated from a short term immersion test. The exponent α, however, must be determined from field data from appropriate environmental conditions and similar binder types.


5.4.6 Model(s) for environmental actions Environmental actions are translated into a surface chloride content Csa mainly from exposure tests in the true environment, for the concrete that is going to be used. Some models for Csa are available, i.e. Lindvall (2003), where Csa can be estimated from the concrete composition (type and amount of binder, w/c etc.), and various environmental parameters, such as sea water temperature and salinity, distance to sea level, distance to road surfaces etc.


5.4.8 Openness/Practicality The model is simple and fully transparent for any user.

5.4.9 Interpretation of real behaviour Today it is obvious from numerous field observations that empirical models require a time-dependent achieved diffusion coefficient Da.

5.4.10 Advantages The main advantage is the simplicity in performing the predictions and that there is a possibility to estimate the input data from testing.

5.4.11 Limitations & Drawbacks Besides the general limitations and drawback for empirical models, see section 5.9, the main limitation and drawback is the lack of understanding of why the apparent diffusion coefficient is time-dependent, besides the obvious effect of densification of concrete during hardening. This lack of understanding gives a large uncertainty in the predictions, since there is a risk that the time-dependency does not continue to follow empirical equations like Equation [5.3.2b], e.g. the apparent diffusion coefficient might stop to decrease after some time.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 37 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.4.12 References Maage & Helland (1991), Bamforth (1993, 2004), Maage et al (1995), Maage et al (1999), Mangat (1994), Mohammed et al (2002), Bjergovic (1994), Poulsen (1995), Lee & Christholm (2000)


5.5 Models A1c: ERFC-model, D(t) & Csa (The DuraCrete model)

Time-dependent “true” chloride diffusion coefficients determined from tests started to be used in the late 1990’s, mainly in the DuraCrete project. By including a number of “correction factors” correlation with field data was ensured. The uncertainties of all parameters are quantified and used as input into the model. 5.5.1 Principle The principle of this model is: 1) Determine a chloride diffusion coefficient for the particular concrete with a short term test

method at a certain age, i.e. 28 days. 2) Calculate an “effective diffusion coefficient” after a certain time of exposure by correcting

the determined diffusion coefficient for the effect of test method, curing, environment and aging during exposure.

3) A (constant) “apparent surface chloride content” Csa, some of the correction factors and the exponent in the time-dependency of the diffusion coefficient are estimated from field exposure of similar concretes in similar environments.

4) Predict the chloride ingress after a certain exposure time, the chloride profile, from the erfc-solution to Fick’s 2nd law, with Csa and the “effective diffusion coefficient” at that age.

5.5.2 Mathematics The original DuraCrete model could be expressed as a model for chloride ingress utilizing these equations

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅⋅=

tDxCtxC

e

sa 2erfc),( [5.5.2a]

n

ceRCMte tt

kkDkD ⎟⎠⎞

⎜⎝⎛⋅⋅⋅⋅= 0

0, [5.5.2b]

- DRCM,0 is the chloride migration coefficient under defined compaction, curing and

environmental conditions, measured at time t0 [m2/s]; - n is an exponent which gives the time-dependency of the effective diffusion coefficient [-]; - kt is a factor which transfers the measured chloride migration coefficient DRCM,0 into a

chloride diffusion coefficient D0 (=kt·DRCM,0) [-]; - ke is a factor which considers the influence of environment on D0 [-]; - kc is a factor which considers the influence of curing on D0 [-]; - Csa is the surface chloride level in [wt % Cl-/binder]; - t is the exposure period [years]; and - t0 is the reference period [years], in this case t0 = 28 days

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 39 of 112 WP4 REPORT – Modelling of Chloride Ingress Later, a modification of the DuraCrete model has been done, Gehlen (2000), by “excluding” a “convection zone” and calculating the chloride penetration from a “peak” in the profile at a certain depth Δx, cf. Figure 1.2,

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅Δ−

⋅= Δ tDxxerfcCtxC

exsa 2

),( , [5.5.2c]

where Δx is in the order of 6-11 mm. This of course means that the modified model only is relevant for long exposure times where this convection zone has been built up and remained more or less constant for a long period of time. The time required to build up such a zone, and the effect of the resistance of the zone, is assumed to be insignificant compared to the total exposure time and the resistance of the rest of the concrete cover.

5.5.3 Required Input data The model has at least 6 independent input parameters, most of which are determined from field data from similar concretes in similar environments.

5.5.4 Input relation(s) to test methods The migration coefficient DRCM,0 is a direct test result from a test method. Parameters n, kt and kc may be estimated from the same laboratory test method.


5.5 Model(s) for environmental actions Environmental actions are translated into a surface chloride content Csa mainly from exposure tests in the true environment, for the concrete that is going to be used. Some models for Csa are available, i.e. Lindvall [2003] originating from the DuraCrete project, where Csa can be estimated from the concrete composition (type and amount of binder, w/c etc.), and various environmental parameters, such as sea water temperature and salinity, distance to sea level, distance to road surfaces etc.


5.5.8 Openness/Practicality The model is simple and fully transparent for any user. It is very straight-forward, but there are some questions on the definitions of some of the parameters, cf. Section 3.8.

5.5.9 Interpretation of real behaviour Since the factors are quantified from available field data the model gives prediction results that coincide with real behaviour. However, the scatter in available data for that quantification was extremely large.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 40 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.6 Advantages The model is simple and very straight-forward with input parameters quantified from separate tests or field exposure data from similar conditions. The model parameters can also be updated in a simple way when new and better field data is available.

5.6.11 Limitations & Drawbacks Besides the general limitations and drawback for empirical models, see section 5.9, the mathematics of the model could be questioned. Equation [5.5.2a] should have the apparent diffusion coefficient Da, instead of the effective diffusion coefficient De, or correction factors according to Equation [3.8.2e]. If not, those correction factors must be included in the environmental factor kt. However, the correction factor in Equation [3.8.2e] includes the exponent n and considers the difference between age and exposure time. Additionally, if the time t in Equation [5.5.2b] is the exposure time, the effective diffusion coefficient is not defined at the start of the exposure. The model must be revised to correctly consider the time-dependency of the diffusion coefficient.

5.5.12 References Alisa et al (1998), Gehlen & Ludwig (1999), Siemes et al (1999), Gehlen (2000), Visser et al (2002), Gulikers (2004).


5.6 Models A1d: ERFC-model, Da(t) & Csa(t) (The False erfc-model)

Some observations in the last few years indicate that the surface chloride content is time-dependent, also for submerged concrete! This model is the simplest empirical model to consider that time-dependency. The model, however, is purely empirical and not the correct mathematical solution to Fick’s 2nd law. 5.6.1 Principle The principle of this model is: 1) Estimate the time-dependency of the “apparent diffusion coefficient” Da(t) and the time-

dependent “apparent surface chloride content” Csa(t), from field exposure tests. 2) Predict the chloride ingress after a certain exposure time t, the chloride profile, from the

erfc-solution to Fick’s 2nd law, with constant parameters Da(t) and Csa(t).

5.6.2 Mathematics The model is based on three simple equations

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅⋅=

ttDxerfctCtxCa

sa )(2)(),( [5.6.2a]

α

⎟⎠⎞

⎜⎝⎛=

tt

DD exaexa [3.8.1], [5.3.2b] & [5.6.2b]

BtttAtC inisa +−+Δ⋅= )ln()( exp [5.6.2c]

where t is the age, texp is the age at exposure and Δtini is an initial binding period (14/365 years). A and B are regression parameters.

5.6.3 Required Input data Required input data are the apparent diffusion coefficient with its time-dependency Da(t) and the time-dependent surface chloride content Csa(t).

5.6.4 Input relation(s) to test methods The apparent diffusion coefficient Daex at the time of exposure can be estimated from a short term immersion test. The exponent α, however, must be determined from field data from appropriate environmental conditions.


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 42 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.6.6 Model(s) for environmental actions Environmental actions are translated into a surface chloride content Csa mainly from exposure tests in the true environment, for the concrete that is going to be used. Models for time-dependent surface chloride contents are available to some extent, from the Hetek-project, see section 5.7.


5.6.8 Openness/Practicality The model is very simple and fully transparent for any user. It is very straight-forward, but there are some questions on the relevance of predictions beyond observation data.

5.6.9 Interpretation of real behaviour Since the model is based on observations for a particular concrete in a particular environment the model gives the real behaviour within the observation time. Beyond that, the model seems to underestimate chloride ingress!

5.6.10 Advantages The model is very simple and very straight-forward with input parameters quantified from field exposure data from true conditions. The model parameters can also be updated in a simple way when more field data is available from longer exposure times. It is a true “empirical” model!

5.6.11 Limitations & Drawbacks Besides the general limitations and drawback for empirical models, see section 5.9, one strong limitation is the lack of mathematical correctness, since the equation [5.6.2a] is not the true solutions to Fick’s 2nd law with the boundary conditions in Equation [5.6.2c]! That gives the underestimation of chloride ingress.

5.6.12 References Nilsson (2002)


5.7 Models A2: Analytical solutions, DF2(t) & Csa(t) (The Mejlbro-Poulsen & Hetek model)

Some observations in the last few years indicate that the surface chloride content is time-dependent, also for submerged concrete! With those boundary conditions, the erfc-equation is no longer a correct solution to Fick’s 2nd law. A completely new equation must be used. 5.7.1 Principle The principle of this model is:

1) Estimate the time-dependency of the “apparent diffusion coefficient” Da(t) and the time-dependent “apparent surface chloride content” Csa(t), from field exposure tests.

2) Predict the chloride ingress after a certain exposure time t, the chloride profile, from the Mejlbro-Poulsen-solution to Fick’s 2nd law, with time-dependent parameters D(t) and Csa(t).

5.7.2 Mathematics Chloride ingress is predicted from this set of equations

⎟⎟⎠

⎞⎜⎜⎝

⎛

⋅Ψ⋅=

ttDxttCtxCa

psa )(2),(),( exp [5.7.2a]

where α

⎟⎠⎞

⎜⎝⎛=

tt

DtD exaexa )( [5.7.2b]

and p

aexisa tt

DttSCttC ⎟⎟

⎠

⎞

⎜⎜

⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

⎛××−×+=

αexp

expexp )(),( [5.7.2c]

where the exponent p depends on how fast Csa(texp,t) increases with time, i.e. mainly on the type of binder and the environment, cf. Figure 5.7.1.

0.0

0.2

0.4

0.6

0.8

1.0

0 10 20 30 40 50 60 70 80 90 100

Time of exposure [years]

Surfa

ce c

hlor

ide

cont

ent

p = 1

p = 0.05

Fig. 5.7.1 An example of the time-dependency of Csa(texp, t) according to Equation [5.7.2b]

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 44 of 112 WP4 REPORT – Modelling of Chloride Ingress Examples of the Mejlbro-Poulsen solutions to Fick’s 2nd law with time-dependent diffusion coefficient and surface chloride content are shown in Figure 5.7.2.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0 0.5 1.0 1.5 2.0Abscissa: z

Ψp (z ) for -0.4 ≤p ≤1; Δ p = 0.2

p = -0.4

p = 1.0

Fig. 5.7.2 Graphs of typical Mejlbro-Poulsen Ψπ-functions.

For p=0 (equal to Csa being constant) the functions are identical to the erfc-equation.

5.7.3 Required Input data Parameters texp, Daex, α and p are required to make a prediction of chloride ingress. In order to make the input data physically understandable, the input can as well be estimates of C1, C100, D1 and D100.

5.7.4 Input relation(s) to test methods The parameters Cspex and Dpex can be quantified by an immersion test and translated into values of C1 and D1.

5.7.5 Required boundary conditions Field observations must be used to quantify the time-dependency of the surface chloride content Csa(t). For concretes and environments close to the ones already studied, excellent models are available.

5.7.6 Model(s) for environmental actions A set of simple models are derived to translate environmental actions into efficiency factors influencing the surface chloride content and the diffusion coefficient, considering the concrete composition and the exposure conditions.

5.7.7 Output data The prediction gives a chloride profile similar to the erfc-solution for each exposure time.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 45 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.7.8 Openness/Practicality All details of the model have been published. The mathematics involved is not simple to quantify, however, and software must be made commercially available. Even so, for ordinary engineers it must be regarded as a ”black box”.

5.7.9 Interpretation of real behaviour Since the parameters are quantified from available field data from one exposure site the model gives prediction results that coincide with real behaviour, limited to that site. However, the parameters can easily be updated if good field data are available from other environments.

5.7.10 Advantages One important advantage is that the model is the true solution to Fick’s 2nd law with time-dependent diffusion coefficient and time-dependent surface chloride content. Another advantage is the models for estimating all the parameters from the concrete composition and the environmental conditions.

5.7.11 Limitations & Drawbacks The limitations are very well described by the model developpers: “The Mejlbro-Poulsen Model do not model the actual physical and chemical processes involved in the chloride ingress into concrete. The aim of the model is to describe the result of these physical and chemical processes of this transport, i.e. the chloride profiles. The model is dependent of good measurements from the environment and the concrete type in question. The measurements must represent a fairly long time of exposure to improve the predictions. These drawbacks will be overcomed with time when the experience grow. The model is very sensitive to the quality of the measurements therefore many of the existing measurements must regarded as unsuitable for prediction purposes with this model” (Frederiksen et al (1997)). An obvious drawback with the model is the complicated mathematics with two different diffusion coefficients DF2 and Da, both being time-dependent.

5.7.12 References Mejlbro (1996), Poulsen (1996), Frederiksen et al (1997).


5.8 Models A3a and A3b: Numerical, DF2(t) & Csa(t) (NIST, LEO, Life-365, etc.))

Solutions to Fick’s 2nd law with any time-dependency of the diffusion coefficients and the surface chloride contents can easily be found numerically. The most elaborated one is Life-365. 5.8.1 Principle The principle of this model is: 1) Estimate the time-dependency of the diffusion coefficient in Fick’s 2nd law DF2(t) and the,

possibly time-dependent, “apparent surface chloride content” Csa(t), from a chloride binding isotherm (models A3a) or from field exposure tests (models A3b).

2) Predict the chloride ingress after a certain exposure time t, the chloride profile, from

numerical solutions to Fick’s 2nd law, with time-dependent parameters D(t) and Csa(t).

5.8.2 Mathematics Fick’s 2nd law is used as a pure mass balance equation, the diffusion coefficient DF2(t) describing the flux of chlorides with the total content of chloride as the driving potential.

2

2

2 xC

Dt

CF ∂

∂⋅=

∂

∂ [5.8.2a]

The diffusion coefficient is time-dependent and temperature dependent

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛−⋅⋅⎟⎟

⎠

⎞⎜⎜⎝

⎛⋅=

TTRU

tt

DTtDref

mref

refF11exp),(2 [5.8.2b]

The boundary conditions are described as Csa(t), in Life-365 increasing at a constant rate up to a certain maximum level after a defined time. Any other description could be used, in principle. The time dependency according to equation [5.8.2b] is used in Life-365 for a limited exposure time only, up to some 25 years. After that, the diffusion coefficient is treated as a constant. Solutions to the mass balance equation with these boundary conditions are found numerical with finite difference methods.

5.8.3 Required Input data Required input data is the diffusion coefficient in Fick’s 2nd law, with its time-dependency DF2(t) and the time-dependency of the surface chloride content Csa(t).

5.8.4 Input relation(s) to test methods Part of the time-dependency of the diffusion coefficient may be derived from laboratory test. The environmental effect, however, must be found from observations.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 47 of 112 WP4 REPORT – Modelling of Chloride Ingress 5.8.5 Required boundary conditions Field observations must be used to quantify the time-dependency of the surface chloride content Csa(t). For concretes and environments close to the ones already studied, reasonable estimates should be possible. In Life-365 the Csa-values are independent of the concrete composition.

5.8.6 Model(s) for environmental actions The boundary conditions, the Csa-values, are completely derived from observations. Field data is translated into a maximum surface chloride content, with a certain rate of accumulation during the early exposure.

5.8.12 Output data The prediction gives a chloride profile similar to the erfc-solution for each exposure time.

5.8.8 Openness/Practicality Life-365 is widely distributed and tested and a manual is published that describes the details of the software and the background for the selection of default input data. The user can freely change all input data, if better data is available. Life-365 could, consequently, be regarded as a mathematical tool. That could be used for various cases by an educated user. Similar models can easily be built by using finite difference methods for any set of assumptions.

5.8.9 Interpretation of real behaviour Comparisons have been made with laboratory and field data, mainly in quantification of the input data, the parameters of the diffusion coefficient and the surface chloride content.

5.8.10 Advantages The main advantage with the Life-365 model is its user-friendliness, especially the data-base covering environmental conditions around North America.

5.8.11 Limitations & Drawbacks Besides the general limitations and drawback for empirical models, see section 5.9, there are uncertainties on the definitions and quantification of the time-dependency of the diffusion coefficient. The exponent m in Equation [5.8.2b] is mostly quantified from observations of the time-dependency of the apparent diffusion coefficient, but in other cases from measurements on old concrete, not exposed to chloride. In Equation [5.8.2a] and [5.8.2b] the diffusion coefficient is not the apparent one. Consequently, the time-dependency should be different, cf. Section 3.8. This must be better clarified and defined.

5.8.12 References Bentz & Thomas (1999), Petre-Lazar et al (2000)


5.9 Conclusions on models based on Fick’s 2nd law General conclusions on empirical models are

1. The effect of the other ions in the pore system is completely neglected in empirical models solving Fick’s 2nd law.

2. The meaning of regression parameters in empirical models is not always easy to understand. They must be very clearly defined.

3. The background for a number of the assumptions made in empirical models could be questioned. This is especially important for the continuous time-dependency of the diffusion coefficients and the surface chloride contents.

4. Empirical models need huge data bases: a Cs for every concrete X, every road Y and every ”water” Z!

Conclusions 1-3 give rise to questions on whether empirical models can be used at all for predictions beyond where data exists. Conclusion 4 gives rise to questions on whether empirical models will ever be practical to use for a new structure, made of a new concrete exposed in a new environment!


6 CHLORIDE INGRESS MODELS BASED ON FLUX EQUATIONS

6.1 General In “physical models” all physical and (electro)chemical processes are described as scientifically correct as possible. True physical models use independently determined input data and no curve fitting to exposure data. Instead, field exposure data is used to validate predictions. If the comparison between predicted results and exposure data is not good enough, the model must be improved, or better data must be determined. Sophisticated physical prediction models for chloride ingress into concrete contains at least two mass balances, and several relations, for chloride and for water, see Figure 3.

Cf

Cb

wvRHvapour

freechloride

liquidwater

boundchloride

Cf

Cb

wvRH

Cf

Cb

wvRHvapour

freechloride

liquidwater

boundchloride

vapour

freechloride

liquidwater

boundchloride

Figure 6.1.1 The mass balance and interaction of chloride and water, in different phases.

The decisive concepts in physical models are concepts in chloride transport and in chloride binding. Required input parameters for chloride transport in physical models are several. The diffusion coefficients should be described at least as functions of temperature, moisture content, degree of hydration, depth from the surface etc. The description of the convection term requires the liquid water flow to be described.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 50 of 112 WP4 REPORT – Modelling of Chloride Ingress 6.3 Boundary conditions concepts in physical models The description of boundary conditions in physical models are much more complicated, since these models require boundary conditions in terms of chloride, temperature, humidity conditions at the very surface of the concrete: C(x=0, t), W(x=0, t), T(x=0,t). An obvious question is: Who’s “responsibility”? Material researchers that develop sophisticated physical models will find no use of them if quantitative boundary conditions are not available. For the submerged zone of marine structures it is fairly simple, but for all other environments it is extremely complicated. The surface conditions at a particular concrete surface must be regarded as dependent on the location, orientation, distance, time etc. We have very little data where the same concrete was exposed in two locations. Every researcher has been exposing his concretes in his own “back yard” with no possibilities to compare data with another researcher. Additionally we usually miss proper documentation of the exposure environment. A “splash zone” is not equal to a “splash zone” somewhere else! A possible approach to make information on boundary conditions for physical models available in the future is to describe the environmental actions in steps from meteorological data up to the surface:

• Regional Climate • Local Climate • Location of the structure • Distance from the source of chloride to the concrete surface • Orientation of the surface

This approach has been analysed for the road environmental actions, cf. Figure 6.2.1. On the regional scale the macro climatic actions are the ones without a road at all! On the meso scale, the environmental actions are the one from the road surface, without even considering the presence of a concrete structure! At the micro scale the effect of the location, the size and the shape of the concrete structure is considered to give the actual environmental actions at a particular concrete surface.

Macro Climate

Air TemperatureAir Humidity

Ground TemperatureRadiation

Rain Wind

Meso (Road)Climate

Rain water splashSalt water splash

Rain water fogSalt water fog

Micro (Surface)Climate

Equivalent airtemperatureSurface RH

WetnessSalt concentration

Traffic intensityand speed

Road geometryRain shelterAir streamsFrequency of de-icing

Response by the structure

T(x,z.t)RH(x,z,t)Cl-(x,z,t)

DistanceHeight

OrientationShape

Surface roughness

Macro Climate

Air TemperatureAir Humidity

Ground TemperatureRadiation

Rain Wind

Meso (Road)Climate

Rain water splashSalt water splash

Rain water fogSalt water fog

Micro (Surface)Climate

Equivalent airtemperatureSurface RH

WetnessSalt concentration

Traffic intensityand speed

Road geometryRain shelterAir streamsFrequency of de-icing

Response by the structure

T(x,z.t)RH(x,z,t)Cl-(x,z,t)

DistanceHeight

OrientationShape

Surface roughness

Fig. 6.2.1 An approach to stepwise quantify the environmental actions at a concrete surface in a road environment, starting from meteorological data.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 51 of 112 WP4 REPORT – Modelling of Chloride Ingress Without the proper boundary conditions available, we are limited to observations of the response by concrete in various environments and during various conditions. We do have some information on the comparison between marine and road climates, cf. Figure 6.2.2.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 10 20 30 40

Depth [mm]

Cl [

wt-%

of c

oncr

ete]

3-50 U1

3-50 P1

3-50 A1

207 BK1

w /B=0.50, 5%Si; 2 years

Rv40

0.1

0.2

0.3

0.4

0.5

0.6

0 10 20 30

Depth [mm]

Cl [

wt-%

of c

oncr

ete]

H4 U1

H4 P1

H4 A1

206 DK6

w /B=0.40, 5%Si; 2 years

Rv

0.0

0.1

0.2

0.3

0.4

0 10 20 30 40 50 60Depth [mm]

Chl

orid

e [w

t-% o

f sam

ple] 2-75 488d Subm.

2-75 488d Splash2-75 488d Atm.236CK6 600d Road

Rv40

Träslöv

Figure 6.2.2. Chloride profiles for a SRPC concrete with w/b of 0.50 (top left), 0.40 (top right) and 0.75 (bottom) exposed to various chloride environments, a marine submerged (upper curves) and atmospheric and road environment (lower curves).

The build-up of chloride profiles during one season is shown in Figure 6.2.3 for the same concrete as in Figure 6.2.2 (bottom). The two figures show the response by the vertical and horizontal surfaces respectively.

236 D (3-6)F

0.00

0.05

0.10

0.15

0.20

0.25

0 10 20 30 40 50 60

Depth (mm)

Cl [

% p

er S

ampl

e]

Feb

March

April

May

Oct

236 D (3-6)Ö

0.00

0.05

0.10

0.15

0.20

0.25

0 10 20 30 40 50 60

Depth (mm)

Cl %

Per

Sam

ple

Feb

March

April

May

Oct

Figure 6.2.3. Chloride build-up and washout during the first winter and summer of the vertical (left) and horizontal (right) surface of a concrete with w/c=0.75 in a road climate.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 52 of 112 WP4 REPORT – Modelling of Chloride Ingress Even though the chloride profiles have a certain uncertainty, some ±20 %, which makes comparisons difficult, some indications could be pointed out. Some differences are clearly larger than the uncertainty. During the winter the chloride content in the surface near region is very large. The build-up seems to be more rapid for the vertical surface, but later during the winter, the horizontal surface gets a larger depth of penetration. This could be due to a longer time-of-wetness for horizontal surfaces, promoting the diffusion and convection of chloride further into the concrete. On the other hand, the wash-out is also more significant for the horizontal surface, more or less for the same reason. After the first winter and summer the chloride profiles are similar for the two orientations, in spite of the differences during the previous winter. In the future, that kind of processes must be described if we want to use physical prediction models for complicated environments. For the time being, these difficulties favour the use of empirical models, because of the simple description of the boundary conditions, a Cs-value. That Cs for every individual concrete could be translated into an “equivalent” chloride concentration at the concrete surface. Then that could be used as boundary conditions for a new concrete, without having to perform an exposure program.


6.3 Models B1a: Models based on Fick’s 1st law of diffusion, no convection (ClinConc, etc.)

One of the most advanced models for predicting chloride penetration into concrete, ClinConc, was presented by Tang & Nilsson [1994] and Tang [1996]. By using a FDM numerical approach, most of the factors involved in chloride penetration are considered in a relevant and scientific way. 6.3.1 Principle The chloride flux is described with the chloride diffusion coefficient in Fick’s 1st law DF1, calculated from a separate migration test at a certain age. The effect of densification of concrete on the diffusion coefficient is included up to a certain age, i.e. six months for Portland cement concrete. Chloride binding is described by a binding isotherm that is a function of pH and temperature. Leaching of alkalis is described as a pure diffusion of hydroxides out of the concrete and the pH- profile is part of the prediction in every time-step. Only the alkali hydroxides are dealt with, e.g. pH remains larger than 12.5. The mass balance equations for chlorides and hydroxides are solved by using separate terms for chloride diffusion and for chloride binding. In every time step the fluxes of chloride is calculated at each depth with Fick’s 1st law. From differences in fluxes the change in total chloride content is calculated at each depth. By using the binding isotherm, the change in total chloride content is divided into free and bound chloride. The chloride ingress is shown as profiles of free and total content of chloride. Both semi-infinite cases and cases with limited thicknesses can be treated.

6.3.2 Mathematics The diffusion coefficient Di,j in Fick’s 1st law is a function of temperature, age and depth:

( ) ( ) ( )D D f T g t f xi j T j j i, ( )Cl 0 D0

= [6.3.2a]

The age dependency of the diffusion coefficient is calculated from

g ttt

t t

t t( ) =

⎛⎝⎜

⎞⎠⎟

<

≥

⎧

⎨⎪

⎩⎪

00

01

β t

[6.3.2b]

The total amount of chloride is split into free (in a moisture content Ws) and bound


)bound()()total( ,,,, jijisjiji CWcC += kg

mCl

concrete3

⎡

⎣⎢

⎤

⎦⎥ [6.3.2c]

where the binding isotherm is expressed by

[ ][ ] B

jib

jiTTR

E

ji cf

WeeC j

b

j

ji

,,

11OH

OH1

, 1000gel)()bound( 01,initial

1,OH ⎟

⎟⎠

⎞⎜⎜⎝

⎛−⎟

⎟⎠

⎞⎜⎜⎝

⎛−

−

−

=α

[6.3.2d]

6.3.2 Required Input data Input data to the model is the complete concrete composition, the diffusion coefficient with its time-dependency and the chloride binding isotherm as a function of pH and temperature. The basic diffusion coefficient is the “true” chloride diffusion coefficient DF1, evaluated from a migration test method. Additionally, the effect of depth from a cast surface can be taken into account. The effect of age is described with a time-dependency of the diffusion coefficient up to a certain age. For a new Portland cement concrete only one parameter has to be determined, the diffusion coefficient for the particular concrete at a certain age. The other parameters in the model follow from the mix composition and the exposure conditions. For a new binder, the parameters of the binding isotherm must be given.

6.3.4 Input relation(s) to test methods The diffusion coefficient at a specified age is determined in an independent, short-term test with no curve-fitting is involved.

6.3.5 Required boundary conditions The boundary conditions can described as sinusoidal annual variation of concentration of chloride and water temperature in the surrounding chloride solution.

6.3.6 Model(s) for environmental actions In later version, the chloride concentration at the surface can be varied in time stepwise.

6.3.7 Output data The predictions give three profiles: free and total concentration of chloride and concentration of alkalis. They are shown graphically and in tables. The profiles do not always have an erfc-shape, depending on temperature and leaching conditions.

6.3.8 Openness/Practicality The software is not yet commercially available but spread among researchers. All details, however, are thoroughly published. The user-friendliness makes the model extremely practical. For a new concrete only one test has to performed, to determine the diffusion coefficient at a certain age.

6.3.9 Interpretation of real behaviour A number of comparisons have been made with laboratory and field data. The predictions for a number of concretes submerged in sea water was once made with the independently

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 55 of 112 WP4 REPORT – Modelling of Chloride Ingress determined diffusion coefficients The predictions coincided very well with the field data from up to two years of exposure. Then, for some concretes, the measured chloride profiles had larger surface chloride contents than the predictions. A time-dependency of the chloride binding had to be empirically included to obtain a better correlation. This time-dependency of chloride binding, however, remains to be explained and independently quantified.

6.3.10 Advantages The most significant advantage is the small number of input parameters and that the only one that has to be determined, can be quantified in a separate, independent test. Another advantage is that the effect of temperature variations and leaching can be considered in a physical way. Recently, ClinConc has been expressed in an engineering way for convenient applications (Tang 2005).

6.3.11 Limitations & Drawbacks The physical part of the model can only treat cases with saturated concrete exposed in submerged conditions. In later versions empirical parts have been added that combines stepwise changed chloride concentrations in the surrounding environment. The concrete, however, still is regarded as saturated. The exclusion of the effects of all ions on the chloride transport process gives rise to questions on the ability to describe future ingress in a correct way.

6.3.12 References Tang (1996, 2005)


6.4 Models B1b: Models based on Fick’s 1st law of diffusion, with convection (Imperial, HetekConv, ConFlux, THI EdF, LERM, Meijers, EPFL etc.)

A number of similar physical models are available where a convection term has been added, to be able to predict chloride ingress in a number of applications where moisture flow and wetting and drying play a dominating role. These models have different simplifications and assumptions, but they are here described together. Examples of details are taken from the Hetek Convection model, partly because it is fairly elaborated and partly because it is best known by the author(s). 6.4.1 Principle The convection of chloride is governed by the moisture flow and the changes in moisture contents will change the chloride concentrations. Consequently, the moisture distributions must be predicted by solving the mass balance equation for moisture. The moisture flow must be separated in two different fluxes, only one carrying the ions. The mass balance equation for chloride, and the separation into free and bound chloride, can be described by taking the moisture content into account, since the free chlorides are limited to the water filled parts of the pore system. The two mass balance equations are coupled and must be solved together. The main part of the solution is to describe the flows of moisture and chloride. The initial conditions are trivial but the boundary conditions must be, until better data are available, a simplification of the true environmental conditions at the concrete surface.

6.4.2 Mathematics The moisture flow is divided into vapour flow and liquid flow, i.e. with vapour content and pore water pressure as driving potentials, respectively

xP

wkxvRHqqq w

Pvlvw ∂∂

−∂∂

−=+= )()(δ [6.4.2a]

The chloride flux is described by Fick’s 1st law, in the water filled part of the pore system, e.g. the diffusion coefficient is a function of the moisture content, and one term giving the convection of chloride with the liquid water flow

q D w T ccx

q cCl Cl ff

l f= − +( , , )''

'∂

∂ [6.4.2b]

A number of relationships must be expressed mathematically, i.e. by considering that the moisture sorption depends on the chloride concentration and by describing the separation of chloride into free and bound chloride, possibly with a non-linear binding isotherm, including the moisture content in the pore system. This is all done differently in the available models.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 57 of 112 WP4 REPORT – Modelling of Chloride Ingress 6.4.2 Required Input data Required input data for this kind of physical models, at least the most sophisticated ones, is tremendous! The chloride and moisture flow coefficients must be given as functions of a number of parameters. A number of relationships between all parameters require a large number of input parameters to be quantified. The input data for boundary conditions require the duration of dry and wet periods at the concrete surface and the temperature and humidity conditions at the very surface during these periods. Additionally boundary conditions must be described as concentrations of free chloride in the solution in contact with the concrete surface, when it is present.

6.4.4 Input relation(s) to test methods The input data for the material properties and relationship can all be determined by test methods. Those test methods, however, are not yet standardized and in some cases not yet invented! As long as this input data is missing to a large extent, physical models can only be used in very special cases where data exists.

6.4.5 Required boundary conditions The boundary conditions at the concrete surface must describe the environmental actions as a function of time in terms of time-varying chloride concentration in a solution in contact with the surface, alternating with drying conditions.

6.4.6 Model(s) for environmental actions For non-trivial cases no models exist today that can translate environmental actions in practice to the required boundary conditions at the concrete surface in physical models.

6.4.7 Output data Physical models of this kind predict the moisture profiles and the profiles of free and total concentration of chloride.

6.4.8 Openness/Practicality Physical models are rarely commercially available or user-friendly at all. In most cases the only possible user is the developer himself! Additionally, the models must treat so many important details in a correct way that it is very difficult for anyone else to evaluate the accuracy and relevance of a physical model. A number of models, however, are very well documented and all details are thoroughly published. The large amount of required input data is not easily available but have to be quantified by laboratory measurements and complicated calculations for every new case. This makes the general practical application of physical models very limited.

6.4.9 Interpretation of real behaviour All models “produce” nice results and frequently predictions are fitted to measured data by adjusting a number of all the input parameters.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 58 of 112 WP4 REPORT – Modelling of Chloride Ingress However, good data that can be used for comparisons is scarce. Simultaneously measured moisture and chloride profiles must be made available, where micro climatic data at the concrete surface and all relevant material properties are well documented.

6.4.10 Advantages Physical models of this kind are the only ones capable of handling cases where moisture flow plays a dominant role. Examples are concrete structures with limited thickness that are exposed to chloride on one side only (tunnels, pool walls, foundation walls), structures exposed to drying and wetting, sometimes with splash from de-icing salts (road structures) or sea water (marine splash and atmospheric zones). They are also capable of handling chloride transport in material combinations.

6.4.11 Limitations & Drawbacks The tremendous amount of required input data, material properties and environmental conditions, most of it as functions of the other parameters, makes the practical application very doubtful. The relevance of all assumptions and limitation is not very well known or quantified. The exclusion of the effects of all ions on the chloride transport process gives rise to questions on the ability to describe future ingress in a correct way.

6.4.12 References Buenfeld (1995), Nilsson (1997, 2000), Hooton, Thomas et al (199x), McLoughlin I.M (1997), Petre-Lazar et al (2003), Meijers (2003), Denarié et al (2003).


6.5 Models B2a, b and c: Physical models based on the Nernst-Planck flux equation (MS-Diff, Li & Page, Johannesson, Stadium)

6.5.1 Principle The fluxes of all ions in the system are predicted by considering how they influence each other by creating an electrical field in the pore solution. This electrical field must be predicted simultaneously. The profiles of each specie are then predicted by solving the mass-balance equations for all ions.

6.5.2 Mathematics The models must describe the flux of every individual type of ion i, which is done with the Nernst-Planck equation

⎟⎠⎞

⎜⎝⎛

∂Φ∂

+∂

∂+

∂∂

−=x

cRT

Fzxa

cxc

Dq iii

ii

iiln

[3.8] & [6.5.2a]

In some of the models a convection term is added to consider the effect of liquid moisture flow. Some models even include the effect of degradation and micro structural changes that will increase the diffusion coefficients over time. The flux equations include the electrical field. This electrical field is created by the species of opposite sign, so that the faster ions will be slowed down and the slower ones will be accelerated. A general expression for the electrical field, Masi et al (1997), is given by:

∑∑ ∂

∂+

−=∂Φ∂

iiii

i

iii

txcDzx

txcDz

SFtI

FRTtx

x ),(

),()(

),( 2 [6.5.2b]

where I(t) is the electrical current and S is the cross-section of the material. The Poisson equation may be used instead of Eq. [6.5.2.b] in order to determine the electrostatic potential (Marchand et al, 2002):

0),(²

²=⎥

⎦

⎤⎢⎣

⎡+

∂Φ∂ ∑

N

ii txczFx ε

[6.5.2.c]

where N is the total number of ionic species and ε is the dielectric permittivity of the medium. For every type of ion the mass-balance equation is solved

ttxC

xtxq

ttxc

p ibiisol ∂

∂−

∂∂

−=∂

∂ ),(),(),( , [6.5.2c]

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 60 of 112 WP4 REPORT – Modelling of Chloride Ingress where the right-hand term gives the binding/interaction between species i and the matrix, e.g. chloride binding or some kind of reaction or liberation.

6.5.3 Required Input data Required input data for this kind of physical models, is tremendous, especially if convection is considered! The ion diffusion coefficients and moisture flow coefficients must be given as functions of a number of parameters. A number of relationships between all parameters require a large number of input parameters to be quantified. The input data for boundary conditions require the duration of dry and wet periods at the concrete surface and the temperature and humidity conditions at the very surface during these periods. Additionally boundary conditions must be described as concentrations of ions in the solution in contact with the concrete surface, when it is present! Parameters of the binding/interaction term must be quantified separately. Different models use various ways to consider e.g. chloride binding. Truc (2000) and Li & Page (1998) used a chloride binding isotherm (models type B2a), Johannesson (2000) described chloride binding as a special ion-solid interaction process (models type B2b) and Samson et al (1999) concentrated on the chemical fixation of chloride (models type B2c).

6.5.4 Input relation(s) to test methods The diffusion coefficients Di for each type of ion cannot be determined in a separate test, since the flux of one ion depends on what other ions are present and at what concentrations. Samson et al (1999) overcame this obstacle in a suitable way by determining a “formation factor” for the pore system with another test method and then simply multiplied the formation factor with the diffusion coefficient for a type of ion in solution. If this is correct, a test method for the formation factor is required. Khitab et al (2004) proposed to measure the chloride diffusion coefficient from a migration test (LMDC-test.) The other ionic species diffusion coefficients are linked to the chloride one by the same relation as in an infinitely diluted solution.

6.5.5 Required boundary conditions The boundary conditions at the concrete surface must describe the environmental actions as a function of time in terms of time-varying chloride concentration in a solution in contact with the surface, alternating with drying conditions.

6.5.6 Model(s) for environmental actions For non-trivial cases no models exist today that can translate environmental actions in practice to the required boundary conditions at the concrete surface in physical models. In reality this means that only simple boundary conditions can be treated, such as direct contact with a salt solution or drying in an atmosphere with a given humidity.

6.5.7 Output data Physical models of this kind predict the moisture profiles, if convection is considered, and the profiles of free and total concentration of all ions. An example of a prediction of the ion profiles and ion fluxes and the electrical potential are shown in Figure 6.5, for a 3 cm thick concrete specimen in a migration test.


Fig. 6.5 Predicted concentration profiles and fluxes for chloride, potassium, sodium and hydroxide in a 3 cm thick specimen in a migration test where 12 V potential difference was applied. The predicted data is shown when chloride penetrated approximately 1 cm, Truc (2000).

6.5.8 Openness/Practicality Physical models are rarely commercially available or user-friendly at all. In most cases the only possible user is the developer himself! Additionally, the models must treat so many important details in a correct way that it is very difficult for anyone else to evaluate the accuracy and relevance of a physical model. A number of models, however, are very well documented and all details are thoroughly published. The large amount of required input data is not easily available but have to be quantified by laboratory measurements and complicated calculations for every new case. This makes the general practical application of physical models very limited.

6.5.9 Interpretation of real behaviour All models “produce” nice results and frequently predictions are fitted to measured data by adjusting a number of all the input parameters. However, good data that can be used for comparisons is scarce. Simultaneously measured moisture and ion profiles must be made available, where micro climatic data at the concrete surface and all relevant material properties are well documented.

6.5.10 Advantages Physical models of this kind are the only ones capable of considering the effects of all ions in a correct way. This makes them especially suitable for explaining observations in laboratory

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 62 of 112 WP4 REPORT – Modelling of Chloride Ingress and field tests that cannot be foreseen by simple empirical and physical models. In this way the confidence in simple models might be significantly improved.

6.5.10 Limitations & Drawbacks The tremendous amount of required input data, material properties and environmental conditions, most of it as functions of the other parameters, makes the practical application very doubtful. Relevant test methods for the diffusion coefficients must be verified. The interaction/binding processes are still not fully understood and possible to quantify.

6.5.12 References Li & Page (1998), Samson et al (1999), Truc (2000), Johannesson (2000), Marchand et al (2002), Khitab et al (2004).


6.6 Conclusions, models based on flux equations The main conclusions on the models based on flux equations are

• Physical models, especially those considering all ions and convection at the same time, give theoretical predictions as close to our present knowledge as possible.

• Physical models may be used to check the relevance of more simple, empirical models and may be explain some the assumptions that have to made in simple models. That may improve the confidence in empirical models.

• Physical models need quantification of a number of material parameters and a number of environmental parameters, each as a function of several parameters,

• The required surface climatic conditions are extremely complicated to model.

• Because of lack of availability, limited user-friendliness and the huge amount of required input data most physical models will remain research tools, not meant for practical applications.


7 SHORT TERM SENSITIVITY ANALYSIS 7.1 Introduction This chapter presents the methodology used to perform the short term (10 year exposure) sensitivity analysis on models for chloride penetration prediction. We need to:

- define a practical situation - define the model - define the input data and their uncertainty (for the considered practical situation) - define a limit state - do probabilities (run the model with several set of input data)

We obtain the uncertainty of the final result and the importance factors of the input data. The models which have been tested are:

- the ERFC model with constant diffusion coefficient and constant surface concentration,

- the LEO model, - the Ms-Diff model.

For more information on these models, see chapter 5 and 6. The work in WP4.2 was mainly performed by X.-S. Nguyen, M. Carcasses, I. Petre-Lazar, G. Martin at INSA and EDF, respectively.

7.2 Methodology for sensitivity analysis Usually, the input data for a model are random variables. The environmental parameters are not very well defined and even if the concrete formulation is well known, its properties vary in all the construction. Though a deterministic approach is not enough and it’s important to know the influence of input data variation on output data for each model. So a probabilistic approach can help to evaluate these variations. 7.2.1 Probabilistic method principle Generally, we search the probability of a function G(X) (where X is the random variable vector of input data for a model) to be greater than 0. G(X) = 0 is the limit state function and separate input data space in two sets: the failure set (G(X) < 0) and the safe set (G(X) > 0). To resolve this kind of problem, different numerical methods exist as simulation methods (for

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 65 of 112 WP4 REPORT – Modelling of Chloride Ingress example the Monte-Carlo one) or others like FORM and SORM (respectively First et Second Order Reliability Method). These methods are available in the probabilistic software PROBAN, Olesen (1992), developed by Det Norske Veritas. With this software, it is possible to determine the failure probability (P(G(X)<0), the most probable failure point and the importance factors, Madsen et al (1986). For each input data, its importance factor measure the relative influence of the input data variability on the failure probability. This means that the higher the importance factor for an input data is, the higher the influence of the input data variability will be on the position in safe or unsafe set. Refer to probability law of a random variable: Mathematically, a random variable is defined as a value which depends on the result of a random test e.g. not previewed in advance with good accuracy. The probability law of a random variable allows calculating these values. A probability law FX can be characterised by a repartition function R: [ ]1,0: →RF X )()( xXPxFx X ≤=→ [7.1] where X is a random variable, x its realisation and FX is an increasing function, continuous on the right as 0)(lim =

−∞→xFXx

and 1)(lim =+∞→

xFXx. So it’s possible to calculate the probability for

X to belong to any interval of R : )()()(,,),( 2 aFbFbXaPbaRba XX −=≤<<∈∀ The random variables are classified as a function of their value validity set. Therefore, a random variable X belongs to a Gaussian distribution law (normal law) if its

values are in R and 2

2

2)(

21)( σ

πσ

mx

X exF−−

= .

The beta distribution law is expressed as follow:

),()(

)()()( 1

11

rtrBabxbaxxF t

rtr

X −−−−

= −

−−−

[7.2]

with

∫ −− −=1

0

11 )1(),( dtttsrB sr [7.3]

This last choice allows to correctly model the data defined by “an inferior limit, a most probable value and a superior limit”. The a and b parameters correspond to the limits and xmax is the most probable value in the set [a,b]. Hence, the mean value m and the standard deviation σ of the variable are :

6

;6

4 max abbxam −

=++

= σ [7.4]

7.2.2 Definition of a practical situation A practical situation has to be defined by one concrete in one particular environment and for a defined service life. We have chosen the case of free diffusion of Cl- in a saturated concrete. The input data for a particular model are based on the experimental data corresponding to this situation obtained for the defined concrete.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 66 of 112 WP4 REPORT – Modelling of Chloride Ingress The environment is pure water with NaCl [Cl-] = 18 g/l in the upstream compartment and 0 in the downstream one. The simulations are made for the concrete defined in the table 1. The composition of the initial solution in concrete porosity is 23 mol/m3 of Na+, 156 mol/m3 of K+, 0 mol/m3 of Cl- et 179 mol/m3 of OH-.

Table 7.1 : Characterisation of the concrete used for simulation

No. Description Unit Quantity 1 Cement Kg/m3 560 Sand Kg/m3 695 Gravel Kg/m3 825 2 Water Litre/m3 224 3 Paste Kg/m3 784 4 Concrete Water porosity % 16 5 Solid density Kg/m3 2710

5 Concrete hydration coefficient

79,0

36,0)32,0/(/

=+−

=cepceh

Powers’ formula 6 Concrete dry density Kg/m3 2304

The concrete effective diffusion coefficient has been measured at 28 day old with the LMDC test:

Sample 1: De1 = 1.71·10-12 m2/s Sample 2: De2 = 1.72·10-12 m2/s Sample 3: De3 = 1.76·10-12 m2/s

The De coefficient is a random variable with a normal distribution law. The mean value is:

smD moyene /1073.13

10)76.172.171.1( 21212

,−

−

⋅=⋅++

=

and its variation coefficient is 1.25%. 7.2.3 Model definition A model allows predicting chloride penetration as a function of time with input data concerning concrete properties and information on the environment. We have studied the sensitivity of three models:

- the ERFC model with constant diffusion coefficient and constant surface concentration,

- the LEO model, - the MS-Diff model.

For some models, the input data presents a great variability and used a probabilistic approach Petre-Lazar (2000). Each parameter is defined by its distribution law. For the measured data,

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 67 of 112 WP4 REPORT – Modelling of Chloride Ingress we proposed a Gaussian distribution law. For the parameters which are not very well known but for which the set of variation is defined, we propose a beta distribution law. In our case, we study the influence of two particular input data which are used in all the models: the chloride diffusion coefficient and the chloride surface content. 7.2.4 State limit definition A limit state is required by the probabilistic method (FORM) in order to perform the sensitivity analysis. This limit state defines the transition between the safe set and the unsafe set. Nevertheless, the choice of the state limit has a great influence on the sensitivity analysis. So in our case we proposed a limit state concerning steel corrosion. For example, we define a critical concentration of chloride which initiate corrosion on the steel surface in concrete even though the definition of such a value is still discussed. We have chosen a cover depth of 5 cm from the concrete exposed surface and a chloride critical concentration expressed by volume of pore solution: Ccrit = 6,5 g Cl-/l. So we consider that corrosion is initiated at a depth of 5 cm when chloride concentration is greater than Ccrit = 6,5 g/l. Hence, the state limit function can be written as follows:

G = Ccrit - C(x=5cm,t) [7.5]

7.2.5 Analysis execution To do the probabilistic analysis, we use the Monte-Carlo simulation and the FORM/SORM method. The Monte-Carlo method is based on a lot of random input data sets. For each set, the response of the model is calculated and a statistical treatment of whole the results is applied. The FORM / SORM and Monte-Carlo methods allow calculating the importance factors. The FORM/SORM method principle can be represented by the Figure 7.1.

Figure 7.1 Geometrical representation of the conception point

The FORM method is used to: - Evaluate the fiability index β. The design point U* is determined by an optimisation

algorithm :

U2

U1

( ( ( )))( ) 0

gG≡ =S X U

U

U*

β

FORM

SORM

β : fiability index U* : design point X : random variable S(X) : system response g(X,S(X)) : state limit function – failure criteria


Numerical model

0))(,(min 2* ≤= USUGUArgU [7.6]

- Obtain a n approximate value of the failure probability.

)( β−Φ≈fP [7.7]

To increase the precision of equation 7.7, the SORM method has been proposed. The idea is to replace the limit state surface by a quadratic surface which probabilistic content is analytically known.

i

n

if kP

ββ

+Π−Φ≈

−

= 11)(

1

1 [7.8]

The Monte-Carlo method can be represented by the following figures:

Figure 7.2 Illustration of the Monte-Carlo method

These methods are available in the PROBAN software. Figure 7.3 illustrate the realisation of the sensitivity analysis:

Unsafe set

Nf

Set of random variables g(S(Xi))

qN

P ff ≈

10k+2 set to evaluate Pf = 10-k


INPUT DATA Random value set

SOFTWARE Probabilistic tool : PROBAN

Numerical model OUTOUT DATA

Range of output data

Importance factors

Figure 7.3 Representation of the sensitivity analysis

The output is a couple of values which characterise the range of the output data. We calculate the fractiles at 95% and 5%. They are determined by the following relations: σ645.195 += mf (superior limit) σ645.15 −= mf (inferior limit) where m is the mean value and σ the standard deviation of all the obtained output data.

7.3 Application of the methodology for ERF 7.3.1 Definition of the model This is the straight forward error function:

⎥⎥⎦

⎤

⎢⎢⎣

⎡

⎟⎟⎠

⎞⎜⎜⎝

⎛−=

tDxerfCtxC

asurf .2

1.),( [7.9]

where C(x,t) is the chloride concentration in time [g of Cl/l], Csurf is the chloride concentration in the external solution [g of Cl/l], Da is the apparent diffusion coefficient [m2/s], t is the exposure time [s] and x is the penetration depth [m]. 7.3.2 Input data Table 7.2 : Input data for ERF

No. Parameter Meaning Statistical distribution 1

Csurf Chloride surface concentration Gaussian distribution Mean value = 18 g/l Coefficient of variation = 5%

2 Da Diffusion coefficient

Gaussian distribution Mean value = 4.96·10-12 m2/s Coefficient of variation = 1.25%

3 x Penetration depth 0.05 m (protection thickness) 4 Ccrit Limit state of corrosion initiation 6.5 g/l

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 70 of 112 WP4 REPORT – Modelling of Chloride Ingress 7.3.3 Results for ERF Chloride profiles:

At t = 10 years:

02468

10121416182022

0 0,01 0,02 0,03 0,04 0,05

depth x (m)

conc

entr

atio

n (g

/l)

free chloridesmean profile

Figure 7.4 : Band of chloride profiles at 10 years for ERF

Importance factors In our case, we interest in the importance factors before 10 years. The results of importance factors are presented in the table 3 :

Table 7.3 : Importance factor in time evolution for ERF

Importance factors t (years) Csurf Da

3 69.4 30.6 4 80.0 20.0 5 87.2 12.8 6 91.8 8.2 7 94.5 5.5 8 96.1 3.9 9 97.2 2.8 10 97.8 2.2


0102030405060708090

100

0 5 10

T (years)

perc

enta

ge (%

)

CsurfDa

Figure 7.5 Importance factor in time evolution for ERF

On the Figure 7.5, we found that the importance factors are time-dependent. In considered case, the variation of the chloride surface concentration is more important than the one of the chloride coefficient.

Curve of the limit state function A limit state defines the transition between the safe set and failure set.

Figure 7.6 Curve of the limit state function for ERF

4,00E-12

4,50E-12

5,00E-12

5,50E-12

6,00E-12

6,50E-12

7,00E-12

7,50E-12

10 15 20 25 30 35

Csurf (g/l)

D (m

2/s)

None passivation zone

Passivation zone

4 years

10 years

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 72 of 112 WP4 REPORT – Modelling of Chloride Ingress 7.4 Application of the methodology for LEO 7.4.1 Definition of the model This is the function of the model:

⎥⎥⎦

⎤

⎢⎢⎣

⎡

⎟⎟⎠

⎞⎜⎜⎝

⎛−−+=

tDxerfCCCtxC

Clinitialesurfinitiale ...2

1)(),(ηα

[7.10]

where C(x,t) is the chloride concentration in time [g/l], Cinitiale is the initial chloride concentration in the pore solution [g/l], Csurf is the chloride concentration in the external solution [g/l], DCl is the effective chloride concentration in the pore solution [m2/s], x is the penetration depth [m], t is the exposure time [s], α is the correction coefficient on the interaction of ionic flux and η is the correction coefficient on the interaction of ion – matrix. 7.4.2 Input data

Table 7.4 : Input data for LEO

No. Parameter Meaning Statistical distribution

1 Cinitiale Initial chloride concentration in

the pore solution 0 (g/l)

2 Csurf Chloride surface concentration Gaussian distribution Mean value =18 (g/l solution) Coefficient of variation = 5%

3 D Effective diffusion coefficient in pore solution

Gaussian distribution Mean value = 10.81·10-12 (m2/s) Coefficient of variation = 1.25%

4 x Penetration depth 0.05 m (protection thickness) 5 Ccrit Limit state of corrosion initiation 6.5 g/l 6 cement Cement quantity/m3 of concrete 560 kg 7 water Water quantity/m3 of concrete 224 kg 8 paste Paste quantity/m3 of concrete 784 kg 9 p Concrete porosity 0.16 10 h Concrete hydratation coefficient 0.79 11 ρ Density of concrete 2304 kg/m3

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 73 of 112 WP4 REPORT – Modelling of Chloride Ingress 7.4.3 Results for LEO Chloride profiles

At t = 10 years :

02468

10121416182022

0 0,01 0,02 0,03 0,04 0,05

depth x (m)

conc

entr

atio

n (g

/l)

free chloruresmean profile

Figure 7.7 Band of chloride profiles at 10 years for LEO

Importance factors

Table 7.5 : Importance factor in time evolution for LEO

Importance factors t (years) Csurf D

4 62.7 37.3 5 72.8 27.2 6 81.5 18.5 7 87.8 12.2 8 92.0 8.0 9 94.6 5.4 10 96.3 3.7


0102030405060708090

100

0 2 4 6 8 10

T (years)

perc

enta

ge (%

)

CsurfD

Figure 7.8 Importance factor in time evolution for LEO

As the previous case, the importance factors are time-dependent and the variation of the chloride surface concentration is more important than the one of the chloride coefficient.

Curve of the limit state function

1.00E-11

1.20E-11

1.40E-11

1.60E-11

1.80E-11

2.00E-11

2.20E-11

15 20 25 30 35 40 45

Csurf (g/l)

D (m

2/s)

4 years

10 years

Figure 7.9 Curve of the limit state function for LEO


Passivation zone

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 75 of 112 WP4 REPORT – Modelling of Chloride Ingress 7.5 Application of the methodology for MS-Diff 7.5.1 Definition of the model The flux of ionic species is described by Nernst Planck equation:

⎥⎦

⎤⎢⎣

⎡∂

∂+

∂∂

+∂

∂∂∂

−

=∂

∂−+

∂∂

xtx

RTFtxcDz

xtx

txtxc

Dt

txcD

x

ttxcC

pt

txcp

iieii

i

iie

iie

iBmii

),(),(),(

),(),(),(

)),(()1(

),(

,,,

,0

ϕγγ

ρ [7.11]

where De.i is the diffusion coefficient of the ionic specie i [m2/s], ci is the concentration of ionic specie i [mol /m3], γi is the activity coefficient of the ionic specie i, zi is the valence of the ionic specie i, ϕ is the electric potential [V], F is the constant of Faraday [96480 J/(V x mol)], R is the constant of perfect gas [8.314 J/(mol x K)], T is the absolute temperature [K], po is the porosity of concrete, ρ is the density of dry concrete [kg /m3] and Cmi.B is the number of bound specie divided by the dry material [mol /kg]. In the MsDiff model, we consider four ionic species Na+. K+. OH- et Cl-. 7.5.2 Input data

Table 7.6 : Input data for MsDiff

No. Descrition Unit Value, incertainty

1 Sample thickness m 0.2 2 Exposed sample surface cm2 95 3 Porosity % 16 4 Dry volume mass kg/m3 2710 5 Freundlich isotherm coefficients α 0.0064 β 0.33 6 Immersion duration days 7 Time step sec 200.000 8 Space step nodes 100

9 Chloride diffusion coefficient (LMDC test)

m2/s

Gaussian distribution

Mean value = 1.73·10-12 m2/s Coefficient of variation = 1.25


10 Ratio Dion/DCl- Na+ 0.05 K+ 0.0735 OH- 2.73

11 Specie concentration in the upstream compartment Na+ mol/m3 Cl-

K+ mol/m3 0

Cl- mol/m3Gaussian distribution

Mean value = 18 g/l = 507 mol/m3

Coefficient of variation 5% OH- mol/m3 0

12 Specie concentration in the interstitial solution and the downstream compartment

Na+ mol/m3 23 K+ mol/m3 156 Cl- mol/m3 1 OH- mol/m3 178

13 External current mA 0 7.5.3 Results for MsDiff Chloride profiles At t = 10 years:

02468

10121416182022

0 0,01 0,02 0,03 0,04 0,05

depth (m)

conc

entr

atio

n (g

/l)

free chloride

Figure 7.10 Band of chloride profiles at 10 years for MsDiff

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 77 of 112 WP4 REPORT – Modelling of Chloride Ingress Importance factors

Table 7.7 : Importance factor in time evolution for MsDiff

Importance factors T (years) Csurf D

4 62.7 37.3 5 72.8 27.2 6 81.5 18.5 7 87.8 12.2 8 92.0 8.0 9 94.6 5.4 10 96.3 3.7

0102030405060708090

100

0 5 10

T (years)

perc

enta

ge (%

)

CsurfD

Figure 7.11 Importance factors in time evolution for MsDiff

As the two previous cases, the importance factors are time-dependent and the variation of the chloride surface concentration is more important than the one of the chloride coefficient.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 78 of 112 WP4 REPORT – Modelling of Chloride Ingress Curve of the limit state function

Figure 7.12 Curve of the limit state function for MsDiff

7.6 Conclusions The sensitivity analysis with probabilistic methods has been done on two input data: the surface chloride concentration and the chloride diffusion coefficient. This study shows that the importance factors are depending on time. This remark is valid for the three tested models in spite of their very different conception. In the considered case, the influence of chloride surface concentration is more important than the one of diffusion coefficient for the three tested models. So that, we spend more efforts to increase the precision of the chloride surface concentration measure in order to increase the precision of the prediction. The chloride surface concentration is function of the environment but also of the interactions between cement paste and chloride. So the knowledge of the binding isotherm is a key of the prediction of chloride ingress as important as the diffusion coefficient. The above conclusions are based on the sensitivity analysis done on an exposure time up to ten years. The conclusions are depending on the exposure time. For longer exposure times, such as 100 years, the conclusions may be different. This is tested in the next chapter.

1,00E-12

1,20E-12

1,40E-12

1,60E-12

1,80E-12

2,00E-12

2,20E-12

15 20 25 30 35 40 45

Csurf (g/l)

D (m

2/s)

10 years


Passivation zone

4 years


8 LONG-TERM SENSITIVITY OF ERFC-MODELS

A second sensitivity analysis has been done in sucha way that it covers also much longer exposure times than 0-10 years. This analysis has been done by Tang Luping at SP. 8.1 Mathematical expressions Since the initial chloride concentration Ci in an ERFC model only change the vertical position of a chloride distribution profile, but not influence the shape of the profile, it is reasonable to assume that Ci ≈ 0, so as to simplify the mathematical expressions. Thus all the models based on ERFC can be expressed in a general form as follows: ( )zCC erfcs ⋅= [8:1] where C could be further expressed as (C - Ci) and Cs as (Cs - Ci) in the application of the models. In equation [8:1],

ttt

kD

xzn

⋅⎟⎠⎞

⎜⎝⎛ ′

=′

002

[8:2]

or more correctly,

tt

tt

ttt

nkD

xznnn

⋅⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛ ′

−⎟⎠⎞

⎜⎝⎛ ′

+⎟⎠⎞

⎜⎝⎛ ′

−

=−− 1

ex1

ex00 11

2

[8:3]

For model A1c, it is the combination of equations [8:1] and [8:2], while for models A1b and A1d, let k = 1 in equation [8:2], and for model A1a, let k = 1 and n = n’ = 0 in either equation [8:2] or [8:3]. Even though in model A1d the surface concentration Cs is a function of time t, it will be seen later that this will only influence the magnitude of Cs, but not the sensitivity of Cs for prediction of C.


8.2 Sensitivity of various parameters for prediction of chloride concentration

8.2.1 Surface concentration Cs The sensitivity coefficient of parameter Cs is as follows:

( )zCC erfc

s

=∂∂ [8:4]

Therefore,

1s

s

=⋅ΔΔ

CC

CC [8:5]

This implies that any relative change in surface concentration Cs will result in an equal relative change in concentration C. From equation [8:5] it can be seen that the value of

CC

CC s

s

⋅ΔΔ has no dependence on any other parameter, such as t. Therefore, no matter if Cs is

time-dependent or not, its sensitivity is the same. 8.2.2 Diffusion coefficient D0 The sensitivity coefficient of parameter D0 can be derived as:

2

0

0

0

zeDzC

DC −⋅⋅=

∂∂

π [8:6]

resulting in the following equation:

s

0

0

22 11

CCezez

CC

CD

DC z

zs−

− ⋅⋅

π=⋅⋅⋅

π=⋅

ΔΔ [8:7]

Apparently, the sensitivity of parameter D0 is related to the ratio of C to Cs (or C/Cs) and the value of z that contains all the variations except for Cs. It can be seen from equation [8:7] that

CD

DC 0

0

⋅ΔΔ is inversely proportional to C/Cs. The quantitative relationships of

CD

DC 0

0

⋅ΔΔ and z

are shown in Figure 8.1. The maximal influence of z on CD

DC 0

0

⋅ΔΔ appears at z ≈ 0.7. It is

relatively less sensitive if the C/Cs is larger than 0.1 and z value is larger than 1, the latter implies a thicker cover x or longer exposure time t.


0

5

10

15

20

25

0 0.5 1 1.5 2 2.5 3

z

C/Cs = 0.01

C/Cs = 0.02

C/Cs = 0.03

C/Cs = 0.05

C/Cs = 0.1

C/Cs = 0.2

C/Cs = 0.5CD

DC 0

0

⋅ΔΔ

Figure 8.1. Sensitivity of parameter D0 for the prediction of chloride concentration C.

8.2.3 Parameter k The sensitivity coefficient of parameter k is similar to that of parameter D0, that is,

20 ze

kzC

kC −⋅⋅=

∂∂

π [8:8]

resulting in an equation equal to equation [8:7]:

s

z

CCez

CD

DC

Ck

kC

2

10

0

−⋅⋅=⋅

ΔΔ

=⋅ΔΔ

π [8:7’]

Therefore, the relations in Figure 8.1 are also valid for parameter k. 8.2.4 Parameter n’ in equation [8:2] The sensitivity coefficient of parameter n’ in equation [8:2] is

⎟⎠⎞

⎜⎝⎛ ′

⋅⋅π

=′∂

∂ −

tt

ezC

nC z 00 ln

2

[8:9]

which results in

⎟⎠⎞

⎜⎝⎛ ′⋅

⋅π′

=′

⋅′Δ

Δ −

tt

CCezn

Cn

nC z

0

s

ln2

[8:10]

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 82 of 112 WP4 REPORT – Modelling of Chloride Ingress Obviously, besides the influences of C/Cs and z values, which is similar to those discussed in

section 2.2, Cn

nC ′

⋅′Δ

Δ is directly proportional to the value of n’ and logarithmically

proportional to the ratio of t0/t. In the case of C/Cs = 0.1 and z = 1, the relationships of

Cn

nC ′

⋅′Δ

Δ and n’ can be shown in Figure 8.2.

-15

-10

-5

0

0 0.2 0.4 0.6 0.8 1

n' in equation [8:2]

0.001

0.002

0.005

0.01

0.02

0.05

0.1

t' 0/t =

Figure 8.2. Sensitivity of parameter n in equation [8:2] for the prediction of chloride concentration C: Influence of value n when C/Cs = 0.1 and z = 1.

8.2.5 Parameter n in equation [8:3] The sensitivity coefficient of parameter n in equation [8:3] is

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

⎟⎠⎞

⎜⎝⎛ ′

−⎟⎠⎞

⎜⎝⎛ ′

+

⎟⎠⎞

⎜⎝⎛ ′

⋅⎟⎠⎞

⎜⎝⎛ ′

−⎟⎠⎞

⎜⎝⎛ ′

+⋅⎟⎠⎞

⎜⎝⎛ ′

+−

−+⎟

⎠⎞

⎜⎝⎛ ′

⋅⋅π

=∂∂

−−

−−

−nn

nn

z

tt

tt

tt

tt

tt

tt

ntt

ezC

nC

1ex

1ex

1exex

1exex

00

1

ln11ln

11ln

2

[8:11]

which results in

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

⎟⎠⎞

⎜⎝⎛ ′

−⎟⎠⎞

⎜⎝⎛ ′

+

⎟⎠⎞

⎜⎝⎛ ′

⋅⎟⎠⎞

⎜⎝⎛ ′

−⎟⎠⎞

⎜⎝⎛ ′

+⋅⎟⎠⎞

⎜⎝⎛ ′

+−

−+⎟

⎠⎞

⎜⎝⎛ ′⋅

⋅π

=⋅ΔΔ

−−

−−

−

nn

nn

z

tt

tt

tt

tt

tt

tt

ntt

CCezn

Cn

nC

1ex

1ex

1exex

1exex

0

s1

ln11ln

11ln

2

[8:12]


In this case the influences of n and t’0/t on Cn

nC

⋅ΔΔ become complicated, and can be

illustrated in Figure 8.3, where C/Cs = 0.1 and z = 1.

-8

-6

-4

-2

0

2

4

0 0.2 0.4 0.6 0.8 1

n in equation [8:3]

0.001

0.002

0.005

0.01

0.02

0.05

0.1

Cn

nC

⋅ΔΔ

t' ex/t = 0.001

t' 0/t =

Figure 8.3. Sensitivity of parameter n in equation [8:3] for the prediction of chloride concentration C: Influence of value n when C/Cs = 0.1 and z = 1.

8.2.6 Parameter t0 The sensitivity coefficient of parameter t0 in both equations [8:2] and [8:3] will be similar, that is,

0

0

0

2

teznC

tC z

′⋅

⋅π⋅

=′∂

∂ −

[8:13]

resulting in

0

0

0

2

CCezn

Ct

tC z−⋅

⋅π

=′

⋅′Δ

Δ [8:14]

Clearly, besides the influences of C/Cs and z values, which is similar to those discussed in

section 2.2, Ct

tC 0

0

′⋅

′ΔΔ is directly proportional to the value of n. In the case of C/Cs = 0.1 and z

= 1, nnCt

tC 18.3100

0

=⋅π

=′

⋅′Δ

Δ .


8.3 Effect of different parameters on the sensitivity In order to compare the effect of different parameters on the sensitivity, we have to make the parameter z constant, because this parameter is directly related to chloride concentration C. For instance, in the case of x = 0.05 m and t = 100 years, it is not difficult to keep z = 1 through the adjustment of parameters D0, n’ or n. Under the conditions of z = 1 and t’0/t = 0.005 (for parameter n only, corresponding to t’0 = 0.5 years and t = 100 years), the effects of different parameters on the sensitivity can be summarise in Figure 8.4. It can be seen that, among different parameters, n or n’ is the most sensitive one, especially when its value is larger than 0.2. Except for parameter Cs, which is independent of other parameters, the sensitivity of all the other parameters is dependent on the ratio C/Cs.

-10

-8

-6

-4

-2

0

2

4

0.1 0.2 0.3 0.4 0.5

C /C s

Cs

Do or k

n = 0.2

n' = 0.2

n = 0.5

n' = 0.5

n = 0.8

n' = 0.8

t0 (n' = 0.2)

t0 (n' = 0.5)

t0 (n' = 0.8)

Cf

fC i

i

⋅ΔΔ

n' as in Eq [8:2]n as in Eq [8:3]

Figure 8.4. Effects of different parameters on the sensitivity under the conditions of z = 1 and t’0/t = 0.005 (for parameter n or n’ only).

8.4 Discussions Under the typical exposure conditions, Cs may be about 5 % of binder. If the threshold chloride content is 0.5 % of binder (e.g. in splash zone), C/Cs will be about 0.1. In this case, the sensitivity of parameters D0 and any types of k will be 2 times as large as that of Cs, while the sensitivity of n, once it is larger than 0.2, will be more than 2 times as large as that of Cs. If the threshold chloride content is 1.5 % of binder (e.g. in submerged zone), C/Cs will be about 0.3. Thus the sensitivity of parameters D0 and any types of k will be less than that of Cs, but the sensitivity of n may still be larger than that of Cs.


8.5 Combined uncertainty of models for prediction of

chloride concentration According to ISO standard expression of uncertainty, 1993 (E), the combined standard uncertainty uc is the positive square root of the combined variance 2

cu , which is given by

∑=

⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

=N

if

ii

ufCu

1

22

2c [8:15]

where f denotes any type of parameter which may contribute to the uncertainty. The coefficient of variation COV is defined as

CucCOV = [8:16]

8.5.1 Model A1a In model A1a, two parameters, Cs and D0, will contribute to the uncertainty, that is,

22

0

22

s

2A1a 0s DC u

DCu

CCu ⋅⎟⎟

⎠

⎞⎜⎜⎝

⎛∂∂

+⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

= [8:17]

2

0

2

s

2

sA1a

0

2

s 1COV ⎟⎟⎠

⎞⎜⎜⎝

⎛⋅

⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜

⎝

⎛⋅

π+⎟⎟

⎠

⎞⎜⎜⎝

⎛=

−

Du

CCez

Cu D

zC [8:18]

8.5.2 Models A1b and A1d In models A1b and A1d, four parameters, Cs, D0, n and t0, will contribute to the uncertainty, that is,

22

0

22

22

0

22

s

2dA1b, 00s tnDC u

tCu

nCu

DCu

CCu ⋅⎟⎟

⎠

⎞⎜⎜⎝

⎛∂∂

+⋅⎟⎠⎞

⎜⎝⎛

∂∂

+⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

+⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

= [8:19]

⎪⎭

⎪⎬⎫

⎪⎩

⎪⎨⎧

⎟⎟⎠

⎞⎜⎜⎝

⎛+⎥

⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛+⎟⎟

⎠

⎞⎜⎜⎝

⎛⋅

⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜

⎝

⎛⋅

π+⎟⎟

⎠

⎞⎜⎜⎝

⎛=

− 2

0

22

0

2

0

2

s

2

sdA1b,

00

2

s ln1COVtu

nutt

Du

CCez

Cu t

nD

zC [8:20]

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 86 of 112 WP4 REPORT – Modelling of Chloride Ingress 8.5.3 Model A1c In model A1c, seven parameters, Cs, D0, kt, ke, kc, n and t’0, will contribute to the uncertainty. The expressions are similar to those for models A1b and A1d, that is,

22

0

22

22

22

0

22

s

2A1c 00s tnk

iDC u

tCu

nCu

kCu

DCu

CCu

i⋅⎟⎟

⎠

⎞⎜⎜⎝

⎛∂∂

+⋅⎟⎠⎞

⎜⎝⎛

∂∂

++⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

+⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

+⋅⎟⎟⎠

⎞⎜⎜⎝

⎛∂∂

= ∑ [8:21]

⎪⎭

⎪⎬⎫

⎪⎩

⎪⎨⎧

⎟⎟⎠

⎞⎜⎜⎝

⎛+⎥

⎦

⎤⎢⎣

⎡⎟⎠⎞

⎜⎝⎛+⎟⎟

⎠

⎞⎜⎜⎝

⎛+⎟⎟

⎠

⎞⎜⎜⎝

⎛⋅

⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜

⎝

⎛⋅

π+⎥

⎦

⎤⎢⎣

⎡= ∑

− 2

0

22

0

22

0

2

s

2

sA1c

00

2

s ln1COVtu

nutt

ku

Du

CCez

Cu t

ni

kDz

C i [8:22]

where ki represents kt, ke and kc, respectively. Apparently, the more parameters in a model, the more sources of error. However, the actual COV for each model is dependent on the actual magnitude of each individual uncertainty. Besides, the most important factor for a prediction model is the degree of agreement with the actual chloride profiles. 8.6 Conclusions Parameter Cs. has constant sensitivity in concentration prediction, that is,

CC

CC s

s

⋅ΔΔ = 1.

The sensitivity of parameter D0 or k (as a multiplier to D0) is related to the ratio C/Cs and the value of z that contains all the variations except for Cs. If z = 1, when C/Cs is larger than 0.2, the sensitivity of D0 or k in concentration prediction becomes less than parameter Cs, Otherwise it becomes larger than Cs. On the other hand, when the value of z is very small or very large (not close to 0.7), the prediction of chloride concentration also becomes less sensitive to parameter D0 or k. Among all the parameters, the age factor n or n’ is the most sensitive one, especially when its value is larger than 0.2. Parameter t0 has relatively less sensitivity in the prediction of chloride concentration, especially if the age factor n or n’ is less than 0.5.


9 BENCHMARKING OF MODELS

9.1 Objectives and over-view of work performed Task 4.3 Bench Marking, by partners IETcc (task leader), SP, LTH and EdF, concerned studying the models’ abilities to reproduce reality, with laboratory test results as input and in-field performance data as comparison. This was done by selecting some field data, which had the concrete composition and properties and the environmental conditions well documented, and ask model developers to make predictions with their models on these field cases. The prediction results were then to be analyzed by using a probabilistic approach. The complete background data and prediction results of task 4.3 are presented in a final working report by the task leader IETcc: “BENCHMARKING OF MODELS”, dated October 25, 2005. This report has been complimented with an “AMENDMENT”, dated January 2006, by the work package leader LTH, adding and correcting some data and analysis. The work performed in task 4.3 was the following:

1. Establishment of criteria for the benchmarking

2. Selection of profiles from the data base

3. Selection of the model producers more reputed in the world

4. Submission of the documentation to the modellers

5. Collection of results and application of benchmarking criteria

6. Discussion on results obtained and suggestions for further actions

9.2 Establishment of criteria for benchmarking Prior to the work, there were established criteria for making the comparison between predicted profiles and real ones, in collaboration between IETcc and EDF. The method proposed for making the benchmarking was selected on the basis of the comparison of the bias of the predicted profiles of each model with respect to the real one, based in the areas between both profiles as is explained in the next section. It has the

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 88 of 112 WP4 REPORT – Modelling of Chloride Ingress advantage to quantify in some manner the deviation and therefore to make the models comparable. 9.2.1 Comparing different models when the measured data is known To verify the performance of the different mathematical approaches made by different researchers, a methodology based in the area between the curves was developed. The prediction result from each model can be compared with the measured curve by comparing the value of the area between the predicted curve and the measured curve in a depth range from X1 to X2 as shown in Figure 9.1. The values used to compare are the Areas with and without sign called:

S1 = A1+A2 S2 = A1 + A2

MeasuredPredicted

c1

c2

x1 x2

A1

A2

X

c MeasuredPredicted

c1

c2

x1 x2

A1

A2

X

c

Figure 9.1 Predicted and measured profiles of chloride concentration

where : c1 : measured curve (total chlorides) c2 : model curve (total chlorides) A1 : area between the measured curve and model curve which measured data is lower (negative sign) [%·mm] A2 : area between the measured curve and model curve which measured data is higher (positive sign) [%·mm] x1, x2 : validation depth range C, X : concentration and depth axis respectively

Further information:

The c1 and c2 curves are cubic spline interpolation of measured and model data points. If the measured data or predicted profile do not cover the whole interval between x1 and x2 an extrapolation was made.


S1 gives the information about how near the model is from the measured data. S2 gives the information about how much higher or lower the predicted curve is, compared with the measured data. S1 > 0 S2 < S1 always S2 < 0 if the measured profile is lower than the predicted S2 = -S1 if the whole measured profile is lower than the predicted S2 = S1 if the whole measured profile is higher than the predicted

The validation range adopted was:

• x1 = 10mm ( this value was adopted to not take into account the skin effect) • x2 = 50mm (except in the situation when the depth of the measured data is smaller).

It is obvious from some cases that this choice of validation range is questionable where the depth of penetration is small. A significant part of the profile falls then between 0 and x1 and is excluded from the comparison. 9.2.2 Comparing different models when the measured data is unknown When the measured data is unknown, for example for predicted chloride profiles at 100 years, the table with the modellers’ data and the graphics are given, but no comparison is made.

9.3 Selection of profiles and documentation prepared for the benchmarking

The purpose is to know the deviation of the predicted data from the real data, represented by a second profile, of the same concrete at a second age. A prediction to 100 years was introduced in order to evidence the sensitivity of the models to take into account the age effect on diffusivity. As the task tries to study the ability of the different available models to predict further evolution of chloride profiles, the prediction can be made based on results (a profile) from one age, accompanied or not with short term results in the laboratory, of the same concrete. The tasks performed were the following: 1) Selection of a set of chloride profiles from the database of ChlorTest following the criteria

given by the leaders of the project (SP) and of WP 4 (LTH). The selection should have been made by WP 3. • A set of 400 profiles was sent to the leader of task 4.3 at the beginning of January 2005. • It was recognized that most of the profiles did not follow the criteria fixed previously in

the meeting of November 2004. • Only profiles provided by SP seem to meet all the criteria • It seemed to leaders of task 4.3 that a wider variety of profiles is necessary to assure

enough objectivity to the benchmarking exercise.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 90 of 112 WP4 REPORT – Modelling of Chloride Ingress Due to the enormous amount of profiles sent to task 4.3 (400) extra work has been to select

some of them to fulfil the criteria given on November 2004. Therefore, IETcc prepared the following set of profiles:

1. Those of SP because they meet the criteria fixed. Two profiles, each at a different age, will be submitted as proposed on November 2004. The prediction exercise will be from the first age profile to the second and to 100 years.

2. A second set of profiles was selected not meeting all the criteria. Only one profile at one age will be sent to the modellers requesting prediction to the second age (profile not sent) and 100 years.

3. Data from old real structures will constitute a third set of profiles to be sent to modellers to predict the 100 year profiles.

2) A brief guide for modellers was prepared together with the profiles. The guide contained:

• Summary of the project • Purpose of benchmarking exercise (task 4.3) • The three set of profiles selected with indications of needed predictions and form for

submitting results • Form for responding on positive or not collaboration with task 4.3 • A questionnaire to specify by the modeller the characteristics of the models • Excel files with the given profiles with space for the predicted ones.

In Annex A of the final task 4.3 working report the profiles submitted for the benchmarking are enclosed. Three series of data were selected with indications of needed predictions. The results of the calculated profiles at the ages requested from each modeller were asked to be given back in the same Excel files. The three series correspond to 22 different cases. The conditions of the profiles were: Series 1: One early of two chloride profiles from two consecutive ages in the same concrete is

given. Results of short term accelerated tests are also included. The modeller is requested: • to predict the second age profile • to predict the profile at 100 years

Series 2: One of two chloride profiles from two consecutive ages in the same concrete is

given. No result of short-term accelerated test is available. The modeller is requested:

• to predict the second age profile • to predict the profile at 100 years

Series 3: One profile from the old structure is given. The modeller is requested:

• to predict the profile at 100 years The modellers were encouraged to model all the cases if possible. The modellers can also model a limited number of cases if their time is limited. In this circumstance, it was suggested to the modellers to follow the order of the case number.


• For series 1, the exercise must be performed with the 4 first profiles supplied. The rest of profiles included in series 1 are voluntary

• In series 2 and 3 the exercise is requested in the case of the first profile. The rest of profiles supplied are voluntary.

In Series 1 the concrete chloride ingress properties were given as a measured chloride migration coefficient. That kind of information was not available in Series 2 and 3. In series 1 the given profiles were after a very short time of exposure, less than 10 % of the exposure time for the required prediction. In Series 2 and 3, however, the profiles were given after some 40 % of the final exposure time. The different cases can be summarized in the following way: Series 1Case Environment Given profile Measured/Predicted Comment

C1 Submerged limitedC2 Splash thicknessC3 AtmosphericC4 CEMI/0.4 Road VF 0.4 4.6 small diff. 0.4-4.6 yC5 5 %Si 10.2C6 20 %FA 10.1C7 CEMI/0.4 Road H 0.4 4.6 small diff. 0.4-4.6 yC8 SubmergedC9 Splash

C10 AtmosphericC11 TidalC12 Road VFC13 Road H

VF= vertical surfaceH = horisontal surface

small diff. 0.4-4.6 y

1.0Submerged

ConcreteExposure time [years]

0.4 4.6

10.2

10.3

0.5 3.0

0.8

CEM I w/c 0.5CEM I w/b 0.4

5 %Si

CEM I

w/c 0.4

CEM I w/b 0.35

CEM I w/b 0.35

5 %Si 10 %FA

0.7

Series 2Case Environment Given profile Measured/Predicted Comment

C1 OPC 7 12C2 SRPC 6 14 small diff. 6-14 yC3 CEM II/B w/c 0.57 Splash 8 18 small diff. 8-18 yC4 SubmergedC5 Tidal

CEM I w/c 0.5

CEMIV/B w/c 0.4

Tidal

1.5none

ConcreteExposure time [years]

Series 3Case Concrete Given profile Measured/Predicted Comment

C1 Tidal 1C2 9C3C4

CEMIII/B

w/c 0.45

Atmos-

pheric

42 none/100

Environ. Over sea [m]

no real difference14


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 92 of 112 WP4 REPORT – Modelling of Chloride Ingress 9.4 Selection of models The selection was made attending two criteria:

1. Models developed by reputed individuals which have published in the scientific literature.

2. Models that are known and currently used, like ”error function” or ”square root”

models. The list of model developers addressed is the following:

• Ilie Petre-Lazars – EDF- France • Tetsuya ISHIDA - University of Tokyo - Japan • Tang Luping - SP-Sweden • Lars-Olof Nilsson - University of Lund - Sweden • Myriam Carcasses- INSA, Toulouse- France • Sander Meijers- Intron - The Netherlands • Kazuo Yamada - Taiheiyo Cement – Japan • Jens M. Frederiksen – Birch & Krogboe - Denmark • Jacques Marchand - Laval University-Canada • Evan Bentz - University of Toronto - Canada • Dale Bentz- NIST- USA • Steinar Helland - Skanska- Norway • Heidi Ungricht- Switzerland • Eugen Brühwiler- University of Lausanne - Switzerland • Joost Gulikers- the Netherlands • Nick Buenfeld- Imperial College- UK • Oladis de Rincon - University of Zulia- Venezuela • Carmen Andrade- IETCC - Spain

9.5 Responses obtained Not all the modellers could participate, for different reasons. The available time was very short and just before the summer vacation 2005. Some models require two profiles as input data and that was not available. The final number of models tested has been 16, which is thought representative enough for the purpose of the project. In total there are 16 models from 9 different modellers. The “model 4” is the traditional Error Function. Table 9.1 shows the profiles analyzed in this work with the series that have been made. “1” represents that the profile in the second age was made by the modeller “2” represents that the profile for 100 years was made by the modeller

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 93 of 112 WP4 REPORT – Modelling of Chloride Ingress Table 9.1 – Models received from the modellers

Series 1 Series 2 Series3 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 C2 C3 C4 C5 C1 C2 C3 C4

Mod1 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 Mod3 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 Mod4 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod5 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod6 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod7 1,2 1,2 1,2 1,2 1,2 1,2 1,2 Mod8 1,2 1,2 1,2 1,2 1,2 1,2 1,2 Mod9 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1

Mod10 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod11 1,2 1,2 1,2 1,2 Mod12 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1,2 1,2 1,2 1,2 1 1 1 1Mod13 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod14 1 1 1 1Mod15 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1Mod16 1,2 1,2 2 1,2 1,2 1,2 1,2 2 1 2

Table 9.2 shows the summary of the models used in this task. The complete answered questionnaires are found in the Annex B of the final working report. Table 9.2 – Summary of the questionnaires answered by the modellers.

1 2 3 5 6 7 8 9 10 11 12 13 14 15 16Based on Theoretical Principles [TP]

or on Empirical Calibration[EC] EC

BO

TH

TP TP EC TP TP EC

BO

TH

TP EC

EC

EC

BO

TH

TP

Based on Fick´s second law

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

NO

Temporal dependence of Cs

YE

S

NO

NO

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

NO

Yes

Yes NO

YE

S

Need to introduce a D

YE

S

YE

S

NO

YE

S

NO

NO

NO

NO

NO

YE

S

YE

S

YE

S

YE

S

YE

S

YE

S

Type of D : steady-state [SS] or snon-teady-state [NSS] N

SS

SS

SS

BO

TH

SS

NS

S

NS

S

NS

S

NS

S

Chloride binding

YE

S

YE

S

YE

S

YE

S

NO

YE

S

YE

S

NO

NO

NO

NO

YE

S

YE

S

Temporal dependence of D

YE

S

YE

S

NO

YE

S

YE

S

YE

S

NO

YE

S

YE

S

NO

YE

S

YE

S

YE

S

YE

S

YE

S

Assume some value when it's not available N

O

YE

S

NO

YE

S

YE

S

YE

S

YE

S

NO

NO

YE

S

YE

S

YE

S

YE

S

NO

MODEL

It is noteworthy that 11 of the 16 models are using a surface chloride concentration that changes with time.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 94 of 112 WP4 REPORT – Modelling of Chloride Ingress 9.6 Comparison of results The result for the series 1 is presented with the schema below for each Case:

• The Table with the predicted and the measured profile at the second age, • The Graphic with the predicted and real profile for the Second age, • The Histogram of the S1 value of the modellers for the Second age, • The Histogram of the S2 value of the modellers for the Second age, • Comments about the results of the Second age, • The Table with the predicted and real profile for the 100 year predictions, • The Graphic with the predicted and real profile for the 100 year predictions, • Comments about the results of the 100 year predictions.

All results are documented in the final working report. Some examples are given in the next section. An analysis is made in the following section.

9.7 Some examples of results Series 1 – Case 01 – 10.3 years of exposure Series 1, Case 01, is a CEM1 concrete with a w/c of 0.40, constantly submerged into sea water. The given profile after an exposure time of 0.78 years is shown in Figure 9.2.

0

1

2

3

4

5

6

7

0 10 20 30 40 50

Depth (mm)

Cl/C

[wt%

of b

inde

r]

Series 1 Case 1

Figure 9.2.- Series 1- given profile after an exposure of 0.78 years

The given profile has a surface chloride content Cs around 2.5-3 % of binder. The predicted profiles after an exposure time of 10.3 years are shown in Figure 9.3 together with the measured profile (“Mod 0”). The differences between the predicted profiles and the measured one are shown in Figures 9.4 and 9.5, expressed as S1 and S2 values.


Chlortest WP4.3 - Series 1 - Case 1- T=10.3

0,001

1,001

2,001

3,001

4,001

5,001

6,001

7,001

0,00 10,00 20,00 30,00 40,00 50,00

Depth

Chl

orid

e C

once

ntra

tion

Mod 0 Mod 1 Mod 2 Mod 3 Mod 4 Mod 5 Mod 6 Mod 7Mod 8 Mod 9 Mod 10 Mod 12 Mod 13 Mod 15 Mod 16

Figure 9.3.- Series 1- profile 1. The profile showing the maximum (Mod 0 ) is the measured profile


0,00

10,00

20,00

30,00

40,00

50,00

60,00

70,00

80,00

Models

S1

Mod 1Mod 2Mod 3Mod 4Mod 5Mod 6Mod 7Mod 8Mod 9Mod 10Mod 12Mod 13Mod 15Mod 16

Figure 9.4- Series 1- profile 1.Value of area S1 between real and predicted profiles.



-60,00

-40,00

-20,00

0,00

20,00

40,00

60,00

80,00

100,00

Models

S2

Mod 1Mod 2Mod 3Mod 4Mod 5Mod 6Mod 7Mod 8Mod 9Mod 10Mod 12Mod 13Mod 15Mod 16

Figure 9.5- Series 1- profile 1.Area S2 between real and predicted profiles

Comments on Series 1, Case 1, 10.3 years The first remark is that the profile after 10.3 years does not represent a semi-infinite case. The profiles have been taken from specimens with a thickness of only 100 mm and penetration from two sides. For an exposure time of 10.3 years, the penetration from each surface is more than 50 mm. Consequently, the measured chloride profile is not fully relevant for comparing with the predicted profiles, since all predictions are made for semi-infinite cases. Having this in mind, the results could be analyzed with some limitations. There are large differences between the predicted profiles. Roughly, there are some models (1, 5, 9, 10 and 13) that have correctly predicted the change with time of Cs and therefore they fit the measured profile much better. Others, not having picked the new Cs, in reality used a constant Cs-value, predict profiles very far from the measured profile in the initial part of the profile near the surface. It is also interesting to notice that there are large differences between the predicted profiles within the two groups using a time-dependent or constant Cs-value, respectively. Even though the input data included a given, “early” profile and a D, the D used in the predictions are very different. However, it is interesting to realize that there are not large differences in the ”front part” of the predicted profiles of those which picked more or less well the new Cs with those considering a constant Cs. That is, at 5 cm depth the predicted profiles do not differ too much. This is of course due to the low D in all cases. As summary, in spite of the difficulty in the prediction, there are profiles which reach very good predictions in this case. These are mainly MOD 5 and MOD 9.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 97 of 112 WP4 REPORT – Modelling of Chloride Ingress Comments on Series 1, Case 1, 100 years Here the prediction is up to 100 years, see Figure 9.5. Now the prediction at 5 cm of course differs much more between the models that have considered an increase of Cs with time compared to those which did not consider this. Therefore, the Cs-change has a strong influence in the long term prediction. There is also a significant difference in the D:s used.

Chlortest WP4.3 - Series 1 - Case 1- T=100

0,00

2,00

4,00

6,00

8,00

10,00

0,00 10,00 20,00 30,00 40,00 50,00

Depth

Chl

orid

e C

once

ntra

Mod 1 Mod 2 Mod 3 Mod 4 Mod 5 Mod 6 Mod 7


Figure 9.6- Series 1- profile 1. 100 years exposure

Comments on Series 1, Case 4, 4.6 years Series 1, Case 4, is a similar concrete as in case 1 but now exposed in a road environment. The given profile is measured after 0.4 years, i.e. during the first year, see Figure 9.7. Actually, the profile was taken just after the first winter, before the rains in the summer and fall, when chloride is leached out. The predicted profiles after 4.6 years are shown in Figure 9.8 together with the measured profile. All predicted profiles are overestimating the chloride contents after 4.6 years! The measured profile after 4.6 years has actually lower chloride contents in the outer part of the concrete that the measured profile after 0.4 years!


0

1

2

3

4

5

6

7

0 10 20 30 40 50Depth (mm)

Cl/C

[wt%

of b

inde

r]Series 1 Case 4

Figure 9.7- Series 1- case 4. given profile after 0.4 years of exposure


0,001

1,001

2,001

3,001

4,001

5,001

6,001

7,001

8,001

0,00 10,00 20,00 30,00 40,00 50,00

Depth

Chl

orid

e C

once

ntra

tion


Mod 8 Mod 9 Mod 10 Mod 11 Mod 12 Mod 13

Figure 9.8- Series 1- profile 4. 4.6 years exposure

Here the measured profile presents very low values and therefore the prediction is more accurate by those models that consider a constant Cs-value equal to the given one. Those assuming an increase of Cs with time have strongly failed. The models having achieved a prediction closest to the measured profile are: MOD: 3, 8 and 10.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 99 of 112 WP4 REPORT – Modelling of Chloride Ingress 9.8 Analysis of all predictions 9.8.1 Series 1 The other cases in series 1 gave similar results. The S1-values for all the cases in series 1 are shown in Table 9.3 and the S2-values in Table 9.4. The tables also give the average S1- and S2-value, and the standard deviation for each model and each case. Table 9.3 Over-view of all predictions for series 1 compared to the measured profiles, expressed as S1-values. Series 1 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13

Mod S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 Average StDev1 21.2 50.3 9.1 12.1 5.0 16.7 12.8 30.7 5.6 7.1 17.9 6.9 3.8 15.3 13.02 34.2 63.1 11.0 8.4 25.4 21.1 8.4 24.5 19.63 75.2 69.7 11.6 13.4 10.2 21.0 15.1 21.4 7.0 21.8 25.0 10.1 12.9 24.2 22.14 35.0 65.0 9.6 15.8 6.6 24.5 14.9 20.4 8.0 6.1 16.7 16.5 20.4 20.0 15.75 15.8 12.4 1.6 6.1 4.3 6.8 7.8 6.5 1.9 5.8 9.9 2.5 1.5 6.4 4.36 66.8 46.3 49.3 63.8 9.9 8.6 59.2 18.8 27.7 5.0 7.7 53.0 59.3 36.6 24.07 52.4 25.3 76.6 89.4 7.9 10.2 43.6 34.68 36.3 65.5 13.5 30.8 6.2 23.4 29.3 20.99 13.7 72.3 4.2 13.0 7.9 18.6 12.7 18.4 3.3 6.4 18.0 8.8 8.3 15.8 17.7

10 36.2 74.4 2.8 6.5 3.0 10.5 3.3 9.5 8.7 6.0 22.0 2.5 5.5 14.7 20.311 22.5 18.9 24.1 30.9 24.1 5.012 50.5 76.4 3.1 6.8 9.4 9.2 7.4 20.6 2.3 7.5 17.7 3.9 7.8 17.1 21.913 47.3 44.4 19.3 44.6 4.7 9.0 46.3 19.6 4.6 4.6 29.3 10.4 12.8 22.8 17.31415 68.2 84.4 2.7 12.4 10.7 19.1 2.3 10.7 19.2 25.5 29.716 49.4 69.1 6.3 21.9 16.6 12.5 29.3 24.6

Average 43.0 58.5 16.5 27.1 7.3 15.5 19.8 18.6 7.3 8.1 17.8 13.9 16.3StDev 19.1 20.4 22.0 26.3 2.6 6.8 18.2 6.0 7.2 5.1 6.3 15.3 17.4

Table 9.4 Over-view of all predictions for series 1 compared to the measured profiles, expressed as S2-values. Series 1 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13

Mod S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 Average StDev1 19.5 50.3 -9.1 -12.1 -5.0 16.7 -12.8 30.7 -5.6 -7.1 16.0 6.9 3.8 7.1 18.92 34.2 63.1 -11.0 3.8 -25.1 -4.2 -8.3 7.5 30.63 75.2 69.7 -11.6 -13.4 8.1 21.0 -15.1 7.9 -6.6 -21.8 -12.0 -9.7 -12.9 6.1 31.74 35.0 65.0 -9.6 -15.8 1.2 -24.5 -14.9 -3.8 -8.0 0.7 -0.7 -16.5 -20.4 -0.9 24.85 4.3 -1.1 1.3 -6.1 -0.6 -5.8 -7.8 -3.4 1.3 -0.7 0.6 2.5 -0.6 -1.2 3.66 66.8 46.3 -49.3 -63.8 7.9 4.1 -59.2 17.5 -27.7 -5.0 -2.1 -53.0 -59.3 -13.6 42.57 52.4 25.3 -76.6 -89.4 5.8 -6.8 -14.9 56.68 36.3 65.5 -13.5 -30.8 2.1 -23.2 6.1 37.69 4.1 72.3 -3.3 -13.0 7.9 -18.6 -12.7 -15.7 -3.3 1.4 2.5 -8.8 -8.3 0.4 23.1

10 36.1 74.4 -0.9 -6.2 1.6 10.4 -2.8 2.9 -8.7 2.9 20.5 -1.4 -5.1 9.5 23.011 -22.5 -18.9 -24.1 -30.9 -24.1 5.012 50.5 76.4 -2.7 -6.8 9.4 7.9 -7.4 20.4 2.2 7.4 13.1 -3.6 -7.6 12.2 24.913 -47.3 44.4 -19.3 -44.6 4.5 8.9 -46.3 19.6 4.6 -3.6 29.3 -6.6 -9.7 -5.1 28.71415 68.2 84.4 2.7 12.4 10.7 19.1 2.3 10.7 17.7 25.4 29.716 49.4 69.1 1.9 -16.6 4.5 5.1 18.9 32.9

Average 34.6 57.5 -15.6 -27.0 4.4 -2.9 -19.8 8.0 -5.3 -1.5 8.2 -11.4 -15.1StDev 32.1 22.9 22.6 26.3 4.6 16.3 18.2 13.6 8.9 8.9 12.1 17.1 18.4

The colours in the tables give some additional information, compared to the final task 4.3 report:


data missing; now calculated and addeddata wrong, now correctedbest comparison

Some of cases (5, 9 and 10) gave excellent comparison between predicted and measured profiles, even though a few of the models predicted profiles that are far from the measured ones. Cases 1-2 gave very bad results for most of the models. The extreme increase of Cs with time, from 0.8 years to 10 years was difficult to predict. One model is outstanding in this comparison, Model 5. However, this model has been compared to most of the data of Series 1 and adjusted accordingly. Four more models are in a second group, Models 1, 9, 10 and 12. They are, however, far from the measured profile in some of the cases. This way of comparing models are, however, doubtful. The magnitude of the S1- and S2-values depends on the magnitude of the chloride contents for each profile. A profile with high chloride concentrations will give much higher S1-values than a profile with small chloride contents. Consequently, the different cases will contribute differently to the comparison, which can be questioned. 9.8.2 Series 2 The comparison between measured and predicted profiles are shown in Tables 9.5 and 9.6. Some predicted profiles are so unexpected low, that they have not been quantified. This is because of a mistake in the originally given profiles that were expressed in % by weight of concrete instead of % of binder. Some of the modellers did not change their predictions after new profiles were distributed. The number of predictions was small for cases 4 and 5. These cases are not further dealt with. Table 9.5 Over-view of all predictions for series 2 compared to the measured profiles, expressed as S1-values. Series 2 C1 C2 C3 C4 C5Mod S1 S1 S1 S1 S1 Average StDev

1 29.4 42.4 94.4 8.7 10.2 37.0 35.04 45.3 34.6 8.5 29.4 19.05 12.7 11.9 7.9 11.3 16.0 12.0 2.96 9.0 9.0 11.8 9.9 1.67 13.3 13.38 10.3 10.39 10.6 11.5 11.3 11.1 0.5

10 18.9 7.9 3.3 10.0 8.015 X X 4.5 4.516 26.6 26.6

Average 18.7 20.6 20.2 10.0 13.1StDev 12.6 13.9 32.9 1.8 4.1

first x1=15mm!first x1=20mm!

X data missing; based on original erroneous data

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 101 of 112 WP4 REPORT – Modelling of Chloride Ingress Table 9.6 Over-view of all predictions for series 2 compared to the measured profiles, expressed as S2-values. Series 2 C1 C2 C3 C4 C5Mod S2 S2 S2 S2 S2 Average StDev

1 -3.6 -41.5 -94.4 -8.7 0.1 -29.6 39.84 45.3 34.6 -8.4 23.8 28.45 0.0 -11.8 -7.9 -11.3 16.0 -3.0 11.66 1.5 -9.0 -11.8 -6.4 7.07 8.2 8.28 10.3 10.39 3.5 -11.5 -10.9 -6.3 8.5

10 13.3 -3.8 2.9 4.1 8.615 X X -2.7 -2.716 26.6 26.6

Average 9.8 -2.3 -19.0 -10.0 8.1StDev 15.4 25.7 33.6 1.8 11.2

first x1=15mm!first x1=20mm!

Once again, Model 5 gave good predictions, now together with Models 9 and 10. These cases are ”new” for Model 5, contrary to the ones in Series 1. Other model gave some good prediction but only dealt with one or two cases. 9.8.3 Series 3 Series 3 had given profiles after 42 years as “calibration” of the models. In spite of this, the comparison between predicted profiles after 100 years of exposure showed significant differences. One example is shown in Figure 9.9, case 1 in Series 3. It is a case with a concrete exposed in the tidal zone, which should be ”easier” than cases 2-4 which are in the atmospheric zone. The given profile after 42 years had a Cs-value of some 2.5. The predicted surface chloride contents vary from some 2.5, over around 4.0 up to 12.0 and above 30! It is also clear from the profiles that the predictions utilize very different diffusion coefficients even though a calibration was possible after 42 years.


Figure 9.9- Series 3- case 1 after 100 years exposure, with given profiles after 42 years as “calibration”.


9.9 Final comments on the benchmarking of models The most relevant remarks to be made are:

1. Regarding the sensitivity of the models to the effect of the constitutive parameters, it seems possible to deduce from the study that one of the most influencing one is the Cs value, both at short and at long term (100 years). This parameter was also identified in task 4.2. Task 4.3 confirms the importance of correct assumptions on the boundary conditions for making reliable predictions.

2. Almost all the models have Cs as the input, except a few models which have the free c as the input and convert free c to Cs with binding factors. The larger influence of Cs and less influence of D from the study is an indication that the assumption of a constant Cs in most of models is far away from the reality.

3. Better models to predict the environmental actions are urgently required. 4. With respect to D-values, they are much less influencing than the Cs-values as was

also identified in task 4.2. One aspect has to be mentioned about D values: they were given to the modellers only in some of the profiles submitted (series 1), which means that the modellers have taken the D-values they have considered the most convenient in the rest of series.

5. Where the same surface chloride conditions have been used, very different D-values have been used in different models, even though an initial profile and a measured D was given as input data. The models obviously treat chloride diffusion coefficients in very different ways.

6. Finally with respect to the methodology used to make the benchmarking, the cases could have been better chosen and the analysis could have been more thorough, but it has been shown very helpful to compare the measured and predicted profiles.


10 CONCLUSIONS The Work Package 4 of ChlorTest has three objectives:

• Analysis of models, • Correlation of lab tests to in-field performance, • Bench marking of models (regarding EU standardization needs (EN206, EC2)

In reality these objectives may be summarized in one sentence: Make sure/exemplify that test results can be used (in practice)! The work package has been divided into three tasks, 4.1 - 4.3, with the objectives “Critical evaluation of models” (4.1), “Sensitivity analysis of models” (4.2) and “Benchmarking of models” (4.3). Conclusions from the critical analysis of models General conclusions on models based on Fick’s 2nd law are

• The effect of the other ions than chloride in the pore system is completely neglected in empirical models solving Fick’s 2nd law.

• The meaning of regression parameters in empirical models is not always easy to understand. They must be very clearly defined.

• The background for a number of the assumptions made in empirical models could be questioned. This is especially important for the continuous time-dependency of the diffusion coefficients and the surface chloride contents.

• Empirical models need huge data bases: a Cs for every concrete X, every road Y and every ”water” Z!

These conclusions give rise to questions on whether empirical models can be used at all for predictions beyond where data exists. The last conclusion gives rise to questions on whether empirical models will ever be practical to use for a new structure, made of a new concrete exposed in a new environment! The main conclusions on the models based on flux equations are

• Physical models, especially those considering all ions and convection at the same time, give theoretical predictions as close to our present knowledge as possible.

• Physical models may be used to check the relevance of more simple, empirical models and may be explain some the assumptions that have to made in simple models. That may improve the confidence in empirical models.


• Physical models need quantification of a number of material parameters and a number of environmental parameters, each as a function of several parameters,

• The required surface climatic conditions are extremely complicated to model. • Because of lack of availability, limited user-friendliness and the huge amount of

required input data most physical models will remain research tools, not meant for practical applications.

Conclusions on the sensitivity analysis of models The sensitivity analysis in task 4.2 on how each model appraises environmental differences, concrete characteristics and corrosion onset was dealt with theoretically, first with a probabilistic approach on the first 10 years ingress for a selection of models and then in a general way on the ERFC-models. The sensitivity analysis with probabilistic methods has been done on two input data: the surface chloride concentration and the chloride diffusion coefficient. This study shows that the importance factors are depending on time. This remark is valid for the three tested models in spite of their very different conception. In the considered case, the influence of chloride surface concentration is more important than the one of diffusion coefficient for the three tested models. So that, we spend more efforts to increase the precision of the chloride surface concentration measure in order to increase the precision of the prediction. The chloride surface concentration is function of the environment but also of the interactions between cement paste and chloride. So the knowledge of the binding isotherm is a key of the prediction of chloride ingress as important as the diffusion coefficient. The above conclusions are based on the sensitivity analysis done on an exposure time up to ten years. The conclusions are depending on the exposure time. For longer exposure times, such as 100 years, the conclusions may be different. The long-term sensitivity analysis shows, however, that the influence of Cs is constant, while the influence of D is dependent on the C/Cs. The most sensitive parameter is the age factor in some empiric models. Conclusions on benchmarking models The bench-marking in task 4.3 aimed at studying the models’ abilities to reproduce reality, with laboratory test results as input and in-field performance data as comparison. The most relevant conclusions are: • Regarding the sensitivity of the models to the variation of the constitutive parameters, it

seems possible to deduce from the study that one of the most influencing parameter is the Cs value, both at short and at long term (100 years). This parameter was also identified in task 4.2. Task 4.3 confirms the importance of correct assumptions on the boundary conditions for making reliable predictions.

• With respect to D-values, they are much less influencing than the Cs-values as was also identified in task 4.2. One aspect has to be mentioned about D values: they were given to the modellers only in some of the profiles submitted (series 1), which means that the


modellers have taken the D-values they have considered the most convenient in the rest of series.

• Almost all the models have Cs as the input, except a few models which have the free c as the input and convert free c to Cs with binding factors. The larger influence of Cs and less influence of D from the study is an indication that the assumption of a constant Cs in most of models is far away from the reality.

• Where the same surface chloride conditions have been used, very different D-values have been used in different models, even though an initial profile and a measured D was given as input data. The models obviously treat chloride diffusion coefficients in very different ways.

• The benchmarking results show that one model (5), which was previously calibrated with most of the profiles from Series 1, gives reasonable good prediction, indicating the importance of calibration with the reliable long-term data.

• Finally with respect to the methodology used to make the benchmarking, the cases could have been better chosen and the analysis could have been more thorough, but it has been shown very helpful to compare the measured and predicted profiles.

Final conclusions For the main objective of WP4, the conclusion is obvious: There are a tremendous number of chloride ingress models available and most of them utilize test results as input data. Test results from any test method will have one or several models that require that particular data as input parameters. However, a very relevant conclusion is also: Most of these models are still not very accurate in predicting reality. The main drawback of the models is the description of the boundary conditions, i.e. the chloride content in the exposed surface in a certain environment. Especially the increasing chloride content with time is still not understood. Most models also need data from field exposure to express the boundary conditions. Results from a test method are not reliable. The time-dependency of chloride diffusion coefficients is still not well understood and obviously described in very different ways in different models. This time-dependency is not the result of any test method. Standardizing the test methods proposed by the project ChlorTest is a large step forward to improve the prediction of chloride ingress. The test methods only will, however, not be the complete answer to the required information for accurate predictions.


REFERENCES Alisa, M.; Andrade, C.; Gehlen, C.; Rodriguez, J.; Vogels, R.: Modelling of Degradation. Brussels : Brite-Euram, 1998. Project No. BE95-1347, R 4/5, 174 pp. Andrade, C. (2004), Calculation of initiation and propagation periods of service-life of reinforcements by using the electrical resistivity, International Symposium on Advances in Concrete through Science and Eng., RILEM Symposium, March 22-24, Evanston (Illinois, USA), 2004. Arliguie, G. (2003)Note de cours sur Durabilité des bétons, DEA Génie Civil, l’Institut National des Sciences Appliquées de Toulouse, France, 2003 Bamforth P.B.(1993), Concrete classifications for R.C. structures exposed to marine and other salt-loaded environments, Proceedings of Structural Faults and Repair, Edimburgh, 1993. Bamforth, P. B.: (2004) Enhanching reinforced concrete durability, Concrete Society Technical Report no 61, UK Bjergovic, D.; Krstic, V.; Mikulic, D.; Ukrinczyk, V.(1994): C-D-c-t diagrams for practical design of concrete durability parameters. Department of Materials, Faculty of Civil Engineering. University of Zagreb.

Boddy, A., Bentz, E., Thomas, M.D.A. and Hooton, R.D. (1999) An overview and sensitivity study of a multimechanistic chloride transport model. Cement and Concrete Research, Volume 29, Issue 6, June 1999, Pages 827-837

Buenfeld, N.R.; Shurafa-Daoudi M.-T.; McLoughlin I.M.(1995): Chloride transport due to wick action in concrete. Paper presented at the RILEM International Workshop on Chloride Penetration into Concrete, October 15-18, Saint-Rémy-les-Chevreuse, France. Collepardi, M.; Marcialis, A.; Turriziani, R.(1970): The kinetics of chloride ions penetration in concrete (in Italian). Il Cemento, Vol. 67, pp. 157-164. Crank, J.C. (1975), The mathematics of diffusion, 2nd edition, Oxford University Press.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 108 of 112 WP4 REPORT – Modelling of Chloride Ingress Denarié, E., Conciatori, D. & Brühwiler, E. (2003), Effect of micro climate on chloride penetration into reinforced concrete 6th CANMET/ACI International Conference on Durability of Concrete, Tessaloniki, Greece, 2003. Francy, O. (1998) Modélisation de la pénétration des ions chlorures dans les mortier partiellement saturés en eau. These du doctorat, LMDC, Université Paul Sabatier, Toulouse 1998 Frederiksen, J. M., L.-O. Nilsson, P. Sandberg, E. Poulsen, Tang L. & A. Andersen A system for estimation of chloride ingress into concrete. Theoretical background. HETEK. Danish Road Directorate Report No 83 1997. Frederiksen J.M., Geiker M.(2000), On an empirical model for estimation of chloride ingress into concrete, Proceedings of the 2 nd International RILEM Workshop "Testing and Modelling Chloride Ingress into Concrete", sept. 11-12, 2000, Paris, France (Ed. by C. Andrade & J. Kropp, RILEM, 2000), pp 355-371. Gehlen, C.; Ludwig, H.-M (1999).: Compliance Testing for Probabilistic Design Purposes: Brussels: Brite-Euram,. Project No. BE95-1347, R 8, 114 pp. Gehlen, C. (2000), Probabilistische Lebensdauerbemessung von Stahlbetonbauwerken – Zuverlässigkeitsbetrachtungen zur wirksamen Vermeidung von Bewehrungskorrosion, Heft 510 DafStb, Germany Gulikers, J. (2004), Critical evaluation of service life models applied on an existing marine concrete structure. NORECON Seminar 2004: Repair and Maintenance of Concrete Structures, Copenhagen April 19-20, 2004 Heinfling G., Courtois A., Hornet P., (1998) Application des méthodes probabilistes à l'analyse du comportement des structures en béton viellisantes: une approche industrielle, Seconde Conférence Nationale "Fiabilité des matériaux et des structures", France, 23-24 Novembre 1998 Hornet et coll., Hornet P., Dupas P., Schneiter J-R., (1998) Sensitivity Analyses on the Behavior of a Pressure Vessel during a Severe Accident using Probabilistic Methods, Severe Accidents and Topics in the NESC Project, ASME, PVP-Vol.362, pp 123-129, 1998 Houdusse O., Marchand J., Hornain H.(1998), Prédiction de l’évolution de la concentration des ions présents dans les matériaux cimentaires : modélisation numérique du transport d’ions par la méthode des éléments finis, Proceedings of the 1 st International Meeting "Material Science and Concrete Properties", march 5-6, 1998, Toulouse, France, (LMDC, 1998), pp 151-158. Houdusse O., Hornain H., Martinet G. (2000), Prediction of long-term durability of Vasco de Gama Bridge in Lisbon, Proceedings of the 5 th CANMET/ACI International Conference on Durability of Concrete, June 4-9, 2000, Barcelona, Spain, SP-192 (Ed. by V.M. Malhotra, ACI, 2000), vol. II, SP 192-20.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 109 of 112 WP4 REPORT – Modelling of Chloride Ingress Johannesson, B. (2000), Transport and sorption phenomena in concrete and others porous media, PhD thesis, report TVBM-1019, div. of Building Materials, Lund Institute of Technology, Lund, Sweden Khitab, A., Lorente, S., Ollivier, J.P.(2004), Chloride diffusion through saturated concrete: numerical and experimental results, Advances in Concrete through Science and Eng., RILEM Symposium, March 22-24, Evanston (Illinois, USA), 2004. Lee N.P., Christholm D.H.(2000), Impediments to the application of marine concrete durability models in design, Proceedings of the 2 nd International RILEM Workshop "Testing and Modelling Chloride Ingress into Concrete", sept. 11-12, 2000, Paris, France (Ed. by C. Andrade & J. Kropp, RILEM, 2000, pp 383-393. Li, L.Y. and Page, C.L. (1998), Modelling of electrochemical chloride extraction from concrete: influence of ionic activity coefficients, Computational Materials Science, 9, 1998, 303-308. Lindvall, A. (2003), Environmental actions on concrete exposed in marine and road environments and its response – consequences for the initiation of chloride induced reinforcement corrosion. PhD thesis, publication P-03:2, dept. of Building Technology – Building Materials, Chalmers University of Technology, Göteborg, Sweden Maage M., Helland S. (1991) Quality Inspection of “Shore Approach” High strength concrete. Proceedings Second CANMET/ACI International Conference on Durability of Concrete, Montreal 1991, ACI SP 126, Detroit USA . Maage, M.; Poulsen, E.; Vennesland, Ø.; Carlsen, J.E.(1995): Service life model for concrete structures exposed to marine environment initiation period. LIGHTCON Report No. 2.4, STF70 A94082 SINTEF, Trondheim, Norway. Maage M., Helland S. Carlsen J. E. (1999) Chloride penetration into concrete with light weight aggregates. Brite EuRam project BE96-3942, CUR P.O.Box 420, NL- 2800 AK Gouda, The Netherlands, May 1999. This can also be found as report R3 in http://www.sintef.no/static/BM/projects/EuroLightCon/Rapportoversikt.htm Madsen et coll., Madsen H.O., Krenk S., Lind N.C., (1986) Method of structural safety, Prentice-Hall Inc, ISBN 0-13-579475-7, 1986 Mangat, P.S., Molloy, B.T.(1994): Predicting of long term chloride concentration in concrete. Materials and Structures, Vol. 27, pp. 338-346. Marchand J., Samson E., Maltais Y., Lee R.J., Sahu S.(2002), Predicting the performance of concrete structures exposed to chemically aggressive environment. Field validation, Concrete science and engineering, vol. 4, n° 15, included in Materials and structures, vol. 35, n° 254, Dec. 2002, pp 623-631. Masi M., Colella D., Radaelli G., Bertolini L.(1997), Simulation of chloride penetration in cement-based materials, Cement and concrete research, vol. 27, n° 10, 1997, pp 1591-1601.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 110 of 112 WP4 REPORT – Modelling of Chloride Ingress McLoughlin I.M. (1997), Modelling of chloride and moisture transport in concrete, PhD thesis, dept. of Civil Engineering, Imperial College, London 1997. Meijers, S.J.H. (2003), Computational modelling of chloride Ingress in concrete, PhD thesis, Delft University, The Netherlands. Mejlbro, L.(1996): The complete solution of Fick’s second law of diffusion with time-dependent diffusion coefficient and surface concentration. Durability of marine concrete structures. CEMENTA, Danderyd. Sweden Mohammed, T. U., Yamaji, T. Hamada, H. (2002) Microstructures and interfaces in concrete after 15 years of exposure in tidal environment. ACI Materials Journal, Vol 99, No 4 – July-August 2002 Nilsson, L.O.; Ollivier, J.P. (editors) (1995): Chloride penetration into concrete. Papers presented at the RILEM “International Workshop on Chloride Penetration into Concrete”, October 15-18, 1995, Saint Rémy les Chevreuse, France. Nilsson, L-O, E Poulsen, P Sandberg, H E Sörensen, O Klinghoffer (1996) Chloride penetration into concrete. State of the Art. HETEK. Danish Road Directorate Report No 53 1996. Nilsson, L.-O.(1997), A model for convection of chloride, chapter 7 in Frederiksen et al (1997) Nilsson, L-O (2000) A numerical model for combined diffusion and convection of chloride in non-saturated concrete. Proceedings of 2nd International Workshop on Testing and Modelling the Chloride Ingress into Concrete, 11-12 September 2000, Paris Nilsson, L-O (2001) Prediction models for chloride ingress and corrosion initiation in concrete structures, Nordic Mini Seminar & fib TG 5.5 meeting, Göteborg, May 22-23, 2001. Publication P-01:6, Dept. Building Materials, Chalmers University of Technology, December 2001 Nilsson, L.-O. (2002) On the uncertainty of service-life models for reinforced marine concrete structures, International RILEM workshop Life Prediction and Age Management of Concrete Structures, Cannes, 16-17 October 2002 Nilsson, L-O (2002) Concepts In Chloride Ingress Modelling Key note paper at 3rd International RILEM workshop on Testing And Modelling Chloride Ingress Into Concrete, Madrid, 9-10 September 2002. Nguyen Xuan Son, (2004) L'approche multi-espèces pour la description de la diffusion des chloures, Mémoire bibliographique de DEA Génie Civil, LMDC, INSA de Toulouse, France, 2004

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 111 of 112 WP4 REPORT – Modelling of Chloride Ingress Nugue, F. (2002) Recherche d’une méthode rapide de détermination du coefficient de diffusion en milieu cimentaire saturé, Thèse de doctorat de l’Institut National des Sciences Appliquées de Toulouse, France, 2002

Olesen R. (1992) PROBAN User's Manual, Technical report Det Norske Veritas, 1992

Ollivier, J.-P., Carcassès, M., Bigas, J.-P. & Truc, O. (2002) Diffusion des chloures dans le béton saturé, Revue française de génie civil, Volume 6 - No.2/2002, pp 227 à 250 Petre-Lazar I., Heinfling G., Marchand J., Gérard B. (2000), Application of probabilistic methods to analysis of behavior of reinforced concrete structures affected by steel corrosion, Proceedings of the 5 th CANMET/ACI International Conference on Durability of Concrete, June 4-9, 2000, Barcelona, Spain, SP-192 (Ed. by V.M. Malhotra, ACI, 2000), pp 557-572.

Petre-Lazar, I. (2000) Evaluation du comportement en service des ouvrages en béton armé soumis à la corrosion des aciers, Thèse de doctorat de l’Université Laval de Québec, Canada, 2000 Petre-Lazar, I., L. Abdou, C. Franco, I. Sadri (2003) THI - A Physical Model for Estimating the Coupled Transport of Heat, Moisture and Chloride Ions in Concrete. 2nd International RILEM Workshop on Life Prediction and Aging Management of Concrete Structures, Paris, 5-6 May 2003, France Poulsen, E.(1990): Chlorides and 100 years of service-life (in Danish) Proceedings of “Dansk Betondag 1990”. Dansk Betonforening, Publikation No. 36. Copenhagen, Denmark. Poulsen, E.(1993): On a model of chloride ingress into concrete having time dependent diffusion coefficient. Proceedings of the Nordic Miniseminar in Gothenburg. Chalmers University of Technology. Gothenburg, Sweden. (Editor: L-O Nilsson). Report P-93:1. Poulsen, E.(1995): The LIGHTCON service life model improved at the RILEM workshop in St-Rémy (in Danish). Internal AECnote. AEC Consulting Engineers (LtD) A/S, Vedbæk, Denmark. Samson E., Marchand J.(1999), Numerical solution of the extended Nernst-Planck Model, Journal of Colloid and Interface Science, 215, 1999, pp 1-8. Samson E., Marchand J., Robert L., Bournazel J.P. (1999), Modeling the mechanisms of ion diffusion in porous media, International journal for numerical methods in engineering, 46, 1999, pp 2043-2060. Samson E., Lemaire G., Marchand J., Beaudoin J.(1999), Modeling chemical activity effects in strong ionic solutions, Computational material science, 15, 1999, pp 285-294. Siemes, A.J.M.; Gehlen, C.; Lindvall, A.; Arteaga, A.; Ludwig, H.-M.: Statistical Quantification of the Variables in the Limit State Functions. Summary: Brussels: Brite-Euram, 1999. Project No. BE95-1347, Draft R 9, 106 pp.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 112 of 112 WP4 REPORT – Modelling of Chloride Ingress Tang, L.(1996) Chloride Transport in Concrete - Measurement and Prediction, Gothenburg, Chalmers University of Technology, 1996 Tang, L. (2005) Engineering expression of the ClinConc model for prediction of free and total chloride ingress in submerged marine concrete, submitted to Cement and Concrete Research. Truc, O. (2000), Prediction of Chloride Penetration into Saturated Concrete. Multi-Species Approach, Thèse de doctorat de l'Université de technologie de Chalmers et de l'Institut National des Sciences Appliquées, Göteborg et Toulouse, 2000, 173 p Tuutti, K. Corrosion of steel in concrete. Research report No.4.82Swedish Cement and Concrete Research Institute (CBI), Stockholm 1982 Visser, J.H.M., Gaal, G.C.M., and de Rooij, M.R (2002), Time dependency of chloride diffusion coefficients in concrete, 3rd RILEM Workshop Testing and Modelling Chloride Ingress into Concrete, Madrid, September 9-10, 2002, to be published

WP5 REPORT – FINAL EVALUATION OF TEST METHODS



ACRONYM: CHLORTEST




EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 38 WP5 REPORT – Final Evaluation of Test Methods

















ACKNOWLEDGEMENT: The present document is deliverables of Workpackage 5 – “Final Evaluation of Test Methods”. The consortium members SP, IETcc, UoA, Chalmers, EDF, TNO,, HSB, ZAG, QUB, LNEC, IBRI, INSA, LCPC, VCLC and LTH were involved in the work of this part of the project. The work was led by SP, assisted by IETcc.





TABLE OF CONTENTS

Page

1 INTRODUCTION 5


3 CONCRETE 6

4 TEST PROCEDURES AND SPECIMENS 7

5 TEST RESULTS AND PRECISION ANALYSIS 9 5.1 Primary examination 9 5.2 Consistency Test 9 5.3 Criteria for rejection of data 9 5.4 Precision analysis and results 9 5.5 Dependency analysis 24

6 DISCUSSIONS 28 6.1 Comparison of the precision results 28 6.2 Comparison of the mean values 30 6.3 Relationships between various transport parameters 32

7 CONCLUDING REMARKS AND SUGGESTIONS 34 7.1 Census opinions 34 7.2 Precision of various methods 35 7.3 Suggestions 35

REFERENCES 36 Appendix 1: Description of Method M6 (by M. Castellote, C. Andrade, and C. Alonso) Appendix 2: Description of Method R1 (by C. Andrade) Appendix 3: Summary of the results from individual laboratories Appendix 4: Received answers to the questionnaire



1 INTRODUCTION The objectives of this workpackage are

• Evaluating four test methods selected based on the results from WP2 [1], and • Producing precision data of the selected test methods according to ISO 5725.

In this workpackage 15 laboratories (14 contracted and 1 voluntary) participated in the inter-laboratory tests for evaluation of the following four selected test methods:

• D2 (NT BUILD 443), as for non-steady state diffusion; • M4 (NT BUILD 492), as for non-steady state migration; • M6 (steady-state migration test based on the conductivity measurement), as for

steady-state as well as non-steady state migration; and • R1 (resistivity test), as for rapid indication of transport property.

Six types of concrete manufactured with the combination of four types of binder (PC, and PC blended with silica fume, slag and fly ash) were used in the evaluation. This report presents the results from the inter-laboratory test and the precision analysis. 2 PARTICIPATING LABORATORIES The laboratories participated in the inter-laboratory test for various test methods are listed in Table 2.1.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 2.1 – Participating laboratories for various test methods.

No. Partner D2 M4 M6 R1

P1 SP X X X X

P2 IETcc X X X X

P3 UoA X X X X

P4 Chalmers X X - X

P7 EDF - X X (failed) X

P8 TNO - X X X

P9 HSB X X X X

P10 ZAG Participated but no result reported

P11 QUB X X X X

P12 LNEC X X X X

P13 IBRI X X X

P14 INSA X X X X

P15 LCPC - X X X

P16 VCLC X - X X

P17* LTH - X (partly) - - X: participated; * Voluntary. 3 CONCRETE Six types of concrete were used in the inter-laboratory test. The mixture proportions of concrete are listed in Table 3.1. Four types of binder were tested. Swedish natural sand and gravel were used as fine and coarse aggregates, respectively. Concrete slabs of size 1200×800×200 mm were cast at SP in May 2004 and cured without demoulding under moist condition (covered with thick plastic foils) at the room temperature for at least four weeks, and afterwards the moist condition was not assured even though the slabs were still covered with thick plastic foils. After about five months storage in the laboratory the cores of size ∅100×200 mm (∅75×200 mm for participant No. 2) were taken from the slabs. Three cores per concrete type were randomly selected (except for participant No. 2 who needed ∅75 mm cores) and distributed to each participating laboratory in December 2004.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 3.1 – Mixture proportions of concrete used in the inter-laboratory test.

Material PC50 PC42 PC35 SF42 FA42 SL42

Cement type Swedish CEM I 42.5 N BV/SR/LA

Norwegian CEM II/A-V 42.5 R

(∼18% fly ash)

Dutch CEM III/B 42.5 LH HS (∼70% slag)

Cement 400 420 450 389.5 410 410

Silica fume (ELEKM)

20.5

Water 200 176.4 157.5 172.2 172.2 172.2

Sand (0-8 mm) 920 926 904 897 901 901

Gravel (10-15 mm)

816 855 904 897 901 901

SP (CemPlux) % of binder

0.5 1.0 0.5 0.5 0.5

w/b 0.5 0.42 0.35 0.42 0.42 0.42 4 TEST PROCEDURES AND SPECIMENS Methods D2 and M4 are already the Nordic standard, NT BUILD 443 and 492, respectively. The descriptions of Methods M6 and R1 are attached in Appendixes 1 and 2. To share the specimens and preconditioning procedures, Method R1 was carried out in combination with Methods M4 and M6. It should be noticed that in M4, saturated lime water is used for vacuum saturation, while in M6 demineralised water is used for vacuum saturation. The test specimens for various methods were cut at individual laboratories according to the schematic shown in Figure 4.1. It should be noticed that, owing to insufficient vibration under casting, some defects (air voids or cavities) were found in some cores, as shown in Figure 4.2. In order to keep away the visual defects, the schematic in Figure 4.1 may not exactly be followed in some laboratories. Depending on laboratories, the unavoidable visual defects in specimens were handled in different ways. Some of laboratories did nothing while some used silicon or other sealant to fill up the voids or cavities. The effect of defects will be discussed later. The inter-laboratory tests started in January 2005. Most of the participating laboratories finished the tests by the end of March and some of them by the end of June.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 38 WP5 REPORT – Final Evaluation of Test Methods Figure 4.1 – Schematic for specimen cutting.

Figure 4.2 – Example of the defects in some specimens (by courtesy of UoA).

20 ± 2 mm for M6 and R1 (3 specimens from 3 cores)


60-70 mm for D2 (3 specimens from 3 cores)

The rest for initial Cl according to D2(one specimen is enough)

Cl exposure surfaceFor M6


Cl exposure surfaceFor D2

Trowelledsurface



60-70 mm for D2 (3 specimens from 3 cores)

The rest for initial Cl according to D2(one specimen is enough)



Cl exposure surfaceFor D2

Trowelledsurface

Trowelledsurface


5 TEST RESULTS AND PRECISION ANALYSIS The test results reported by individual laboratories are summarised in Appendix 3. These results were double-checked by each laboratory to assure the correctness. The precision analysis was carried out according to ISO 5725:94. 5.1 Primary examination The raw data were visually examined for possible abnormal values. Two values in the parameter Dns from Method M6 seem significantly abnormal, that is, Lab 11: PC35 test 2, the value 105.73 in (24.07, 105.73, 19.8) Lab 11: SL42 test 2, the value 741.11 in (20.57, 741.11, 9.88) These two values were excluded before the consistency test. 5.2 Consistency Test Similar to WP2, two methods as described in ISO 5725-2:94 were used for the consistency test, that is, Mandel’s k-statistic and Cochran’s test for within-laboratory consistency, and Mandel’s h-statistic and Grubbs’ test (single only) for between-laboratory consistency. According to the standard, the significant level α = 0.01 was used for the criteria of outlier and α = 0.05 for the criteria of straggler. The consistency test was carried out only once, but if a laboratory was excluded as outlier, that is, all the data from this laboratory were rejected due to many outliers, the consistency test was carried out again. 5.3 Criteria for rejection of data To assure the proper rejection of data, the consistency test results from both methods should be consistent. Therefore, to reject one set of data, it should be classified as outlier by both Mandel’s k -statistic and Cochran’s tests or by both Mandel’s h -statistic and Grubbs’ (single) tests. 5.4 Precision analysis and results 5.4.1 Method D2 From both Mandel’s h -statistic and Grubbs’ tests it was found that most of the data from Lab 16 were outliers, as shown in Figure 5.1 and Table 5.1. Therefore, this laboratory was rejected as outlier, and the consistency test was carried out again after rejection of Lab 16. The results from the second consistency test are shown in Figure 5.2 and Tables 5.2 and 5.3. Based on Figure 5.2 and Table 5.2, two outliers (PC35 and FA42 from Lab 13) were excluded from the precision calculation. Since the value Cs is accompanied with Dns and the value K is mathematically derived from Dns, the same rejection as applied to Dns was applied to both Cs and K without further consistency test. The precision results are shown in Tables 5.4-5.6.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 38 WP5 REPORT – Final Evaluation of Test Methods Figure 5.1 - Mandel’s k-statistic and h-statistic of value Dns from Method D2. Table 5. 1 – Grubbs’ test of value Dns from Method D2

Level j PC50 PC42 PC35 SF42 FA42 SL42Valid lab p 8 9 8 9 10 9

Single G cr(1%) 2.274 2.387 2.274 2.387 2.482 2.387Single G cr(5%) 2.126 2.215 2.126 2.215 2.29 2.215Single low G 1 0.620 1.132 0.619 0.430 0.389 0.641

Single high G p 2.383 2.327 2.449 2.663 2.843 2.610Single-low outlier criterion Correct Correct Correct Correct Correct Correct

Single-low outlier LabSingle-high outlier criterion Outlier Straggler Outlier Outlier Outlier Outlier

Single-high outlier Lab Lab 16 Lab 16 Lab 16 Lab 16 Lab 16 Lab 16

Within-laboratory Consistency StatisticAccording to ISO 5725

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42

1% signif icance for PC50


Between-laboratory Consistency StatisticAccording to ISO 5725

-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 38 WP5 REPORT – Final Evaluation of Test Methods Figure 5.2 - Mandel’s k-statistic and h-statistic of value Dns from Method D2, after rejection of all the data from Lab 16.


0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42

1% signif icance for FA42



-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.2 - Cochran's test results for value Dns from Method D2, after rejection of all the data

from Lab 16 Level j PC50 PC42 PC35 SF42 FA42 SL42

Valid lab p 7 8 7 8 9 8Mean n 3 3 3 3 3 3C cr(1%) 0.664 0.615 0.664 0.615 0.573 0.615C cr(5%) 0.561 0.516 0.561 0.516 0.478 0.516

Cochran's test statistic C 0.615 0.492 0.720 0.373 0.609 0.371Outlier criterion Straggler Correct Outlier Correct Outlier Correct

Outlier Lab Lab 13 Lab 13 Lab 13 Table 5.3 - Grubbs’ test results for value Dns from Method D2, after rejection of all the data

from Lab 16 Level j PC50 PC42 PC35 SF42 FA42 SL42

Valid lab p 7 8 7 8 9 8Single G cr(1%) 2.139 2.274 2.139 2.274 2.387 2.274Single G cr(5%) 2.02 2.126 2.02 2.126 2.215 2.126Single low G 1 0.961 1.610 1.732 1.714 1.487 1.438


Single-low outlier LabSingle-high outlier criterion Correct Correct Correct Correct Correct Correct

Single-high outlier Lab Table 5.4 – Results of precision analysis for value Dns [×10-12 m2/s] from Method D2.

Level j PC50 PC42 PC35 SF42 FA42 SL42Number of valid lab p 7 8 6 8 8 8

General mean m 17.54 15.65 4.66 4.073 1.455 1.498Repeatability variance s r

2 16.55 24.205 0.33918 0.21113 0.05042 0.15886Between-lab variance s L

2 15.545 6.844 1.13061 0.96741 0.02827 0.04298Reproducibility variance s R

2 32.095 31.049 1.46979 1.17854 0.07869 0.20184Repeatability std. dev. s r 4.07 4.92 0.582 0.459 0.225 0.399

Reproducibility std. dev. s R 5.67 5.57 1.212 1.086 0.281 0.449Repeatability COV(s r) 23.2% 31.4% 12.5% 11.3% 15.5% 26.6%

Reproducibility COV(s R) 32.3% 35.6% 26.0% 26.7% 19.3% 30.0%γ = s R/s r 1.39 1.13 2.08 2.37 1.25 1.13

Repeatability limit r = 2.8s r 11.4 13.78 1.63 1.285 0.63 1.117Reproducibility limit R = 2.8s R 15.9 15.6 3.39 3.04 0.79 1.26

Number of excluded outliers 1 1 2 1 2 1Outlier type G G C, G G C, G G

Outlier laboratories Lab 16 Lab 16 Lab 13, 16 Lab 16 Lab 13, 16 Lab 16

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.5 – Results of precision analysis for value Cs [mass% of sample] from Method D2.




2 0.0111602 0 0.0234493 0.0222022 0.0104164 0Reproducibility variance s R





Number of excluded outliers 1 1 2 1 2 1Outlier type G G C, G G C, G G

Outlier laboratories Lab 16 Lab 16 Lab 13, 16 Lab 16 Lab 13, 16 Lab 16 Table 5.6 – Results of precision analysis for value K [mm/√yr] from Method D2.

Level j Level 1 Level 2 Level 3 Level 4 Level 5 Level 6Number of valid lab p 7 8 6 8 8 8








Number of excluded outliers 1 1 2 1 2 1Outlier type h1 h1 k2, h1 h1 k2, h1 h1

Outlier laboratories Lab 16 Lab 16 Lab 13, 16 Lab 16 Lab 13, 16 Lab 16

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 38 WP5 REPORT – Final Evaluation of Test Methods 5.4.2 Method M4 The results from the consistency test are shown in Figure 5.3 and Tables 5.7 and 5.8. Only one outlier (FA42 from Lab 15) was excluded according to the criteria. The precision results are shown in Table 5.9. Figure 5.3 - Mandel’s k-statistic and h-statistic of value Dns from Method M4.


0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42




-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.7 - Cochran's test results for value Dns from Method M4


Mean n 3 3 3 3 3 3C cr(1%) 0.475 0.475 0.450 0.450 0.450 0.450C cr(5%) 0.392 0.392 0.371 0.371 0.371 0.371

Cochran's test statistic C 0.219 0.311 0.301 0.205 0.604 0.350Outlier criterion Correct Correct Correct Correct Outlier Correct

Outlier Lab Lab 15 Table 5.8 – Grubbs’ test results for value Dns from Method M4





Single-high outlier Lab Table 5.9 – Results of precision analysis for value Dns [×10-12 m2/s] from Method M4.








Repeatability limit r = 2.8s r 3.75 5.49 2.069 2.528 1.439 1.1Reproducibility limit R = 2.8s R 7.2 9 2.76 4.49 2.02 1.65

Number of excluded outliers 0 0 0 0 1 0Outlier type C

Outlier laboratories Lab 15

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 38 WP5 REPORT – Final Evaluation of Test Methods 5.4.3 Method M6 For the value Ds, the results from the consistency test are shown in Figure 5.4 and Tables 5.10 and 5.11. Three outliers (PC42 from Lab 11, FA42 and SL42 from Lab 16) were excluded according to the criteria. The precision results are shown in Table 5.12. Figure 5.4 - Mandel’s k-statistic and h-statistic of value Ds from Method M6.


0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42




-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.10 - Cochran's test results for value Ds from Method M6


Mean n 3 3 3 3 3 3C cr(1%) 0.504 0.504 0.504 0.504 0.504 0.504C cr(5%) 0.417 0.417 0.417 0.417 0.417 0.417

Cochran's test statistic C 0.312 0.525 0.267 0.361 0.708 0.583Outlier criterion Correct Outlier Correct Correct Outlier Outlier

Outlier Lab Lab 12 Lab 16 Lab 16 Table 5.11 - Grubbs’ test results for value Ds from Method M6





Single-high outlier Lab Table 5.12 – Results of precision analysis for value Ds [×10-12 m2/s] from Method M6.









Number of excluded outliers 0 1 0 0 1 1Outlier type C C C

Outlier laboratories Lab 12 Lab 16 Lab 16

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 38 WP5 REPORT – Final Evaluation of Test Methods For the value Dns, the results from the consistency test are shown in Figure 5.5 and Tables 5.13 and 5.14. It should be noticed that two abnormal values from Lab 11 as mentioned in Chapter 1 have primarily been excluded before the consistency test. Four outliers (PC50 and PC35 from Lab 13, SF42 from Lab 15 and FA42 from Lab 8) were excluded according to the criteria. The precision results are shown in Table 5.15. Figure 5.5 - Mandel’s k-statistic and h-statistic of value Dns from Method M6.


0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42




-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 19 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.13 - Cochran's test results for value Dns from Method M6


Mean n 3 3 3 3 3 3C cr(1%) 0.504 0.536 0.504 0.504 0.504 0.504C cr(5%) 0.417 0.445 0.417 0.417 0.417 0.417

Cochran's test statistic C 0.884 0.462 0.723 0.609 0.615 0.302Outlier criterion Outlier Straggler Outlier Outlier Outlier Correct

Outlier Lab Lab 13 Lab 13 Lab 13 Lab 15 Lab 8 Table 5.14 - Grubbs’ test results for value Dns from Method M6




Single-low outlier LabSingle-high outlier criterion Straggler Correct Correct Straggler Correct Correct

Single-high outlier Lab Lab 13 Lab 15 Table 5.15 – Results of precision analysis for value Dns [×10-12 m2/s] from Method M6.









Number of excluded outliers 1 0 1 1 1 0Outlier type C C C C

Outlier laboratories* Lab 13 Lab 13 Lab 15 Lab 8 * Two abnormal values in Lab 11: PC35-test2 and SL42-test2 have primarily been excluded.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 20 of 38 WP5 REPORT – Final Evaluation of Test Methods 5.4.4 Method R1 For the resistivity measured on the specimens for Method M4 or alike, the results from the consistency test are shown in Figure 5.6 and Tables 5.16 and 5.17. Three outliers (PC50 from Lab 11, PC42 from Lab 2 and SL42 from Lab 8) were excluded according to the criteria. The precision results are shown in Table 5.18. Figure 5.6 - Mandel’s k-statistic and h-statistic of the resistivity (R1 on M4 specimens).


0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42




-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 21 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.16 - Cochran's test results for the resistivity (R1 on M4 specimens)


Mean n 3 3 3 3 3 3C cr(1%) 0.504 0.475 0.475 0.504 0.504 0.504C cr(5%) 0.417 0.392 0.392 0.417 0.417 0.417

Cochran's test statistic C 0.557 0.284 0.327 0.427 0.457 0.577Outlier criterion Outlier Correct Correct Straggler Straggler Outlier

Outlier Lab Lab 11 Lab 8 Lab 8 Lab 8 Table 5.17 - Grubbs’ test results for the resistivity (R1 on M4 specimens)




Single-low outlier LabSingle-high outlier criterion Correct Outlier Correct Correct Correct Correct

Single-high outlier Lab Lab 2 Table 5.18 – Results of precision analysis for the resistivity [Ω⋅m] (R1 on M4 specimens).








Repeatability limit r = 2.8s r 10.08 18 26.29 62.8 91.14 213.11Reproducibility limit R = 2.8s R 28.8 25.6 73.3 116.5 303.6 543.8

Number of excluded outliers 1 1 0 0 0 1Outlier type C G C

Outlier laboratories Lab 11 Lab 2 Lab 8

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 22 of 38 WP5 REPORT – Final Evaluation of Test Methods For the resistivity measured on the specimens for Method M6 or alike, the results from the consistency test are shown in Figure 5.7 and Tables 5.19 and 5.20. Two outliers (PC42 and SL42 from Lab 11) were excluded according to the criteria. The precision results are shown in Table 5.21. Figure 5.7 - Mandel’s k-statistic and h-statistic of the resistivity (R1 on M6 specimens).


0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

k-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42




-3

-2

-1

0

1

2

3

Lab 1

Lab 2

Lab 3

Lab 4

Lab 7

Lab 8

Lab 9

Lab 10

Lab 11

Lab 12

Lab 13

Lab 14

Lab 15

Lab 16

Lab 17

Laboratory

Man

del's

h-s

tatis

tic

PC50

PC42

PC35

SF42

FA42

SL42



EU-Project CHLORTEST G6RD-CT-2002-00855 Page 23 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 5.19 - Cochran's test results for the resistivity (R1 on M6 specimens)


Mean n 3 3 3 3 3 3C cr(1%) 0.664 0.615 0.615 0.615 0.615 0.615C cr(5%) 0.561 0.516 0.516 0.516 0.516 0.516

Cochran's test statistic C 0.530 0.732 0.423 0.364 0.535 0.637Outlier criterion Correct Outlier Correct Correct Straggler Outlier

Outlier Lab Lab 11 Lab 14 Lab 11 Table 5.20 - Grubbs’ test results for the resistivity (R1 on M6 specimens)





Single-high outlier Lab Table 5.21 – Results of precision analysis for the resistivity [Ω⋅m] (R1 on M6 specimens).








Repeatability limit r = 2.8s r 69.75 30.94 68.49 95.17 185.89 166.21Reproducibility limit R = 2.8s R 201.3 160.3 174.3 228.3 438.9 621

Number of excluded outliers 0 0 0 0 0 0Outlier type

Outlier laboratoriesNumber of excluded stragglers 0 1 0 0 0 1

Straggler type C CStraggler laboratories Lab 11 Lab 11


5.5 Dependency analysis The weighted regression procedure as specified in ISO 5725 was used for dependency analysis. Since in most cases the correlation coefficients for three types of relationship are very similar, as shown in Figure 5.8, the simplest one, that is, s = a⋅m, was used for describing the dependency. In such a relationship, the slope a describes the mean COV (Coefficient of Variation), and thus it is very easy to compare the precision of different test methods. The results are shown in Figures 5.9 to 5.12.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 5 10 15 20

General mean m (D ns from Method M4, ×10-12 m2/s)

Repr

oduc

ibili

ty s

td. d

ev. s

R

Actual points

SR = 0.236m, R² = 0.932

SR = 0.31 + 0.153m, R² = 0.932

SR = 0.364m^0.733, R² = 0.921

Figure 5.8 – Example of dependency analysis using three different relationships.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 5 10 15 20


Rep

eata

bilit

y/Re

prod

ucib

ility

s

Actual points of repeatability std dev

Actual points of reprodicibility std dev

Sr = 0.152m, R² = 0.857

SR = 0.236m, R² = 0.932

Figure 5.9 – Dependency of repeatability/reproducibility on mean value (Method M4).


0

1

2

3

4

5

6

7

0 5 10 15 20 25

General mean m (D ns from Method D2, ×10-12 m2/s)

Repe

atab

ility

/Rep

rodu

cibi

lity

s Actual points of repeatability std dev


Sr = 0.201m, R² = 0.938

SR = 0.283m, R² = 0.993

0

2

4

6

8

10

12

0 10 20 30 40 50 60 70 80

General mean m (K from Method D2, mm/yr½)

Repe

atab

ility

/Rep

rodu

cibi

lity

s



Sr = 0.09m, R² = 0.757

SR = 0.135m, R² = 0.88

0

0.05

0.1

0.15

0.2

0.25

0.3

0 0.2 0.4 0.6 0.8 1 1.2

General mean m (C s from Method D2, mass% of sample)

Repe

atab

ility

/Rep

rodu

cibi

lity

s



Sr = 0.177m, R² = 0.143

SR = 0.223m, R² = 0.412

Figure 5.10 – Dependency of repeatability/reproducibility on mean value (Method D2).


0

0.5

1

1.5

2

2.5

3

0 1 2 3 4

General mean m (D s from Method M6, ×10-12 m2/s)

Repe

atab

ility

/Rep

rodu

cibi

lity



Sr = 0.22m, R² = 0.918

SR = 0.759m, R² = 0.949

0

5

10

15

20

25

0 5 10 15 20 25


Rep

eata

bilit

y/Re

prod

ucib

ility



Sr = 0.357m, R² = 0.388

SR = 0.872m, R² = 0.9

Figure 5.11 – Dependency of repeatability/reproducibility on mean value (Method M6).


0

50

100

150

200

250

0 100 200 300 400 500 600 700

General mean m (ρ from Method R1/M4, Ω ⋅m)

Rep

eata

bilit

y/Re

prod

ucib

ility

s



Sr = 0.105m, R² = 0.968

SR = 0.251m, R² = 0.998

0

50

100

150

200

250

300

350

400

0 100 200 300 400 500 600 700 800

General mean m (ρ from Method R1/M6, Ω ⋅m)

Repe

atab

ility

/Rep

rodu

cibi

lity

s



Sr = 0.16m, R² = 0.705

SR = 0.485m, R² = 0.953

Figure 5.12 – Dependency of repeatability/reproducibility on mean value (Method R1).

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 28 of 38 WP5 REPORT – Final Evaluation of Test Methods 6 DISCUSSIONS 6.1 Comparison of the precision results The mean COV-values of repeatability and reproducibility from both WP2 and WP5 are summarised in Table 6.1 and Figure 6.1. It can be seen that the precision of Method D2 becomes significantly better in WP5, probably due to the fact that the operators in WP5 became better experienced after the practice through WP2. There is no remarkable difference in precision of Methods M4, M6 (Ds) and R1 (M4) from both WP2 (50-60 mm thick specimens saturated with demineralised water) and WP5 (50 mm thick specimens saturated with lime-water), indicating that the defects (air voids or cavities) may not significantly influence the precision of these test methods. However, there is remarkable increase in COV for Method M6 (Dns), probably due to the fact that the air voids or cavities can make chlorides more rapidly penetrating through the specimen. The amount of chlorides penetrating through the specimen via the air voids or cavities may be limited, depending on how many defects in the specimen. This amount of chloride may significantly influence the measurement of time-lag for calculating Dns in Method M6, but may not significantly influence the total flux of chlorides, resulting in less variation in the measurement based on the flux of chlorides. The variation in R1 (M6) is larger than that in R1 (M4), probably due to the fact that the defects more significantly affect the thinner specimens (20 mm in M6 and 50 mm in M4). Therefore, it can be concluded that Method M6 is more sensitive than the other methods to the defects in specimens. Table 6.1 – Summary of precision data for various test methods.

Valid number of laboratories*

Repeatability COV (%) Reproducibility COV (%) Method

WP2 WP5 WP2 WP5 WP2 WP5

D2 (Dns) 6∼8 6∼8 29.1 20.1 42.2 28.3

D2 (Cs) 6∼8 6∼8 - 17.7 - 22.3

D2 (K) 6∼8 6∼8 13.2 9.0 20.9 13.5

M4 (Dns) 7∼8 12∼13 17.3 15.2 23.1 23.6

M6 (Dns) 5∼7 10∼11 23.7 35.7 45.4 87.2

M6 (Ds) 5∼7 10∼11 20.8 22.0 79.9 75.9

R1 (M4)** 6∼7 10∼12 9.3 10.5 23.2 25.1

R1 (M6) - 7∼8 - 16.0 - 48.5 * Excluding rejected outliers. ** In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 29 of 38 WP5 REPORT – Final Evaluation of Test Methods Note: In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water. Figure 6.1 – Comparison of COV for various methods. From Table 6.1 it can be seen that the parameter K from Method D2 reveals the lowest COV values (repeatability COV 9.0% and reproducibility COV 13.5%). It should be noted,

however, that K is proportional to the square root of Dns, resulting in ns

ns

21

DD

KK ∂

⋅∝∂ , that is,

the variation of K is a half of the variation of Dns, as seen in Table 6.1. Therefore, it can be concluded that, among all the methods for diffusion coefficient (Dns and Ds), Method M4 reveals the best precision (repeatability COV 15.2% and reproducibility COV 23.6%). The resistivity measurement using thick specimens, that is, Method R1 (M4), reveals a better repeatability COV (10.5%) and a similar reproducibility COV (25.1%) when compared with Method M4. However, if Method R1 needs to be calibrated against any of the methods for diffusion coefficient, the variations from both R1 and the calibration method have to be summed.

0

10

20

30

40

50

Rep

eata

bilit

y C

OV,

% WP5WP2

0

20

40

60

80

100

D2(Dns)

D2(Cs)

D2 (K) M4(Dns)

M6(Dns)

M6(Ds)

R1(M4)

R1(M6)

Rep

rodu

cibi

lity

CO

V, % WP5

WP2


6.2 Comparison of the mean values The mean values of diffusion coefficient and resistivity from both WP2 and WP5 are summarised in Table 6.2 and Figure 6.2, where the error bars indicate the standard deviation of reproducibility. It can be seen that the values of diffusion coefficient measured from both WP2 and WP5 are in most cases reasonably comparable, especially for the Portland cement concrete with w/c in the range of 0.42 to 0.50, in which the measured values vary not very much. This is in agreement with the previous Nordic investigations [2,3], as shown in Figure 6.3. Therefore, it can be concluded that the qualities of concrete used in WP 5 are in the normal range. Again, Figure 6.2 shows that the deviation of Dns in Method M6 from WP5 is remarkably higher than that from WP2, due to the same reasons as discussed in 6.1, that is, the relatively thin (20 mm thick) specimens used in Method M6, coupled with the presence of defects, might help to promote the deviation. Usually it should be easier to saturate a thinner specimen as used in M6 than in M4, resulting in a lower resistivity. The results in Table 6.1 show, however, that the resistivity measured with M6 specimen is in general higher than that measured with M4 specimen. Since the liquid used in M6 (demineralised water) for saturation is different from that used in M4 (saturated lime water) in WP5, the higher resistivity measured by M6 might be attributed to the less conductive liquid used for saturation, indicating that the liquid used for saturation of specimens plays important role in this method, especially when specimens contain more capillary pores, e.g. in Concrete PC50 (w/c 0.50). Table 6.2 – Summary of mean values of diffusion coefficient and resistivity. Diffusion coefficient, ×10-12 m2/s Resistivity, Ω⋅m

Concrete Mix D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds) R1 (M4) R1 (M6)

PC 0.50 17.5 15.4 20.0 2.86 54.4 106

PC 0.45 (WP2) 17.6 16.8 17.8 2.9 70* -

PC 0.42 15.7 14.7 19.4 2.05 60.7 88.4

PC 0.35 4.66 5.94 11.1 1.38 124 172

SF 0.42 4.07 7.44 14.7 1.62 155 207

SF 0.40 (WP2) 2.04 2.61 5.11 1.16 332* -

FA 0.42 1.46 2.31 6.08 0.59 323 343

FA 0.45 (WP2) 7.43 5.4 8.19 1.01 273* -

SL 0.42 1.50 1.77 12.2 0.37 555 606

SL 0.45 (WP2) 2.5 2.51 2.67 0.45 382* - * In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 31 of 38 WP5 REPORT – Final Evaluation of Test Methods Figure 6.2 – Comparison of the mean values from various methods.

0

10

20

30

40

0.2 0.3 0.4 0.5 0.6 0.7 0.8

w/c

Dns

by

D2

and

M4,

x10

-12 m

2 /s D2 in WP2M4 in WP2D2 in WP5M4 in WP5D2 in [2]M4 in [2]D2 in [3]M4 in [3]

Figure 6.3 – Comparison of the Dns values by Methods D2 and M4 from various projects.

0

10

20

30

40

50

D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds)

D, x

10-1

2 m2 /s

PC50 PC45(WP2) PC42 PC35

0

5

10

15


D, x

10-1

2 m2 /s

SF42 SF40(WP2)

02468

101214


D, x

10-1

2 m2 /s

FA42 FA45(WP2)

0

2

4

6

8

10


D, x

10-1

2 m2 /s

SL42 SL45(WP2)


6.3 Relationships between various transport parameters Since Method D2 is based on natural immersion under a concentrated condition, which is the most close to the real situation among all the methods evaluated in WP5, the values from this method could be used as reference for comparison, as shown in Figure 6.4. According to Andrade et al [4], resistivity ρ can be converted to a steady state diffusion coefficient Ds, using the following equation:

ρ= Clk

Ds (6.1)

where kCl is a constant. When the dimension of Ds is in cm2/s and ρ in Ω⋅cm, the value of kCl is of 2×10-4 Ω⋅cm3/s, which was used in WP2 and in this report for comparison. It can be seen that the diffusion coefficients measured by Methods M4 and D2 are fairly comparable. This is in agreement with the previous investigations [2,3], as shown in Figure 6.5, where the data from previous investigations are included. Obviously, the most measurement data are in the range of ± s, where s is the reproducibility standard deviation of Method D2 (for x-axis) and M4 (for y-axis), respectively. Therefore, Method M4 should be a good substitution of Method D2 when considering the rapidity, simplicity, cost-effectiveness and measurement precision. As expected, the steady state diffusion coefficient Ds (Method M6) is always smaller than the non-steady state one Dns (Methods D2, M4 or M6). The theoretical explanations have been given elsewhere [5-7]. The values of Ds in Method M6 from WP2 are better comparable with Method D2 than from WP5. As discussed in 6.1, the defects, if any, in a specimen or any leakage from the curved surface of the specimen may cause a shorter time-lag which, in return, results in a larger diffusion coefficient. Figure 6.6 shows the relationship between the diffusion coefficients measured by Method M6 and converted from Method R1. It seems the data from both WP2 and WP5 are in fairly good agreement with the relationship given by Equation (6.1), with the consideration of large deviation in Method M6. Therefore, it can be concluded that Method R1 should be a good substitution of Method M6 when considering the rapidity, simplicity, cost-effectiveness and measurement precision.


0

5

10

15

20

25

0 5 10 15 20 25

D ns from Method D2, x10-12 m2/s

D fr

om o

ther

met

hods

, x10

-12

m2 /s

M4 (Dns)

M6 (Dns)

M6 (Ds)

R1* (M4)

R1* (M6)

* Converted by Eq. (6.1) with kCl = 200 x10-12

Ω⋅m3/s

WP2 data are in solid points

Note: In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water. Figure 6.4 – Relationships between diffusion coefficients from different methods.

0

10

20

30

40

0 10 20 30 40

D ns from Method D2, x10-12 m2/s

Dns

from

Met

hod

M4,

x10

-12 m

2 /s WP2

WP5

Ref [2]

Ref [3]

x ± s

y ± s

Figure 6.5 – Relationship between diffusion coefficients measured by Methods D2 and M4.


0

1

2

3

4

5

0 1 2 3 4 5

D s from Method M6, x10-12 m2/s

Ds

from

Mth

od R

1*, x

10-1

2 m

2 /s R1(M4),WP2

R1(M4),WP5

R1(M6),WP5

x ± s

y ± s

Note: In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water. Figure 6.6 – Relationships between diffusion coefficients measured by Method M6 and

converted from Method R1. 7 CONCLUDING REMARKS AND SUGGESTIONS 7.1 Census opinions Since the opinions to a test method are in many cases very subjective and strongly dependent on the persons, a statistic way was taken to obtain census opinions to various test methods. To do so, a questionnaire was prepared and distributed to all partners. Based on the answers from 10 laboratories to the questionnaire (see Appendix 4), the opinions to the four selected methods can be summarised in Table 7.1, where the variations (COV%) are given in parentheses, which reflect the spread of opinions. It can be seen that the opinions to Method R1 are relatively closer (always lowest variations), while to Method D2 are very discrepant (always highest variations). According to the census opinions, it can be concluded that the order of rapidity, simplicity and easiness of the method is R1, M4, M6 and D2.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 35 of 38 WP5 REPORT – Final Evaluation of Test Methods Table 7.1 – Census opinions to various methods.

D2 M4 M6 R1Rapidity regarding test duration 1.4 (36%) 4.3 (3%) 3.2 (4%) 5 (0%)

Simplicity in equipment requirement and operation 2.4 (22%) 3.7 (9%) 2.9 (12%) 4.4 (6%)

Easiness in handling specimens 2.2 (27%) 3.6 (9%) 3.7 (7%) 4.3 (5%)

Easiness in measurements 1.7 (28%) 3.8 (6%) 3.5 (6%) 4.4 (7%)

Easiness in calculating results 2.4 (15%) 4 (3%) 2.9 (10%) 4.8 (3%)

Sumary 2 (26%) 3.9 (6%) 3.2 (8%) 4.6 (4%)

Test methods

7.2 Precision of various methods From Chapter 6 it can be concluded that the precision data obtained from WP5 are in most cases comparable with or better than those from WP2, except for Method M6 (Dns), which was probably affected by some defects in the specimens. Therefore, the data from WP2 should be used for this method. Thus the precision of various methods can be listed in Table 7.2. Table 7.2 – Precision of methods for testing chloride resistance. Method Para-

meter Repeatability standard deviation sr

Reproducibility standard deviation sR

NT BUILD 443 (D2) Dns Cs K

sr = 0.201Dns (R2 = 0.94) sr = 0.177Cs (R2 = 0.14) sr = 0.090K (R2 = 0.76)

sR = 0.283Dns (R2 = 0.99) sR = 0.223Cs (R2 = 0.41) sR = 0.135K (R2 = 0.88)

NT BUILD 492 (M4) Dns sr = 0.152Dns (R2 = 0.86) sR = 0.236Dns (R2 = 0.93)

Steady state migration test (M6)

Ds Dns

sr = 0.220Ds (R2 = 0.92) sr = 0.237Dns (R2 = 0.76)*

sR = 0.759Ds (R2 = 0.95) sR = 0.454Dns (R2 = 0.92)*

Resistivity** (R1) ρ sr = 0.105ρ (R2 = 0.97) sR = 0.251ρ (R2 = 998) * Data from WP2; ** with 50 mm specimen and lime-water saturation. 7.3 Suggestions Based on the results and experiences from both WP2 and WP5, it can be suggested that

• For the measurement of diffusivity, Method M4 is a good substitution of Method D2 and Method R1 is a good substitution of Method M6, when considering the rapidity, simplicity, cost-effectiveness and measurement precision;

• Thick specimens, e.g. 50 mm, should be used in Method R1 for resistivity measurement in order to avoid the influence of defects like air voids or cavities; and


• Defects like air voids or cavities in the specimen or insufficient sealing of the curved surface of the specimen may significantly influence the measurement of parameter Dns when using Method M6.

Although method D2 has the lowest census favourite, it is the only one among all the evaluated methods, whose exposure condition is the closest to the reality (natural diffusion). Besides, this method can give some rough information about chloride binding though the value of Cs. Therefore, it is proposed that the following three methods can be recommended for further standardisation at the European level:

• Immersion test (based on NT BUILD 443) for determination of non-steady state diffusion coefficient, Dnssd, and surface total chloride content, Cs;

• Rapid migration test (based on NT BUILD 492) for determination of non-steady state migration coefficient, Dnssm, under the standardised laboratory exposure condition; and

• Resistivity test (based on the version used in Appendix 2) for determination of resistivity ρ as an indirect measurement of the transport property of concrete

All the above three proposed methods have the precision in an acceptable range, that is, repeatability COV ≤20% (11%∼20%) and reproducibility COV ≤30% (24%∼28%). Therefore, they are suitable for data exchanges and industrial applications.

Method M6 gives values of both steady state and non-steady state diffusivity, from which chloride binding can also be estimated. It needs, however, further development for improving its precision before being standardised as an EN-method for data exchanges and industrial applications.

REFERENCES [1] Castellote, M. & Andrade, C., “Pre-evaluation of different test methods”, Chlortest

WP2 report, version 2005-02-25. [2] Frederiksen, J.M., Sørensen, H.E., Andersen, A. & Klinghoffer, O., ‘HETEK, The

effect of the w/c ratio on chloride transport into concrete - Immersion, migration and resistivity tests’, HETEK Report, ed. by J.M. Frederiksen, published by the Danish Road Directorate, 1997.

[3] Tang L. & Sørensen, H.E., “Precision of the Nordic Test Methods for Measuring the

Chloride Diffusion/Migration Coefficients of Concrete”, Materials and Structures, Vol. 34, Oct. 2001, pp. 479-485.

[4] Andrade, C., Alonso, C., Arteaga, A. & Tanner, P., “Methodology based on the

electrical resistivity for the calculation of reinforcement service life”. Supplementary papers of the proceedings of the Fifth International CANMET/ACI Conference on Durability of concrete. Barcelona, Spain, 4-9 June 2000, pp 899-915.

EU-Project CHLORTEST G6RD-CT-2002-00855 Page 37 of 38 WP5 REPORT – Final Evaluation of Test Methods [5] Atkinson, A. & Nickerson, A.K., “The diffusion of ions through water-saturated

cement”, J. Materials Sci., 19 (1984) 3068-3078. [6] Nilsson, L.-O., “A Theoretical Study on the Effect of Non-linear Chloride Binding on

Chloride Diffusion Measurements in Concrete”, Div. of Building Materials, Chalmers University of Technology, Publication P-92:13, Gothenburg, Sweden, 1992.

[7] Tang, L., “Chloride Transport in Concrete - Measurement and prediction”, Doctoral

thesis, Publication P-96:6, Dept. of Building Materials, Chalmers Universities of Technology, Gothenburg, Sweden, 1996.


1

MULTI-REGIME METHOD: MEASUREMENT OF THE STEADY AND NON-STEADY STATE CHLORIDE DIFFUSION COEFFICIENTS IN A MIGRATION TEST BY MEANS OF MONITORING THE CONDUCTIVITY CHANGE IN THE ANOLYTE CHAMBER.

M. Castellote, C. Andrade, and C. Alonso Institute of Construction Science “Eduardo Torroja”, CSIC, Madrid, Spain.

1. SCOPE This procedure is for the determination of the steady and non-steady state chloride diffusion coefficients by monitoring the conductivity of the electrolyte in the anolyte chamber in a migration experiment. 2. FIELD OF APPLICATION The method is applicable to hardened specimens cast in the laboratory or drilled from field structures. The method requires cylindrical specimens of any diameter and a thickness of 15-20 mm, sliced from cast cylinders or drilled cores. The development of this method is fully described in [1]. 3. REFERENCES [1] M. Castellote, C. Andrade, C. Alonso, Measurement of the steady and non steady state chloride diffusion coefficients in a migration test by means of monitoring the conductivity in the anolyte chamber. Comparison with natural diffusion tests, Cement and Concrete Research, 31 (2001), 1411-1420. [2] ASTM Standard C 1202–97, Standard Test Method for Electrical Indication of Concrete’s ability to Resist Chloride Ion Penetration. [3] C. Andrade, Calculation of chloride diffusion coefficients in concrete from ionic migration measurements, Cement and Concrete Research, 23, (3) (1993) 724-742. 4. DEFINITIONS

• Migration: Movement of ions under the action of an external electrical field. • Time-lag : In a diffusion or migration cell, time taken by the chloride ions to

establish a constant flux through the specimen • Steady-state chloride diffusion coefficient, Ds: Parameter that characterises the

chloride transport in concrete in conditions of constant flux taking into account only transport phenomena.

• Non steady-state chloride diffusion coefficient, Dns: Parameter that involves the combined processes of transport and binding of chlorides in concrete.

Appendix 1

2

5. TEST METHOD 5.1 Principle The test consists of applying a voltage drop to a test specimen between two compartments containing, upstream, a chloride solution and, downstream, a solution without chlorides. The negative electrode is placed within the upstream compartment and therefore the external electrical potential applied forces the chloride ions to migrate through the specimen towards the downstream compartment. The test is based in measuring the amount of chlorides arriving in the downstream cell (anolyte) by means of measuring the conductivity of that solution instead of analysing chlorides in it. The concentration of chlorides is calculated through a previous empirical correlation between chloride concentration and conductivity for the specific conditions of the test. The steady-state coefficient is calculated then from the flux of chlorides through the specimen calculated from the measurement of the conductivity of the anolyte in the anodic compartment. The calculation of the non steady-state diffusion coefficient is made from the time taken by the chloride ions to establish a constant flux, that is to say, from the so-called time-lag. 5.2 Reagents and apparatus 5.2.1 Reagents

• Distilled or de-ionised water • Sodium chloride: NaCl, chemical quality. • Chemicals for chloride analysis (as required by the analysis method employed)

5.2.2 Apparatus

• Water-cooled diamond saw. • Vacuum container: capable of containing at least three specimens. • Vacuum pump: capable of maintaining a pressure of less than less than 1 mm Hg in

the container. • Migration cell: Different designs can be accepted, provided they include two

different compartments separated by the specimen without any leak. (see figure 1) • Cathode: black steel grid of the appropriate diameter for the cell or black steel bars. • anode: black steel grid of the appropriate diameter for the cell or black steel bars. • Power supply: capable of supplying 12 V DC regulated voltage with an accuracy of

±0.1 V. • Conductivimeter and conductivity cell • Voltmeter, capable of displaying voltage ±1 mV. • Two reference electrodes • Thermometer or thermocouple with readout device capable of reading to ±1 ºC.

Appendix 1

3

• Slide calliper with a precision of ±0.1 mm. • Equipment for chloride analysis (as required by the analysis method employed) • Magnetic stirrer (optional for automatic collection of data) • Data-logger (optional for automatic collection of data)

5.3 Preparation of the test specimen 5.3.1 Test specimens If a drilled core is used, the external surface has to be removed in order to avoid the interference of the chloride ions possibly contained in the surface of the core; only the inner part of the specimen has to be used. Therefore, if possible, about 30 mm thick layer should be cut off ( with a water-cooled diamond saw) and the next 15-20 mm thick slice should be cut as the test specimen. If a cast cylinder is used, cut the specimen from the central portion of the cylinder as the test specimen. 5.3.2 Preconditioning (water-saturation) Prior to the test, the specimens have to be pre-conditioned by saturating it under vacuum in order to avoid the transport of chlorides by capillary absorption. The procedure adopted in this method is those given in [2], which can be summarised as follows:

• Place the test specimens in a container (previously rinsed with demineralized water) or vacuum desiccator. Both end faces of specimen must be exposed.

• Seal desiccator and start vacuum pump. Pressure should decrease to less than 1 mm Hg (133 Pa) within a few minutes.

• Maintain vacuum for 3 h. • Fill separatory funnel or other container with de-aerated water (vigorously boiled

tapwater). • With vacuum pump still running, open water stopcock and drain sufficient water

into the container to cover the specimens (do not allow air to enter desiccator through this stopcock).

• Close water stopcock and allow vacuum pump to run for one additional hour. • Close vacuum line stopcock, then turn off pump. • Turn vacuum line stopcock to allow air to re-enter desiccator. • Keep the specimens under water (the water used in the previous steps) for 18 ± 2 h.

5.4 Procedure 5.4.1 Measurement of dimensions

• Measure the diameter of the test specimens at two positions at right angles. The mean of the measurements is taken as the diameter D. (see fig. 1-a)

Appendix 1

4

• If the cell device implies the use of some of these area for sealing the cell, only the diameter of free path for the ions has to be taken into account. (see fig. 1-b)

• Measure the thickness of the test specimen at 4 positions equally spaced around the

circumference. The mean of the 4 measurements is taken as the thickness, l. (see fig. 1-c)

a) b) c) a) b) c) Figure 1: Schematic representation of the measurement of dimensions of the specimen 5.4.2. Assemblage of the cell Fit the specimen in the cell, by the appropriate mechanism depending on the type of cell, introducing the electrodes into the corresponding holes (depending if using bars or plates for the electrodes, the electrodes have to be mounted from the beginning or after having the cell assembled). (see figure 2) Fill in the downstream compartment with distilled or de-ionised water. Wait for a few minutes and check that there are no leaks. Then fill in the upstream solution (1 M NaCl solution). The exact amount introduced in both compartments, Va, Vc (ml) has to be recorded precisely and the exact concentration of the chloride solution has to be known by analysis. Data can be taken automatically by means of a conductimeter and voltmeter connected to a data-logger. If so, introduce the conductivity cell into the anodic compartment and the reference electrodes in both compartments as close as possible from the sample. If the conductivity cell is not equipped with a temperature sensor and correction of temperature, introduce also a termocouple and connect it to the data-logger. It is very important to calibrate the conductivity cell, with the appropriate standard solution, prior and at the end of the test. Data can be also taken manually; if so, the conductivity cell and reference electrodes have to be introduced in the corresponding compartments only in the moment of taking the data, and they are kept outside until next measurement. In this case, it is recommended to calibrate the conductivity cell before each measurement.

Appendix 1

5

Connect the cathode to the negative pole and the anode to the positive pole of the power supply.

∆V

Magnetic stirrer

Power supply12 V DC

- +

- +

reference electrodes Conductivity cell

Temp. meas.Black steelelectrodes

specimen

NaCl1 M

Distilled water

∆V

Magnetic stirrer

Power supply12 V DC

-- +

-- +

reference electrodes Conductivity cell

Temp. meas.Black steelelectrodes

specimen

NaCl1 M

Distilled waterUPSTREAM

(CATHOLYTE)

DOWNSTREAM(ANOLYTE)

Figure 2: Schematic representation of the set up of the test for automatic data collection. 5.4.3. Migration test Turn on the power supply. Set the voltage at 12 Volts ± 0.1 V. At periodic intervals, the effective voltage, the conductivity and temperature in the anolyte compartment have to be measured. For doing that, as a first step, the effective voltage drop has to be taken by means of placing two reference electrodes near both faces of the specimen (see detail in figure 3) and then, the current has to be switched off during 5 minutes during which the anolyte liquid is stirred by the magnetic stirrer. After these minutes, the stirrer is switched off and the

Appendix 1

6

measurement of the temperature and conductivity of the anolyte can be taken. Then, connect again the power supply.

∆V

Power supply12 V DC

-- +

-- +

reference electrodes Figure 3: Schematic representation of the set up of the test: taking the data of effective voltage drop If the conductimeter is equipped with automatic correction of the values to the reference temperature of 25ºC, measurement of the temperature is not necessary. If data are taken automatically the frequency of measurements can vary depending on the equipment; as a general rule a data every 3-6 hours can be enough. If data are taken manually, at least a recording of the corresponding data every 12 hour has to be taken and a magnetic stirrer is not necessary as the liquid is homogenized when introducing and taking out the different electrodes and conductivity cell. It has to be pointed out that the oxidation of the anodic electrode (ordinary steel) takes place and therefore the anodic solution becomes brownish. This is necessary for avoiding evolution of chlorine gas at the anode. These oxides do not perturb neither the experiment nor the measurement of conductivity. The evolution of the conductivity in the anodic chamber follows the trend in three steps given in figure 4. The test is finished when step 3 has been clearly reached, and depends on the type of sample. As a general rule, two weeks are enough for most of samples.

Appendix 1

7

0 50 100 150 200 250 300 350

time

0

10

20

30

40

50

60mmol Cl- anolyte

steady-state

τ

Step1 Step

2

Step3

0 50 100 150 200 250 300 350

time

0

10

20

30

40

50

60mmol Cl- anolyte

steady-state

τ

Step1 Step

2

Step3

Figure 4: Schematic representation of the evolution of the conductivity and amount of chlorides in the anolyte during the test. 6. TEST RESULTS AND CALCULATIONS To make this part easy to understand, an example is given simultaneously with the description of the steps. Data for the example are given in table 1. For the calculation of the steady state and non steady state diffusion coefficients by this method, the following steps have to be followed: 1: Record the evolution during the test of the following parameters: effective voltage Ve (V), conductivity (mS/cm) and temperature (ºC) (if no automatic correction for conductivity is provided by the conductimeter). 2: Calculate the values of conductivity at the reference temperature of 25ºC using equation (1) (1) TT cTcc )25(02.025 −+=Where: c25 = conductivity (mS/cm) at 25ºC. cT = conductivity (mS/cm) measured at temperature T (ºC). 3: Calculate the values of accumulated chlorides according to equation (2) [1] Cl-(mmol) = (-1.71+ 11.45 * c25 ) Va. (2)

Appendix 1

8

where Va is the volume of the anolyte (l) If according to the equation (2) [Cl-] (mmol) <0, then take Cl-(mmol) = 0 Voltage applied (V) 12 l (cm) 1initial [Cl] catholyte (mol/l) 1.011 φ (cm) 7.5 Va (l) 0.414

time effective voltage conductivity, cT Temperature corrected conductivity, c25 Cl anolyte (hours) (V) mS/cm (ºC) mS/cm mmol

0.00 3.20 0.09 19.50 0.10 0.003.00 4.00 0.17 19.50 0.19 0.1916.67 6.77 0.50 18.50 0.57 1.9724.50 5.14 0.33 18.50 0.37 1.0640.75 6.33 0.36 18.00 0.41 1.2447.92 6.70 0.44 18.00 0.50 1.6766.09 8.27 0.97 17.00 1.13 4.6272.25 7.92 1.39 16.50 1.63 7.0088.00 8.65 2.74 15.50 3.26 14.7497.75 9.46 3.89 17.00 4.51 20.68115.42 9.86 5.30 14.00 6.47 29.94139.75 10.30 6.54 14.00 7.98 37.11150.00 10.50 6.52 15.00 7.82 36.37

Table 1: Example of the test results and calculations. Steps 1-3. 4: Draw the evolution of the amount of chlorides in function of the time and from it, select the period of the steady state and therefore the points indicating the steady state initiation (ssi) and the end of the steady state (ssf). During this period, the corresponding points have to be fitted to a linear regression (see figure 5). In the case of the example, this period can be taken from 66.09 (ssi) to 115.42 hours (ssf), where the correlation obtained is very good.

0

5

10

15

20

25

30

35

40

0 50 100 150

time (hours)

mm

ol C

l ano

lyte

ssi

ssf

y = 0.5196x - 30.28R2 = 0.9977

-40

-30

-20

-10

0

10

20

30

40

50

60

0 50 100 150

time (hours)

mm

ol C

l ano

lyte

Figure 5: Example of the test results and calculations. Step 4.

Appendix 1

9

5: Calculate the steady-state diffusion coefficient, Ds, from the Modified Nernst-Planck equation (eq.3) [3].

∆ΦγCFz

RTlJ = D Cls

1

(3)

where:

( ))(

31

ifss

ssissf

StEmmolmmol

J−

−−=

J = Flux of chlorides through the specimen during the steady-state period (mol/cm2s) S : surface area of test specimen exposed to the chloride solution (cm²) t: time (in seconds) of duration of the select the period of the steady state. From the data of the example: J = 3.226E-9 mol/cm2s C1= Cl- concentration in the catholyte (mol/cm3)= 1.011E-3 γ = Activity coefficient of the catholyte solution= 0.657 ∆Φ= Averaged effective voltage (V) through the specimen during the steady-state period= 8.9 l = thickness of the sample (cm) =1. R : perfect gas constant (1.9872 cal/mol K) T : Average temperature during the test (K) =289.55 z: ion valence, for chloride, z = 1; F : Faradays constant = 23060 (cal/ V eq)

According to these data, Ds = 1.36E-8 cm2/s 6: Calculate the non steady-state diffusion coefficient, Dns, from the time-lag, t, obtained by the intersection of the straight line of chloride flux characteristic of the steady state with the X axis, and later application of equation (4).

−= 2

2coth

22

2 vvvl

Dns τ (4)

where: τ = time-lag in the migration test (s); It is obtained from the linear correlation obtained in step 4, calculating the time for chloride in anolyte equal to zero.

Appendix 1

10

In our example, τ = 58.3 hours = 209880 s l = thickness of the specimen (cm) v= ze(∆φ)t/kT; where k= Boltzmann´s constant. T : Average temperature during the test (K) =289.55 (∆φ)t: Averaged effective voltage (V) through the specimen from the beginning of the test until the time lag = 5.92 v= ze(∆φ)t/kT= 1.16E4 x 5.92 /289.55 = 237.26 For the values of voltage drop usually used in the test, equation (4) can be simplified into equation (5)

2

2 )2(2vvl

ns τ−

=D (5)

Therefore Dns = 3.98E-8 cm2/s 7. TEST REPORT The test report should include the following information:

• Name and address of the test laboratory. • Date of arrival of the material or object tested. • Brief description of the material or object tested, including if it is a core of a cast

specimen, composition, curing age and date of arrival. • The name of the method. • Name and address of the person who performed the test. • Date of the test. • Test results, including all the items given in table 2 and the obtained values for the

averaged effective voltage during the steady-state period, averaged effective voltage from the beginning of the test until the time lag, and the time-lag obtained in the migration test.

• Calculated Ds and Dns • Any observation of incidences during the test or any deviation from the test method. • Date and signature.

Appendix 1

DETERMINATION OF ELECTRICAL RESISTIVITY IN CONCRETE SPECIMENS. DIRECT METHOD. (Translation of UNE 83XXX) made by C. Andrade 1. Scope and field of application This standard describes the method for the determination of electrical resistivity of hardened concrete in specimens or cores. The testing samples will be concrete specimens or cores not containing reinforcements and of regular geometry. The specimens will be of the shape and size described in the Standards UNE83-301: 1991. Concrete resistivity of a water saturated concrete is an indirect measurement of concrete pore connectivity. In a non saturated concrete it is an indication of its degree of satuaration. 2. Standards for consult UNE-83301: 1991 – Ensayos de hormigón. Fabricación y conservación de probetas. UNE-83302: 1984 – Ensayos de hormigón. Fabricación y conservación de testigos. 3. General requirements for the testing 3.1. Number of determinations The number of determinations will be of two in each specimen or core. 3.2. Expression of electrical resistance, resistivity, dimensions and results

The electrical resistances will be expressed in Ohms, the dimensions in meters, the areas in square meters and the resistivities in ohms per meter. The results of resistivity of tests, given as the average of the two determinations, will be expressed in ohm·meter in full numbers. 3.3. Conditions of testing room The temperature of the testing room has to be between 20±2ºC and the relative humidity, HR, cannot be smaller than 45%. 3.4. Conditions for maintenance of testing specimens or cores The samples to be tested will be cured or maintained as indicated in paragraph 7 of this standard. In any case the specimens or cores cannot remain out of the saturation

Appendix 2

in water (humid chamber or under water) more than 5 minutes. In this period of time the measurements have to be carried out. 4. Fundamentals of the method

It is based in the measurement of the electrical resistance of a concrete sample by means of current lines parallel to the basement of the sample. This method is considered as the reference one for the measurement of resistivity in concrete specimens or cores. 5. Definitions 5.1. Electrical resistance of concrete, Re It is the relation between the voltage drop (V) between the internal electrodes and the current (I) that circulates between the two external ones through the specimen or core. 5.2. Electrical resistivity of concrete, ρ It is the electrical resistance by unit of volume of concrete. It is the inverse of the conductivity and it is obtained from the relation between the voltage drop and the current refereed to a standarized geometry (a cube of 1m of edge). The resistivity of a concrete saturated in water is an indirect measurement of the concrete pore connectivity. In a non water-saturated concrete represents an indication of the degree of saturation. 5.3. Cell constant (k) It is the relation between the electrical resistance and the resistivity. It is obtained through Ohm’s law. NOTE: The quantitative parameters of all definitions are not object of conformity. 5.4. Vacuum saturation It is such obtained through immersion under vacuum in the conditions specified in paragraph 7 o this standard. 6. Apparatus 6.1. Resistivimeter It is an equipment able to apply a stable voltage or current to the test specimen and to measure respectively the current or voltage drop generated. It is recommended that the equipment use an alternating signal in such a way that in the circuit circulates a current of about 40mA at a frequency between 70-100Hz. It will be also acceptable to use an alternating current source with two external multimetres to measure the current and voltage drop. A datalogger for monitoring and simultaneous measurement is also allowed.

Appendix 2

6.2. Electrodes Metallic nets (openings < 2mm) or plates will be used as electrodes. They can be of steel, copper or any other well conductor metal, free of superficial impurities (deposits, rust, oxides, etc.) and of equal dimensions of that of the transversal section of the specimen or core (see figure 1). 6.3. Contact sponges Two thin sponges (thickness < 5mm) of equal size than the electrodes, that will be placed between them and the two parallel faces of the specimen. 6.4. Water for wetting the sponges The water for wetting the sponges will be tap water. 6.5. A mass of about 2 kg Any object with at least a place face with similar dimensions than the sample face, made of 9 non conductor material, that will be used to press the upper electrode against the contact sponge and the concrete (see figure 1). 7. Sample preparation The test must be performed in a fully water saturated specimen. These conditions are fulfilled in specimens cured submerged in water. In this case care should be taken to avoid any superficial evaporation from the top surface of the specimen durring the first 24 hours when the specimens remain inside the moulds. To avoid this evaporation plastic covers are suitable. For the rest of the cases, water saturation will be achieved through the following procedure: the samples will be introduced in a hermetic container (for instance a desiccator) with two holes in order to allow making vacuum and introducing the water. After the sealing f the container, vacuum will be made to reduce the internal pressure to less than 1 mm Hg. This vacuum level will be maintained during 3 hours. With the vacuum pump functioning, the decarbonated water will be introduced in such a quantity to completely cover the samples. The vacuum will be maintained during 1 further hour. After switching off the vacuum pump, air will be allowed to enter in the container. The samples will be immersed during 18±2 additional hours. Then, the specimens will be maintained immersed or in a chamber at RH>95% until the resistivity determination. samples will be immersed during 18±2 additional hours. Then, the specimens will be maintained immersed or in a chamber at RH>95% until the resistivity determination. 8. Testing method The steps of the test will be the following: 8.1. Connections arrangement

Appendix 2

The connection arrangement is that of figure 1. The two electrodes are plugged to the resistivimeter. Mass of

2Kg

spongesELECTROD

SPECIMEN

HOLDE

Resistivimeter

Figure 1. Arrangement for the measurement of resistivity by the direct method. 8.2. Wetting and measurement of resistance of the sponges The sponges will be wetted with tap water complying with the standard UNE 83301. They will be slightly dripped in order to remove the excess of water. They will be placed between the two electrodes and the mass of 2 kg will be placed on the top of the upper electrode. Then, the current will be applied with the resistivimeter and the potential drop measured as soon as the measurement is stable (after around 5 seconds) in order to obtain the electrical resistance of the sponges, Rsp, alone. This Rsp will not be higher than 100 Ω. 8.3. Measurement of the electrical resistance Just before the measurement, the dipping water will be removed from the surface of the samples, in particular the lateral surfaces, with a cotton cloth. The sample will be placed vertically lying on one electrode and a contact sponge as shown in figure 1. On the top, the other sponge and electrode will be placed and on the top of them, the mass of 2Kg. Then, the measurement will be carried out and the electrical resistance of the sample plus the sponges (Re+sp) will be recorded after stabilization of the signal. 9. Calculation and expression of results 9.1. Calculation of the electrical resistance of the sample The electrical resistance will be obtained from the expression:

Appendix 2

Re= Re+sp - Rsp

where Re is the electrical resistance of the sample Re+sp is the electrical resistance of the sample plus the sponges

Rsp is the electrical resistance of the sponges 9.2. Calculation of cell constant The cell constant for cylindrical shape in specimens of 15cm in diameter and 30cm in height will be calculated as follows:

LSK =

where S the area of the sample face where the sponge is placed in m2

L is the height of the specimen or core in m 9.3. Calculation of the resistiviy The resistivity is then calculated from

Rek=ρ where ρ is the resistivity in Ω·m k is the cell constant in m Re is the electrical resistance in Ω 9.4. Expression of result The resistivity will be given in Ω·m in full numbers and will be calculated for each sample as the average of the six measurements carried out. It will be indicated the temperature in ºC at which the measurement was made. 10. Precision and bias No reproducibility and repetitivity indications are available. 11. References APM –219 : 1996. Laboratory AEC (Denmark) ASTM – G 57 – 78 (modification 1984). RILEM Recommendation.- Test methods for on-site measurement of resistivity of concrete.- Materials &Structures vol 33, December 2000, 603-611.

Appendix 2

Annex 1 (Informative) This standard can be applied to concrete for public works and buildings made with standard aggregates (for instance: siliceous, limestone, etc.). It cannot be applied therefore to concretes made with highweight, heavy or conductor aggregates, as well as concretes reinforced with metallic fibers.

Appendix 2

Appendix 3 1 (8)

Appendix 3: Summary of the results from individual laboratories Collection of the Original Data y ij Method: D2 x10-12 m2/s

Decimal: 1 1 2 2 2 2Laboratory PC50 PC42 PC35 SF42 FA42 SL42

Lab 1, test 1 19.3 19.8 6.31 5.37 1.72 1.87test 2 26.5 17.2 5.74 4.75 1.3 1.46test 3 19.9 18.8 5 4.03 1.86 1.06

Lab 2, test 1 15.9 16.2 4.06 4.57 1.92 0.93test 2 1.35test 3

Lab 3, test 1 12.7 7.7 2.91 2.61 1.19 0.91test 2 13.8 10.3 1.91 2.66 1.29 1.13test 3 13.1 10.3 3.57 2.83 0.93 1.12

Lab 4, test 1 19 28.8 5.31 4.07 1.51 1.74test 2 28.6 19.4 5.07 5.56 1.57 1.82test 3 22.8 18.5 5.94 5.19 1.59 1.82



Lab 9, test 1 14.5 4.9 1.58 2.24test 2 13.6 4.56 1.41 1.32test 3 15.8 1.62 2.49



Lab 12, test 1 13.6 14.8 4.03 2.2 1.24 1.19test 2 12.8 13.7 4.07 2.33 1.28 1.24test 3 13 10.8 3.38 2.59 0.87 1.26

Lab 13, test 1 17.3 7.1 5.47 4.44 0.32 1.6test 2 15.6 13 9.07 3.9 1.94 0.82test 3 31 22 4.83 3.91 1.55 2.19





Appendix 3 2 (8)

Collection of the Original Data y ij Method: D2 (Cs) mass%



Lab 2, test 1 0.673 0.622 0.626 0.68 0.674 0.673test 2 0.667test 3





Lab 9, test 1 0.612 0.902 1.386 0.608test 2 0.634 1.033 1.183 0.799test 3 0.613 1.149 0.62









Appendix 3 3 (8)

Collection of the Original Data y ij Method: D2 (K) mm/year1/2


Lab 1, test 1 72 71.1 36.4 34.7 22.6 20.8test 2 79.1 65.6 35.4 32.4 20 19test 3 70.7 67.2 35.6 32 22.9 17.5

Lab 2, test 1 55 55.7 27.9 29.5 19.1 13.3test 2 16test 3

Lab 3, test 1 56 47 37 35 23 20test 2 59 54 29 34 24 24test 3 62 52 40 38 22 19

Lab 4, test 1 69.4 81.4 34.4 29.7 20.8 19.8test 2 83.2 69.1 33.7 34.4 21.6 19.8test 3 74.8 66.7 35.6 34.2 20.6 20



Lab 9, test 1 56 36 32 23test 2 55 36 22 19test 3 59 23 24



Lab 12, test 1 58.7 60.3 31.7 25.4 16.8 16.1test 2 57 58.4 32.4 25.5 19 13.8test 3 58.6 51 31.2 25.4 16.8 16.8

Lab 13, test 1 63 43.3 31.8 31.3 10 17.4test 2 54.1 50.2 38.9 30.4 22.3 14.7test 3 73.7 61.9 29.6 29.9 19.8 16.1



Lab 16, test 1 113 82.9 61.2 101.9 52.3 29test 2 111.4 97.9 74.9 105.1 63.3 29.1test 3 123.7 84 85 109.9 54 28.2


Appendix 3 4 (8)

Collection of the Original Data y ij Method: M4 x10-12 m2/s


Lab 1, test 1 15.5 13.9 5.53 7.51 2.19 2.37test 2 15.1 20.9 5.46 5.26 2.36 2.28test 3 15 19.7 5.14 7.97 2.08 2.19

Lab 2, test 1 14.7 14.9 7.29 3.25 1.88test 2 16.5 13.3 6.08 5.76 3.09 1.57test 3 17.1 11.5 7.1 5.24 3.82 1.9

Lab 3, test 1 14.2 15.1 4.51 7.59 2.08 2.04test 2 17.1 14.4 3.79 9.63 2.55 1.84test 3 13.7 13 6.31 8.73 2.62 1.39

Lab 4, test 1 19.3 20.7 5.1 8.2 1.62 1.61test 2 16.8 17.5 5.54 9 1.36 2.37test 3 15.9 17 5.22 6.92 2.22 1.36

Lab 7, test 1 11.1 11 5.82 7.07 2.83 1.5test 2 11.6 9.8 7.64 6.56 3.12 1.84test 3 11.8 10 8.71 6.06 3.63 1.56


Lab 9, test 1 13.4 12.9 5.85 8.71 3.43 1.61test 2 17 11.6 6.03 6.51 1.95 1.22test 3 17.3 9.8 6.9 6.6 2.21 1.16


Lab 11, test 1 17.9 14.9 5.37 7.39 1.8 1.05test 2 16.1 17.1 7.47 6.78 1.6 1test 3 16.2 20.1 5.26 6.39 1.56 0.94


Lab 13, test 1 21.3 15.7 5.56 8.56 1.54 0.78test 2 17.7 17.4 5.78 9.8 1.25 1.19test 3 18.1 5.34 10.37 1.38 2.39


Lab 15, test 1 17.7 18 6.05 8.19 2.06 1.48test 2 18.3 13.1 6.2 9.15 5.95 1.6test 3 16.7 19.1 6.81 10.61 2.23 2.02


Lab 17, test 1 5.38 8.16 1.9 2.73test 2 5.87 8.69 2.94 3.54test 3 5.54 10.16 2.81 2.29

Appendix 3 5 (8)

Collection of the Original Data y ij Method: M6 (Dns) x10-12 m2/s


Lab 1, test 1 7.99 9.14 4.84 12.7 8.55 7.49test 2 7.46 10.7 4.34 10.31 17.26 12.45test 3 6.53 16.61 6.99 3.83 3.67


Lab 3, test 1 2.76 3.97 1.8 3.97 1.05 1.03test 2 1.65 2 1.84 6.35 0.74 1test 3 6.91 2.68 2.35 1.77 1.53 1.09







Lab 12, test 1 9.06 7.64 2.16 3.32 1.26 1.11test 2 10.44 9 2.4 3.93 1.56 1.2test 3 13.2 14.21 4.09 5.08 1.42 1.64

Lab 13, test 1 158.91 47 21.13 28.32 5.8 6.4test 2 71.15 65.02 22.31 26.15 5.43 5.64test 3 39.34 30.37 6.8 5.57


Lab 15, test 1 92.2 23.82 34.13test 2 26.15 82.01 14.14 25.52test 3 64.91 59.29 27.92 40.04 26.37 43.13



Appendix 3 6 (8)

Collection of the Original Data y ij Method: R1 (M4) Ω⋅m



Lab 2, test 1 129 157 210 502test 2 95test 3 74 110 91














Appendix 3 7 (8)

Collection of the Original Data y ij Method: R1 (M6) Ω⋅m







Lab 8, test 1 72 118 215 304 508test 2 67 131 200 305 524test 3 67 123 193 312 571










Appendix 3 8 (8)

Collection of the Original Data y ij Method: M6 (Ds) x10-12 m2/s


Lab 1, test 1 1.01 0.92 0.56 0.63 0.24 0.21test 2 1.1 0.9 0.37 0.71 0.29 0.46test 3 1 0.96 0.54 0.18 0.28






Lab 9, test 1 0.1 0.88 0.13 0.76 0.08 0.08test 2 0.1 1.01 0.17 0.62 0.08 0.08test 3 0.11 1 0.16 0.72 0.09 0.07




Lab 13, test 1 6.19 4.2 1.97 3.02 1.4 0.68test 2 3.03 4.22 2 1.95 1.39 0.78test 3 4.25 3.66 1.99 2.36 1.47 0.87


Lab 15, test 1 10.28 2.79 2.72 0.77 0.29test 2 7.32 3.73 2.38 3.68 0.84 0.31test 3 9.2 3.1 3.2 3.53 0.73 0.28



Appendix 4 1 (3)

Appendix 4: Received answers to the questionnaire Questionnaire for the test methods evaluated in WP 5 P1

D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 2 4 4 5

Easiness in handling specimens (easest 5 point) 2 4 4 4

Easiness in measurements (easest 5 point) 1 4 3 5

Easiness in calculating results (easest 5 point) 2 4 3 5Mean 1.6 4 3.4 4.8

Questionnaire for the test methods evaluated in WP 5 P2

D2 M4 M6 R1





Easiness in calculating results (easest 5 point) 2 4 4 5Mean 1.2 3.8 4.2 5


D2 M4 M6 R1







D2 M4 M6 R1






Please give 1-5 point to each method




Appendix 4 2 (3)


D2 * M4 M6 R1





Easiness in calculating results (easest 5 point) 2 4 1 5Mean 1.4 3.8 2 4.6

Note *: comment on D2 is based on extensive experience in the past


D2 M4 M6 R1





Easiness in calculating results (easest 5 point) 1 4 3 5Mean 1 4.6 3.6 3.2


D2 M4 M6 R1







D2 M4 M6 R1





Easiness in calculating results (easest 5 point) 3 3 3 3Mean 2.8 4.4 3.2 4.4





Appendix 4 3 (3)


D2 M4 M6 R1





Easiness in calculating results (easest 5 point) 2 4 3 5Mean 1.6 3.6 3.2 4.4


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 3 5 4

Simplicity in equipment requirement and operation (simplest 5 points) 4 5 3

Easiness in handling specimens (easest 5 point) 2 5 4

Easiness in measurements (easest 5 point) 3 5 4

Easiness in calculating results (easest 5 point) 4 5 4Mean 3.2 5 3.8



Documents

Guideline for Practical Use of Methods for Testing the ...cristina/RREst/Aulas_Apresentacoes/07... · Guidelines for practical use of methods for testing the resistance of concrete