Construction and Building Materials 23 (2009) 3475–3484

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier .com/locate /conbui ldmat

Evaluation of moisture susceptibility of porous asphalt concrete using watersubmersion fatigue tests

L.D. Poulikakos *, M.N. Partl 1

Empa, Swiss Federal Laboratories for Material Testing and Research, Überlandstrasse 129, CH-8600 Dübendorf, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history:Received 13 March 2009Received in revised form 25 June 2009Accepted 5 August 2009Available online 29 August 2009

Keywords:Porous asphaltTwin layMoisture susceptibilityMechanical testsThin sectioningFluorescent microscopy

0950-0618/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.conbuildmat.2009.08.016

* Corresponding author. Tel.: +41 44 823 4479; faxE-mail addresses: lily.poulikakos@empa.ch (L.D.

empa.ch (M.N. Partl).1 Tel.: +41 44 823 4113; fax: +41 44 821 6244.

Porous asphalt concrete due to its open structure is exposed to water and therefore susceptible to waterdamage. Effect of water, temperature and frequency on mechanical properties of porous asphalt concretewas investigated. An innovative test method developed in Switzerland is used to mechanically test150 mm diameter cylindrical cores from eight materials in dry state and while being submerged underwater. Although the conventional indirect tensile strength ratio delivers useful data about the water sen-sitivity of porous asphalt specimen, the coaxial shear test provides vital information about the develop-ment of fatigue damage in the material. Application of coaxial shear test (CAST) for a twin lay porousasphalt as well as conventional porous asphalt shows a reduction in complex modulus due to fatigueloading after each thermal cycle and due to detrimental effects of water submersion. Moisture suscepti-bility results using CAST reflected the field inspections of surface degradation. In addition, investigationof the microstructure has given insight into the mechanical behavior of selected materials.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Porous asphalt (PA) or open graded asphalt concrete is an envi-ronmentally friendly road material using advanced technology inpavement design. PA used in the top layers, usually has an air voidcontent of 20 vol%. Due to higher proportions of coarse aggregatesand lower sand content, interconnected micro voids are createdwhich, in wet weather, drain the water through a series of microconduits. This drainage system prevents aquaplaning on the roadsurface and improves visibility. The high porosity of PA also re-duces traffic induced noise emissions significantly.

Although PA has environmental benefits, it can suffer fromproblems which can affect both its performance and its service life.This is largely due to the microstructure of PA; the open structureexposes a large binder surface area to the oxidative effect of air andthe damaging effect of water resulting from increased interfacialmoisture content, leading to rapid aging of the binder, moisturedamage of the bitumen aggregate bond and structural distress ofthe compounds.

Moisture damage is an important cause of failure of asphaltconcrete pavements. Existence of moisture in the pavement canmanifest itself in the loss of cohesion within the bituminous binderitself or the loss of adhesion between binder and the aggregates.

ll rights reserved.

: +41 44 821 6244.Poulikakos), manfred.partl@

The latter results in the stripping phenomenon. This observed phe-nomenon is the de-bonding of bitumen films from aggregate sur-faces due to greater affinity of the aggregate for water than forbitumen. Furthermore, stripping can be caused by hydraulic scour-ing resulting from repeated generation of pore water pressure.Hence, stripping leads to a weakened pavement that is susceptibleto pore pressure damage and premature cracking.

Water damage in asphalt concrete mixtures in general and inporous asphalt concrete in particular is dependent on many factorsincluding but not limited to aggregate structure and type as well asbinder type and amount. Quantifying water damage in mixtures isa complicated task and the subject of research worldwide [1–3].Significant economic benefits derive from understanding the fun-damental mechanisms of failure and moisture effects in porous as-phalt in order to prolong the service life of this environmentallyfriendly pavement material.

Currently, using European standards water damage susceptibil-ity of PA is determined using empirical techniques such as indirecttensile test (IDT) [4,5], which do not always deliver conclusive re-sults [6,7]. There is a need for meaningful test procedures that pro-vide fundamental knowledge about material behavior under loadand in the presence of water.

Switzerland started using PA in 1979 with mixed results andaccording to a survey taken in 2004, only nine of the 26 Cantonsuse PA. A national cooperative research project funded by theSwiss federal government was initiated to explore the mechanicalperformance of PA in Switzerland [8]. In this study, variousmechanical tests were performed on laboratory prepared specimen

3476 L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484

and cores taken from selected pavements chosen based on feed-back from various Cantons obtained in the survey in 2004. Labora-tory tests allowed the comparison of core performance with that oflaboratory prepared specimen as well as comparison with fieldperformance. This paper focuses on the fatigue tests which wereperformed on water submerged cores from different pavementsusing the coaxial shear tests (CAST) and comparison to empiricalstandardized test results and field performance. In addition knowl-edge gained from the investigation of the microstructure of se-lected specimen is used to understand their mechanicalperformance.

2. Material selection

Based on the results of the survey from the Cantons, sites were selected for cor-ing and further investigation. The selection criteria were age, binder type and dis-tress. Table 1 lists the material, mix properties and age. The binder for all themixes except AG2 are polymer modified. At the time of coring, the sections werebetween 0 and 13 years old, with binder contents between 4.3% and 6% and air voidcontent between 14.8% and 26.1%. It is important to keep in mind that these werecores and at different states of distress and age.

Before coring the selected sections a visual inspection according to the Swissstandards [9] was carried out. To this end, the surface degradation characteristicI1 was assessed by inspection. The index I1 combines degradation in terms of sur-face distress (polishing, bleeding), pavement degradations (wear, loss of aggregate,cracks of joints, cracking) and pavement strain (rutting). The overall state is deter-mined as a function of the surface area Ai and severity of degeneration Si leading tothe total value Mi = Ai � Si. From Mi and a weighing factor Gi, a global weighted indexI1 is deduced with a rating scale which is divided in intervals of good (0 and 1), med-

Table 1Measured material properties (PmB stands for polymer modified binder).

Designation VD2 VD3 VD4 VD5

Type SN 640433b PA11 PA11 PA11 PA11Binder type PmB colflex N PmB practiplast M40 PmB styrelf

13/80PmB CTSrubber a

Sieve size (mm) % Passing (by Mass)

31.5 – – – –22.4 – – – –16 100 100 100 10011.2 98 95 95 978 65 65 49 525.6 25 25 20 204 17 17 15 152.8 13 13 13 132 10 10 11 111 7 7 7 80.5 6 6 6 70.09 5 4 4.8 4.8Binder (% by wght.

of total mix)4.8 4.8 4.8 6.0

Air voids content(% by vol.)

23.3 23.4 22.8 16

Years in service 13 12 7 6(constr date) (1991) (1992) (1997) (1998)

Table 2Summary of surface inspections.

Material designation Overall rating index I1a State

VD2 3 Sufficient to criticaVD3 1.4 MediumVD4 2 Medium to sufficieVD5 2.6 SufficientVS6 1.6 MediumAG2 2.7 SufficientAG3 – –VD10 1.4 Medium

a 0–1 = good, 1–2 = medium, 2–3 = sufficient, 3–4 = critical, 4–5 = bad.

ium (1–2), sufficient (2–3), critical (3–4) and bad (4–5). The global weighing factorGi, differentiates between the various types of degradation. For example the weigh-ing factor for structural degradation is 3 whereas for patches is one. Table 2 summa-rizes the results indicating that the state of the inspected sections varied betweengood and sufficient. Three pavements were found to be in sufficient condition(VD2, VD5, AG2), four were in medium condition (VD3, VD4, VS6, VD10) and onepavement was new (AG3).

3. Investigation of the microstructure of porous asphalt usingpolarizing and fluorescent microscopy

The behavior of all materials depends highly on their micro-structure. Defects initially in form of micro cracks and voids atthe microstructure level can eventually lead to the failure of thematerial under load and improvements in the material can bemade by understanding the deterioration process and improvingthe microstructure.

Asphalt concrete is a non-homogeneous material which typi-cally consists of crystalline aggregate structure and amorphousbitumen structure and air voids making the various componentsof its microstructure easily recognizable with an optical micro-scope. Here, observations have been made using a polarizingmicroscope (Axioplan from Zeiss, Germany) which allows enlarg-ing from 12.5� up to 400�. The polarizing microscope differs froman ordinary microscope in that it has a revolving stage, a polarizingdevice below it (polarizer) and a similar device above called ananalyzer. They can also be referred to as upper and lower polars.Each polar transmits light waves vibrating in one direction only

VS6 AG2 AG3 VD10 twinlayer

PA11 PA11 PA11 PA8/PA22+

dditivePmB styrelf13/80

B 55/70S +trinidad NAF 501

PmB E 70/100 +calcium hydroxide

PmB styrelf13/80

– – – – 100– – – – 95100 100 100 – 6098 98 95 100 2667 60 65 95 2128 16 24 35 1617 13 16 22 1411 12 13 15 129 11 11 13 107 8 8 10 75.5 7 7 8 63.5 4.8 4.1 4.5 44.5 4.8 + 1.8 additive 5.2 5.0 4.3

20.3 26.1 19.3 24.1 14.8

5 5 0 4(1999) (1999) (2004) (2000)

Remarks

l Significant polished aggregates and loss-no structural degradationSignificant aggregate loss, minor rutting

nt Medium aggregate lossSignificant polished aggregates and loss

Significant aggregate lossNew pavement – not inspectedMinor aggregate loss minor rutting

L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484 3477

[10]. It is possible to observe thin specimen with the light micro-scope and thin and thick specimen with the fluorescent micro-scope. Fluorescent microscopy allows the observation of thevoids when they are impregnated with dyed epoxy resin as theincident light excites the fluorescence dye in the epoxy resin.

In this investigation two PA mix types AG1 and VD4 were inves-tigated. AG1 is the laboratory prepared specimen using the samemix design as AG2. Although not an ideal candidate for compari-sons with VD4 cores, due to lack of cores for AG2 and similarmechanical performance with AG2 [8], AG1 was used. The binderfor AG1 is straight run bitumen with Trinidad NAF501 natural as-phalt whereas, VD4 with polymer modified bitumen (Table 1).The selection of these two material types was due to their poorand well field performance in terms of raveling and mechanicalperformance, respectively [8]. For each material type four samplesof approximately 28 mm � 47 mm � 10 mm were cut using a dia-mond blade saw and running cold water. They were cut from thebottom center of 150 mm diameter cylindrical specimen. Afterdrying in a 50 �C oven for three days the samples were placed un-der 1 mbar vacuum for 2 h. The impregnation process makes thevoids detectable and helps keep the sample intact during the pol-ishing process. Not impregnating has led to crumbling problemsduring cutting and polishing [11]. The samples were impregnated

Fig. 1. Thin section of asphalt concrete (28 mm � 47 mm � 25 . . . 30 lm) demonstratinSample shown from mix type AG1 top and VD4 bottom. (For interpretation of the referearticle.)

with four parts epoxy (Araldit BY158) mixed with 1 part hardener(Aradur 21) and less than 0.5% fluorescent dye (EpoDye) under vac-uum. This dye fluoresces at approximately 440 nm in visible lightand in the ultraviolet region between 256 nm and 285 nm [11].Dead weight loading on the samples was used during curing.

After the impregnation process and cure, the excess epoxy andthin layer of asphalt from the impregnated side of the section wereremoved using a thin diamond blade saw. The impregnated sam-ples were sent to the mineralogy and petrographic institute ofthe University of Basel for thin sectioning. To prepare the thin sec-tions the samples were polished first with 600 grit silicon carbide.Thereafter they were further abraded and polished by hand with800 and 1500 silicon carbide abrasive paper. The polished surfaceof the sections was glued cold with a high-strength epoxy to apetrographic glass slide and ground to a ±5 lm level finish. Thethin section specimen size was 28mm � 47mm � 25 . . . 30 lmthickness.

3.1. Properties of the microstructure through qualitative observations

The microstructure of AG1 and VD4 can be seen in the thin sec-tions in Fig. 1. Both types of mixes display aggregates with desir-able features such as sharp edges. However, better coverage of

g the microstructure with voids filled with resin in yellow, aggregates and mastic.nces to colour in this figure legend, the reader is referred to the web version of this

3478 L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484

the aggregates with binder is easily distinguished in case of VD4compared to AG1. Some AG1 aggregates are partially or completelynot covered by the binder. The micrographs display that the distri-bution of the voids in the two types of mixes are quite different.General comparison of the AG1 material and VD4 material showsa better distribution of voids and mastic in the VD4 material. Sim-ilar to the findings of Shih et al. for a comparable material [12], ex-tra long and large voids along aggregate edges were seen for AG1.In addition, AG1 has larger clusters of voids which are probablywhat led to the raveling as indicated in the field inspection (Table2). Unlike the findings of Elseifi [13] air void formation in the mas-tic was observed at several locations as shown in the examples inFig. 1. VD4 samples are made from field cores and have experi-enced the effects of traffic in the form of cracks in the mineralaggregates as seen in several locations shown in Fig. 2. Fig. 1 alsoshows that there is more mastic and it is distributed more evenlyin VD4 than in AG1. VD4 which is the material exposed to the ele-ments and traffic shows areas where the mastic has been washedaway. Additionally, a more homogeneous interlocking of particlescan be seen in VD4 whereas in AG1 there are clusters ofinterlocking.

In summary, preliminary observations of the micro structure ofAG1 has shown indications of sub optimal performance in compar-ison with VD4. However more concrete criteria need to bedeveloped.

4. Mechanical testing

4.1. Water sensitivity investigation using the indirect tensile test (IDT)

The purpose of this European standard test [4] is to determinethe effect of saturation and accelerated water conditioning on theindirect tensile strength (ITS) of cylindrical specimen. The standardtest was modified as stated below. Six specimens (150 mm indiameter) were equally divided into a dry and a wet group. Thedry group was stored at room temperature (20 ± 5)�C while thewet group was vacuum saturated at an absolute pressure of300 mbar for 5 min then stored for 24 h at 60 �C. The specimenwere thereafter further conditioned for 24 h at the test tempera-ture of 25 �C. A deformation of 50 mm/min was applied to thespecimen until failure. The strength at failure ITS is noted. The indi-rect tensile strength ratio ITSR is calculated as follows:

Fig. 2. Fluorescent micrograph showing cracks in the aggregates. Sample shown from mreader is referred to the web version of this article.)

ITSR ¼ 100� ITSw

ITSd½%�

where ITSw is the average indirect tensile strength of the wet group[kPa] and ITSd is the average indirect tensile strength of the drygroup [kPa].

4.2. CAST

The coaxial shear test (CAST) is a cyclic, axial loading system todetermine the complex modulus (E*) of an asphalt pavement. Thetest was developed at Empa in 1987 [14] and further developedin various stages [15,7,16]. Tests are performed in a conventional,temperature controlled, servo-hydraulic tension–compression ma-chine (Fig. 3). The shear load is applied perpendicular to the spec-imen’s circular surface through the central core, with lateralconfinement provided by a metal ring surrounding the specimen.This format has shown to work well for PA [7,17,18] and here itsapplication for a twin lay PA (VD10) is demonstrated for the firsttime.

Virgili et al. [17] have used a continuum damage approach todetermine damage level of each specimen tested with CAST. Partlet al. [18] tested both dense graded and open graded mixtures atdifferent air void contents; the water sensitivity of both sets wasdetermined by comparison of fatigue performance of the dry andwet sets.

The setup allows loading along the same axis as that of trafficwhile the lateral confinement simulates a semi-infinite situationexperienced on the road. The basic setup was modified so thatthe specimen is immersed in water with at least 10 mm of waterabove the top surface of the specimen (Fig. 3). This allows to pro-duce a vertical water flow and pumping action of water similarto the field situation. The test method produces mechanical dam-age due to repeated loading, temperature cycles and water immer-sion avoiding pronounced thermal stresses as described in [6] andmodeled in [17]. In this case four temperature cycles from 25 �C to30 �C and 30 �C to 25 �C each for 18,000 s were used (Table 3 andFig. 4). The tests were conducted under deformation control witha sinusoidal deformation applied displacing the central core verti-cally. Complex modulus and phase angle of the material is calcu-lated using the finite element method as explained in detailelsewhere [16].

ix type VD4. (For interpretation of the references to colour in this figure legend, the

Fig. 3. Experimental setup of the coaxial shear test (CAST) above dry condition,below water submerged [17].

Table 3Test parameters and specimen dimensions.

Test parametersTempering period 7200 sFrequency 10 HzDeformation amplitude 0.01 mmTemperature program Repeated ramps between 25 �C and 30 �CNumber of cycles 4Duration of test 7200 + (18,000 � 2) � 4 = 151,200 s

Typical specimen dimensionHeight 60 mmOuter radius 75 mmInner radius 28.5 mmDiameter 150 mm

0

1000

2000

3000

4000

0 500000 1000000 1500000 2000000

Number of Loading Cycles

Com

plex

Mod

ulus

E* [

MP

a],

0

45

Pha

se A

ngle

[°],

Tem

pera

ture

[°C

]

Modulus

Temp

Phase

0

1000

2000

3000

4000

5000

0 500000 1000000 1500000 2000000Number of Loading Cycles

Com

plex

Mod

ulus

E* [

MP

a],

0

45

Pha

se A

ngle

[°],

Tem

pera

ture

[°C

]

Modulus

Temp

Phase

Cycle 4 Cycle 3 Cycle 2 Cycle 1

E*max E*min E*final

Fig. 4. Evolution of modulus, phase angle and temperature as a function of loadingcycle and four thermal cycles. Sample of specimen in dry condition above, sample ofspecimen in wet condition below. Both from the twin lay VD10.

L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484 3479

5. Results and discussion

5.1. Water sensitivity tests using ITSR

The results of the ITSR at 25 �C for the selected field cores areshown in Fig. 5. With the exception of AG3 the cores were all ina long term aged and trafficked state (Table 1). AG3 cores were ta-ken shortly after construction and therefore were short term aged.However, the water sensitivity criteria in the standard [5] are de-fined for an un-aged specimen. Still, after having aged and traf-ficked for various amounts of years all cores with the exceptionof AG2 and VD5 meet the criteria of ITSR P70% [4,5]. However,VD2 which also received a low surface inspection rating (Table 2)met the criteria.

5.2. Water sensitivity using water submerged fatigue CAST

For the CAST tests two specimens were tested for each condi-tioning state (dry and wet) for eight types of materials which re-sulted in 32 tests. A sample of the results shown in Fig. 4 showsthat at the beginning of the fatigue test during the first tempera-ture cycle, the modulus drops considerably. This is consistent withprevious research [6,7,17,18] and the reasons for it are twofold.Firstly, due to an initial settling effect in the material the internalstresses redistribute and secondly, due to the initial thermal statein the material in the first temperature cycle which is significantlydifferent from the later cycles [6,7]. The thermal history can clarifythis phenomenon as before the beginning of the test the specimenare kept for at least two hours at 25 �C to ensure thermal equilib-rium which is not the case at the beginning of later cycles in spiteof the comparatively slow temperature cycles. The temperature re-corded is the chamber temperature and not the specimen temper-ature. Since the test protocol is the same for all specimens themaximum and minimum modulus values could be compared.The maximum modulus (Fig. 4), E�max, (also labeled as E�initial for fur-ther evaluations) occurs at the end of the first thermal cycle andthe minimum, E�min occurs at the middle of the last thermal cycleand E�final occurs at the end of the last thermal cycle. Fig. 4 showsthe typical development of the modulus, phase angle and temper-ature as a function of loading cycles or time and temperature cy-cles. A comparison of the wet state and the dry state shows thatfor all materials with the exception of AG3 the modulus decreasesmore rapidly in wet state (Table 4 and Fig. 10). Among the samplestested AG2 followed by VD2 and VD5 showed the most pro-nounced reduction in modulus and AG3 and the twin lay VD10the least. AG2, VD2 and VD5 also had the worst surface deteriora-tion characteristics based on the surface inspections (Table 2).

0

20

40

60

80

100

120

AG2

AG3

VD2

VD3

VD4

VD5

VS6

VD10

Material Designation

Cas

t Rat

io o

r ITS

R [%

]ITSRCAST ratio maxCAST ratio minCAST ratio final

Fig. 5. Comparison of the indirect tensile strength ratio (ITSR) at 25 �C and CAST ratio at 25–30 �C.

Table 4Complex modulus reduction.

Material des. Specimen Effect of wetting Effect of fatigue loading

Modulus (MPa) Reduction (%) Modulus (MPa) Reduction (%)

E�max E�min max min E�initial E�final dry wet

AG2-dry 121 2824 1945 77 80 2824 2414 15 42AG2-wet 111 649.6 389.2 649.6 378.9AG3-dry 217 653.3 375.2 5 �6 653.3 572 12 2AG3-wet 221 623.7 397.4 623.7 609.5VD2-dry 315 4012 2911 59 66 4012 3633 9 25VD2-wet 312 1627 983.9 1627 1216VD3-dry 414 2388 1666 53 54 2388 2119 11 21VD3-wet 421 1117 766.9 1117 883VD4-dry 516 2140 1418 21 33 2140 1813 15 32VD4-wet 514 1697 956.9 1697 1161VD5-dry 615 3361 1942 50 57 3361 3087 8 27VD5-wet 613 1693 840.8 1693 1231VS6-dry 714 2320 1386 40 43 2320 2000 14 21VS6-wet 713 1388 783.4 1388 1090VD10-dry 1113 2421 1533 3 �4 2421 2100 13 8VD10-wet 1114 2354 1596 2354 2155

Average 1954 1243.2 38 40 1928 1620 12 22

Fig. 6. Black diagram for all downward temperature ramps of AG2 dry specimen under repeated loading left, and wet specimen under repeated loading right for four thermalcycles.

3480 L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484

Fig. 7. Black diagram for all downward temperature ramps of AG3 dry specimen under repeated loading left, and wet specimen under repeated loading right for four thermalcycles.

Fig. 8. Evolution of the complex modulus as a function of loading cycle, dry state.

Fig. 9. Evolution of the complex modulus as a function of loading cycle, wet state.

L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484 3481

0

1000

2000

3000

4000

5000

AG2

AG3

VD2

VD3

VD4

VD5

VS6

VD10

Com

plex

Mod

ulus

[MPa

]

Initial-Dry

Final-Dry

Initial-wet

Final-wet

Fig. 10. Loss of modulus due to fatigue loading demonstrated using the modulus at the end of first thermal cycle and last thermal cycle in dry and wet state.

0

500

1000

1500

2000

2500

3000

3500

4000

0 500000 1000000 1500000 2000000

No Loading Cycles

Co

mp

lex

Mo

du

lus

[MP

a]

Fig. 11. Complex modulus obtained for two specimen of VS6.

Fig. 12. Development of complex modulus as a function of temperature cycles.Example shown for material type VS6 in wet state.

3482 L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484

Observing the behavior at each cycle for example for AG2 andAG3 using black diagrams of the downward ramp of the tempera-ture cycles as in [6], it can be seen that the wet AG2 specimen isalready displaying a reduction in modulus after the first thermalcycle in comparison to the dry specimen (Fig. 6). However thenew pavement AG3 shows very little sensitivity to water (Fig. 7)as was also seen in the ITSR (Fig. 5). As seen in Fig. 6, in the caseof AG2 a significant reduction of modulus can be observed from cy-cle to cycle. This phenomenon does not happen in AG3 leading tothe conclusion that the modulus of AG3 is not dependent on waterexposure. Another aspect is the flattening of the slope in case ofAG2 indicating less sensitivity to temperature in wet state andthe increasing of the slope in case of AG3 in wet state indicatingmore sensitivity to temperature. Further comparison with the ITSRcan be achieved by introducing the CAST ratio as follows.

CAST ratio max = 100 � ((E�max)dry � (E�max)wet)/(E�max)dry

CAST ratio min = 100 � ((E�min)dry � (E�min)wet)/(E�min)dry

CAST ratio final = 100 � ((E�initial)dry � (E�final)wet)/(E�initial)dry

The results shown in Fig. 5 indicate that the CAST ratio calcu-lated in three ways delivers similar results. With the exception ofAG3 and VD4 the CAST ratio shows AG2, VD2, VD3 and VS6 consid-erably more water sensitive than ITSR and VD10 less water sensi-tive than the ITSR.

It is interesting to note the fatigue performance of the doublelayer porous asphalt or twin lay (VD10). Although this section

has improved performance in terms of reduced clogging, in termsof mechanical performance it is in par with the other well perform-ing sections.

Of particular interest here is the reduction in the modulus dueto fatigue loading. Two methods were used to analyze the data.Firstly, the dry–wet reduction in modulus was assessed and sec-ondly the evolution of wet-fatigue behavior and dry fatigue behav-

Fig. 13. Complex modulus vs. temperature in wet condition example shown for up ramp at the last temperature cycle.

L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484 3483

ior was assessed (Table 4). The evolution of complex modulus forall the materials in dry state and wet state is shown in Fig. 8 andFig. 9. It can be seen that the oldest core VD2 retained the highestmodulus whereas AG3 had the lowest modulus in dry state. How-ever when modulus loss under fatigue loading is assessed AG3ranks the best (Table 4). It is also noteworthy that the performanceof the twin lay VD10 in dry state was very similar to the majority ofthe materials. The surface inspection of VD10 also indicates goodperformance (Table 2). In wet state the twin lay had the highestmodulus with the AG2 ranking the worst. However most of thesections remained in a band defined by VS6 and VD2.

The comparison of initial and final modulus shown in Table 4 al-lows comparison of the material at the same level of thermal con-ditioning. In dry condition the loss of modulus ranged from 8% to15% whereas in wet state this was significantly larger and from2% to 42%. Although AG3 has the lowest modulus but it is at thesame time the least water sensitive which corroborates the resultsfrom ITSR. In situ inspection ranked AG2, VD5 and VD2 as the mostdeteriorated whereas ITSR ranked AG2 and VD5 as the most watersensitive. The fatigue tests dry–wet ranked AG2 with 80% and VD2with 66% and VD5 with 57% modulus reduction as the materialsmost susceptible to water damage (Table 4).

An example of the development of the complex modulus withrespect to temperature is shown in Figs. 12 and 13. The reductionof modulus with each thermal cycle can be seen in the example inFig. 12 for VS6 in wet state. The comparison of all materials duringthe up ramp of the last thermal cycle (starting at 1.28–1.44 millioncycles) where the material has experienced the most mechanicaldamage can be seen in Fig. 13. VD10 was at this last cycle thematerial with the highest modulus but also with greater tempera-ture dependency. AG3 and AG2 were the materials with the lowestmodulus with the remainder of the materials staying within theband defined by VD2 and VD3.

This is a relatively new experimental technique and a vast num-ber of statistically significant data is not as yet available to analyzethe reliability and repeatability of the data. However it should benoted as stated earlier that during this project two specimen weretested for each conditioning state (dry and wet) for eight types ofmaterials which results in 32 tests. Good repeatability was seenbetween the two specimens on most of the data as shown in theexample in Fig. 11. Among the 32 tests two were deemed unusableas the results varied significantly with the second set of the samematerial in same condition and with the other materials. It is rec-

ommended that at least three samples should be tested for eachmaterial and condition.

6. Conclusions

Fatigue tests using the coaxial shear test, CAST, was used to as-sess water sensitivity of field cores of conventional porous asphaltas well as twin lay. These results were compared with field assess-ment and the more empirical ITSR tests. Examination of fatiguebehavior of eight PA mix types from the field showed that fatiguebehavior of dry specimen can differ significantly from water sub-merged specimen. The tests showed for all but two material typesa pronounced reduction in the modulus in wet condition. This de-crease was increased with each temperature cycle. The only mixtype not made with polymer modified bitumen (AG2) was one ofthe most water sensitive. Using fatigue data more insight wasgained into the material performance exemplified by the develop-ment of complex modulus under cyclic load and in wet and drycondition. The fatigue tests showed the twin lay to be performingwell under wet conditions. Such insight could not be gained by theconventional ITSR and can aid in appropriate material selection.CAST test results corroborated field surface degradation results indistinguishing degraded pavements where it was not possible withITSR.

Preliminary observations of the micro structure were promisingand have shown indications of sub optimal performance for mixesthat did perform suboptimal in laboratory tests and field inspec-tions. However more concrete criteria need to be developed.

Acknowledgements

The authors would like to acknowledge the financial support ofthe Swiss Federal Roads Office (FEDRO, ASTRA), the Swiss NationalScience Foundation and cooperation of the local Cantonal authori-ties and EPFL for their cooperation and in providing vital data aswell as specimen.

References

[1] Mallick R, Gould J, Bhattacharjee S, Regimand A, James L, Brown E. In:Development of a rational procedure for evaluation of moisture susceptibilityof asphalt TRB annual meeting; 2003.

3484 L.D. Poulikakos, M.N. Partl / Construction and Building Materials 23 (2009) 3475–3484

[2] Cheng D, Little D, Lytton R, Holste J. Use of surface free energy properties of theasphalt aggregate system tzo predict damage potential. J Assoc Asphalt PaveTechnol AAPT 2002.

[3] Birgisson B, Roque R, Page G. In: Evaluation of water damage using hot mixasphalt fracture mechanics AAPT proceedings; 2003.

[4] EN 12697-12, European standard, bituminous mixtures – test method for hotmix asphalt – Part 12: determination of the water sensitivity of bituminousspecimens, December 2003.

[5] SN 640 431-7NA, Asphalt – offenporiger asphalt (porous asphalt) 2003 [inGerman and French].

[6] Gubler R, Baida LG, Partl MM. A new method to determine the influence ofwater on mechanical properties. Int J Road Mater Pave Des 2004:259–79.

[7] Gubler R, Partl MN, Canestrari F, Grilli A. Influence of water and temperatureon mechanical properties of selected asphalt pavements. Mater Struct2005;38.

[8] Poulikakos L, Pittet M, Junod A, Arnaud L, Simond E, Gubler R, Partl MN,Dumont AG. Mechanical properties of porous asphalt, recommendations forstandardization; 2006.

[9] SN 640 925b, 2003 Swiss Standards, Erhaltungsmanagement der Fahrbahnen(EMF), Anleitung zur visuellen Zustanderhebung und Indexbewertung mitdem Schadenkatalog. August 2003. Guidelines for surface inspection [Originalin German and French].

[10] Wahlstrom, Ernest Eugene. Optical crystallography. 5th ed. New Yorka.o.: Wiley & Sons; 1979.

[11] Scott EB. Polymer/fiber modified asphalt fracture mechanisms andmicrostructure relationships to distresses and environmental factors. PhDdissertation Michigan State University; 1993.

[12] Shin EE, Bhurke A, Scott E, Rozeveld S, Drzal LT. Microstructure, morphologyand failure modes of polymer-modified asphalts, Transportation ResearchRecord (1535), p. 61–73.

[13] Elseifi MA, Al-Qadi IL, Yang S-H, Carpenter S. Validity of asphalt binder filmthickness concept in hot-mix asphalt. In: Transportation Research Board, 87thannual meeting. Washington, DC. USA; 2008.

[14] Junker JP. Entwicklungen zur Bestimmung mechanischer Materialkennwertean bituminösen Baustoffen, insbesondere an Asphalt, Empa report no. 215;1987.

[15] Younger KD, Part1 MN, Fritz HW, Gubler R. Asphalt concrete shear testing withthe co-axial shear tester at Empa. In: DiBenedetto, Franken, editors.Mechanical tests for bituminous materials. RILEM; 1997.

[16] Sokolov K, Gubler R, Partl MN. Extended numerical modeling and applicationof the coaxial shear test for asphalt pavements. J. Mater Struct2005;279(June):515–22.

[17] Virgili A, Partl MN, Grilli A, Santagata F. Damage model for environmentalconditioned fatigue test with CAST. Fatigue Fract Eng Mater Struct2008;31:967–79.

[18] Partl MN, Canestrari F, Grilli A, Gubler R. Characterization of water sensitivityof asphalt mixtures with coaxial shear test. Int J Road Mater Pave Des RMPD2008;9:247–78.