Determining the impact of degree of blending and quality of reclaimed asphalt pavement on predicted...

Construction and Building Materials 48 (2013) 473–478

Contents lists available at ScienceDirect

Construction and Building Materials

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

Determining the impact of degree of blending and quality of reclaimedasphalt pavement on predicted pavement performance using pavementME design

⇑ Corresponding author. Tel.: +1 856 2565327.E-mail address: mehta@rowan.edu (Y. Mehta).

Sean Coffey a, Eric DuBois a, Yusuf Mehta b,⇑, Aaron Nolan c, Caitlin Purdy c

a Graduate Research Assistant Rowan University, United Statesb Associate Professor, Department of Civil and Environmental Engineering, Rowan University, Glassboro, NJ 08028, United Statesc Research Assistant Rowan University, Glassboro, NJ 08028, United States

h i g h l i g h t s

� Predicted rutting is not significantly impacted by the assumed degree of blending and quality of RAP.� Predicted fatigue cracking performance was significantly different between mixes with different qualities of RAP.� As the quality of RAP becomes better, the difference in predicted performance between the full and actual blending conditions is lower.

a r t i c l e i n f o

Article history:Received 1 March 2013Received in revised form 5 June 2013Accepted 6 June 2013Available online 8 August 2013

Keywords:RAPRuttingFatigueDCMPavement ME

a b s t r a c t

Past studies indicated binder from reclaimed asphalt pavement (RAP) aggregates only partial degree ofblending (DOB) within hot mix asphalt (HMA). Most state agencies assume full blending, which may leadto under asphalting. This study focuses on determining the impact of it on predicted performance usingPavement ME level I analysis of six 25% RAP mixes. Dynamic complex modulus (DCM) tests were con-ducted with full and ‘‘actual’’ blending, all greater than 85%, using varying conditions comparing ruttingand fatigue performance. The results indicate high DOB RAP mixes with uniform gradations have negli-gible effect on predicted performance and would not compromise performance.

1. Introduction

The asphalt industry, which is under pressure to cut cost andproduce a more environmentally sustainable product, is increas-ingly incorporating RAP into HMA. With more mixes includingRAP and at higher percentages it has become imperative to under-stand the impact of RAP on performance, yet little is known. RAPquality as well as how well the RAP blends with the new mixmay be impacting performance. Many state agencies assume fullblending, which may lead to under asphalting or a relatively stiffermix. However, it is unclear how this will affect the predicted pave-ment performance. While RAP quality can affect the degree ofblending, RAP quality can impact performance and can be classi-fied separately. There is a need to conduct a level I analysis todetermine the impact of DOB and RAP quality on predicted pave-ment performance.

2. Objective

The objective of this study is to determine the impact of RAPquality and degree of blending, in a 25% RAP hot mix asphalt, onpredicted rutting and fatigue performance using Pavement MElevel I analysis. The research methodology to achieve the objectiveis shown in Fig. 1 below for each RAP source. Each sample will belabeled by a mix label as follows: Mix-X-X (as shown in Table 1).

3. Background information

3.1. Degree of blending

Degree of blending represents the percentage of the total RAPbinder that is mobilized in an asphalt mix. Degree of blending isdependent on binder content and properties, as well as gradation;therefore, samples were prepared under the ‘‘Actual’’ and fullblending conditions to assess the impact of the DOB [1]. Many state

Fig. 1. Research methodology for each of the three RAP sources.

Table 1Mix and analysis designation.

RAP source Degree of blending Mix ID

1 Full Mix-1-FActual Mix-1-A

474 S. Coffey et al. / Construction and Building Materials 48 (2013) 473–478

agencies assume the full blending condition, where all of the RAPbinder mobilizes and contributes to the total binder of the mix.However, in actuality, the DOB falls somewhere between 0% and100%, depending on the RAP and mix properties, and is referredto as the ‘‘actual’’ DOB [2]. DOB is significant as it has an impacton the binder content and the stiffness of the final mix. If an incor-rect DOB is assumed, it could result in under asphalting or overasphalting. Under asphalting can occur if the DOB is expected tobe higher than it actually is, leading to a lower then optimum bin-der content and usually a stiffer mix; while over asphalting is theleads to higher then optimum binder content. Under-asphaltingcan lead to fatigue and thermal cracking, while over asphaltingcan lead to rutting [3].

3.2. Dynamic complex modulus (DCM)

The new AASHTO Mechanistic–Empirical (M–E) Design Guideuses the dynamic complex modulus as the primary test protocolto characterize the modulus response of hot mix asphalt. Dynamiccomplex modulus or E* is the ratio of stress to strain under dy-namic conditions, refer to Eq. (1).

jE�j ¼ re

where r is the amplitude of stress, e is the amplitude of strain.The test was conducted at three temperatures 4, 20, and 40 �C,

as well as multiple frequencies ranging from 0.1 to 10 Hz. Subse-quently, a master curve was developed using the procedure inAASHTO PP-62 [4] developed to extrapolate more data points.

4. Methods and materials

The product of milling an old road, reclaimed asphalt pavement (RAP) is com-posed of aggregates, aged binder known as RAP binder and anything else that hasworked its way into the pavement over its lifetime. Unlike virgin aggregates whichusually come from a single source with little variability, RAP aggregates usuallycome from multiple sources with plants amassing large stock piles that range invariability.

RAP was obtained from three separate asphalt plants to assess the impact ofRAP quality. The RAP was collected by sampling from five different locations withineach of the plant’s RAP stockpiles, which ensured a more representative sample ofthe entire stockpile. All virgin aggregates were from a single source, and a virginbinder of Performance Grade (PG) 70-28 was used.

To assess the impact of the DOB of each RAP source the mixes were prepared tothe full and ‘‘actual’’ blending conditions. The full blending condition was used,since most state agencies assume this condition; therefore, this condition repre-sents what is done in practice. An earlier study conducted by the authors [1]showed that the ‘‘actual’’ DOB condition depends on the gradation of RAP aggre-gates, the RAP binder content, and the difference in the stiffness of the virgin binderand the RAP binder. Therefore, the different characteristics of each RAP supplydetermine the DOB condition that occurs. The degree of blending was experimen-tally determined using the following procedure:

Step 1: Prepare hot mix asphalt with ignited RAP aggregates composing 25% ofthe mix to determine the necessary binder content to achieve a 4% air voidSuperpave mix at the design gyrations of 75.Step 2: Prepare the mix with RAP, working under the assumption of 70% blend-ing. The 70% was selected based on a study conducted by Shirodkar et al. [5]. Ifthe volumetrics do not meet the 4% air voids at the design gyrations, the virginbinder is adjusted until the mix achieves 4% air voids. If no adjustment is nec-essary then the assumed DOB is correct. However, if the binder content needs tobe adjusted, the difference in the effective binder content between the mixesprepared in Step 1 and Step 2 determines how much the actual DOB differs fromthe assumed DOB of 70%.Step 3: The actual DOB is calculated based on the equation shown below:

DOB ¼ 100

�Weight of binder expected with 70% DOBþ weight of mix� difference of Pbe;est

� �

Total weight of binder on the RAPð2Þ

The detailed analysis to determine DOB is beyond the scope of this paper. Oncethe DOB is determined the DCM is determined by preparing samples by followingAASHTO T-342 [6]. After the samples are prepared they are aged according to AASH-TO R30 [7]. They are then tested by following another part of AASHTO T-342 [6].Once the data is compiled the master curves are developed following AASHTOPP-62 [4].

Fig. 2. RAP source gradations.

S. Coffey et al. / Construction and Building Materials 48 (2013) 473–478 475

5. Results

5.1. RAP source properties

The gradation of the entire mix was kept the same to prevent thegradation from becoming a variable between mixes, such that thesignificance of the gradation of each RAP source could be analyzed.A representative sample from the stockpile was collected. Fifteensamples of RAP from the stockpile were also collected to determinethe variability of the RAP sources. RAP sources 1 and 2 had a maxcoefficient of variation of 2.25% on the ½’’ sieve and a max coefficientof variation of 1.11% on the #8 sieve. RAP source 3 was highly vari-able, with half of the sieves having standard deviations over 3%.RAP sources 2 and 3 had mostly fines in the mixes while RAP source1 had a well blended mix due to fractionation. The RAP source rep-resentative gradations can be seen below in Fig. 2.

RAP binder from sources 1, 2, and 3, all had a high PG grade of82 �C. The RAP sources 1, 2, and 3 were mixed with virgin binderPG 70-28, and their DOB were 89.6%, 90.0%, and 84.7%, respec-tively. A recent study by authors [1] regarding DOB from fiveRAP sources in the state of New Jersey showed that four of the fivehad DOB of around 85% and greater. Therefore, the DOB valuesabove are typical of what is observed in the state.

5.2. Mixture properties

All attempts were made of maintaining a similar gradation foreach mix. Fig. 3 shows gradation for each of the mixes. The grada-tions are shown against the Maximum Density Line (MDL) and theother lines connect the control points specified in NJDOT construc-tion specifications.

Master curveThe following equation can model the master curve to allow

data to be extrapolated beyond the testing range.

LogðE�Þ ¼ dþ a=ð1þ ebþcðlog red½freq�ÞÞ ð3Þ

The terms d and d + a represent Log (E*)’s range and the terms band c shape the curve to fit the given situation. Fig. 4 shows themaster curves of all the mixes for actual and full DOB condition.

5.2.1. Mix 1For lower temperatures or high frequencies, the actual DOB had

higher E* values than in the full blending condition. At roughly1.5 Hz, the master curve of actual DOB starts to be stiffer than that

of full blending. As the frequencies go up, Mix-1-A is 18% stifferthan Mix-1-F. At the higher temperatures of the master curve theMix-1-F is stiffer by 2%. Mix-1-A is stiffer at intermediate to lowertemperatures.

5.2.2. Mix 2At high frequencies, Mix-2-F is only 6% stiffer than Mix-2-A. At

the intermediate frequencies Mix-2-A is stiffer, though Mix-2-F be-comes stiffer again for the high temperatures.

5.2.3. Mix 3The master curve for Mix-3-F is consistently above the master

curve for Mix-3-A through the entire curve. The maximum differ-ence between the stiffness values was at lower frequencies whereMix-3-A is about 20% lower than Mix-3-F.

6. Pavement ME design

Pavement ME software evaluates the major flexible pavementdistresses, permanent deformation (rutting), and fatigue cracking(alligator and longitudinal cracking). The software uses traffic data,climatic data, and the structure of the pavement and asphalt layerproperties to predict performance [8]. For the asphalt layer proper-ties data, Pavement ME has three levels of inputs with level 3 usingdefault values for Performance Grades, level 2 using some binderproperties and level 1 using dynamic complex modulus test resultsand binder information [8].

As mentioned earlier, Pavement ME will be used to predictpavement performance using measured dynamic complex modu-lus values to evaluate actual and 100% degree of blending. Thereare three design inputs in this program: traffic data, climate data,and structure design. The traffic data used for the Pavement MEanalysis was as follows:

� An initial two way AADTT of 1700, 5000, and 12000, three levelsof traffic, operational speed is set at 60 mph.� 2 Design lanes, 50% of trucks in the design direction and 25% of

the trucks in the design lane.� A range of three climates will be used: Trenton, NJ, Bismarck,

ND, and Los Angeles, CA.

The two structures selected for analysis were determined fromthe four LTPP sections in New Jersey. Structure 1 consists of fourlayers.

Fig. 3. Gradation for the three mixes versus MDL and NJDOT control points.

Fig. 4. Master curves for all of the mixes at reference temperature of 20 �C.

Table 2Binder blending conditions for binder testing.

RAP source DOB condition RAP binder (%) Virgin binder (%)

1 Actual (89.6) 20 80100% 22 78

2 Actual (90.0) 23 77100% 26 74

3 Actual (84.7) 19 81100% 21 79

Layer 1: A 3 inch thick 25% RAP HMA layer. Level 1 analysis willbe conducted.Layer 2: A 6 inch asphalt concrete layer. The asphalt concretelayer uses a PG 76-22 binder with a basic gradation with:3=4’’sieve with 0%. Cumulative Percent Retained (CPR), 3/8’’ sievewith 17% CPR, #4 sieve with 36% CPR, and the #200 sieve with8% CPP.Layer 3: A 7 inch crushed gravel layer, (default modulus values).Layer 4: Semi-infinite A-1-a material subgrade layer (defaultmodulus values).

Structure 2 consist of 3 layers adding up to the same total thick-ness as Structure with Layer 1 of 9 inches and layer 3 and layer 4same as above.

Table 3Binder testing results.

Mix-DOB Mix-1-A Mix-1-F

Temperature (�C) G* Phase angle G* Phase angle

70 5.84 69.9 5.59 70.476 3.07 72.4 2.89 72.782 1.63 75.2 1.56 75.6

70 5.54 71.0 7.04 69.976 2.90 73.5 3.63 72.282 1.54 76.3 1.98 74.8

70 5.75 70.3 6.03 70.276 3.02 72.7 3.12 72.682 1.60 75.5 1.69 75.3

6.1. Property of the binder within the HMA prepared in the laboratory

Pavement ME requires G⁄ and phase angle binder properties fortesting at levels 1 or 2 analysis. To determine the properties of thebinder in the hot mix asphalt with actual DOB, the binder couldnot be extracted and recovered from the mix itself because theextraction and recovery process would have blended all the RAP bin-der with the virgin binder. Therefore, instead the RAP binder wasextracted and recovered separately then blended with the virginbinder according to their respective Degrees of blending usingAASHTO T319 [9]. Table 2 is the blending ratios necessary to createthe binder for each mix. Table 3 contains the G⁄ and phase angle dataat three temperatures required in Pavement ME for the binders thatwere blended in the laboratory according to their respective DOB.This was done to get a representation of the properties of the bindersfor the level 1 analysis in Pavement ME.

6.2. Pavement ME predicted performance

6.2.1. Discussion6.2.1.1. Fatigue damage. The mix type had a significant effect on thedifference in the fatigue performance prediction between actualand full blending condition. For Mix 1, full blending conditioncaused higher fatigue cracking than actual blending condition.However, for Mix 2 and Mix 3, the reverse trend was observed.The difference in the fatigue performance prediction between ac-tual and full blending condition for all three mixes was around70% more for structure 2 than that of structure 1. This was becausestructure 2 had a thicker surface layer. The increase in traffic fromlow level (AADTT = 1700) to moderate (AADTT = 5000) increasedthe difference in performance prediction by approximately 300%.This appears to be consistent with the proportion of increase intraffic between the two levels. A similar proportional increase indifference in performance prediction between actual and fullblending condition of 2500% was observed between moderateand high traffic level (AADTT = 12000). The difference in the fatigueperformance prediction between actual and full blending conditionfor New Jersey climate was 65% less than that of Los Angeles cli-mate and 47% more compared to Bismarck climatic region.

Table 4Predicted fatigue cracking at 20 years for structures 1 and 2.

Fatigue Damage at 20 years, %

Structure Climate TrafficLevel

Mix-1-A

Mix-1-F

Differencebetweenactual DOBand Full DOB

Averagebasedonclimate

Mix-2-A

Mix-2-F

Mix-3-A

Mix-3-F

1 Trenton,NJ

Low 6.13 6.33 �0.20 �0.73 5.23 3.99 1.24 4.58 5.33 4.23 1.10 4.07Moderate 18.00 18.60 �0.60 15.40 11.70 3.70 15.70 12.40 3.30Heavy 43.30 44.70 �1.40 36.90 28.10 8.80 37.60 29.80 7.80

LosAngeles,CA

Bismark,ND

Averagebased ontrafficlevel

Low �0.25 1.42 1.22Moderate �0.73 4.20 3.60Heavy �1.73 10.00 8.60

Average for each mix �0.90 5.21 4.47

2 Trenton,NJ

LosAngeles,CA

Bismark,ND

Table 5Predicted rutting at 20 years for structures 1 and 2.

Rutting at 20 years, inches

Structure Climate TrafficLevel

Mix-1-A

Mix-1-F

Mix-2-A

Mix-2-F

Mix-3-A

Mix-3-F

1 Trenton,NJ

LosAngeles,CA

Bismark,ND

Low 0.15 0.15 0.00 �0.01 0.13 0.09 0.04 0.07 0.13 0.10 0.03 0.05Moderate 0.24 0.25 �0.01 0.22 0.16 0.07 0.22 0.17 0.05Heavy 0.37 0.38 �0.01 0.34 0.24 0.10 0.34 0.27 0.07

2 Trenton,NJ

LosAngeles,CA

Bismark,ND

S. Coffey et al. / Construction and Building Materials 48 (2013) 473–478 477

6.2.1.2. Rutting. The difference in the rutting performance predic-tion between actual and full blending condition was insensitiveto structure and traffic level. However, the difference in predictionwas more sensitive to climate, with Los Angeles climate indicated1.5 to 2 times more difference in rutting prediction compared toNew Jersey and Bismarck. In addition, the mix type appeared tobe sensitive to difference in performance prediction between fulland actual blending.

6.2.1.3. Comparison between mixes. Using Table 4 to compare thepredicted performance of each mix, Mix-2 performed the best withthe least fatigue cracking and rutting, followed by Mix-3 and Mix-1performing the worst. The three mixes can be ranked in order ofquality with Mix-1 of the highest quality, as it had the tightest gra-dation amongst the three and was fractionated by the plant, fol-lowed by Mix-2, which had the next tightest gradation but wasnot fractionated, and Mix-3, with the most variable gradationand the plant did not fractionate. No trend developed between per-formance and DOB, possible because the DOB of each mix weresimilar. Quality of RAP as measured by sieve analysis and planthandling, either the plant fractionated or did not fractionate itsRAP, once again did not show a trend among mixes. However,it was noticed that the difference in performance between fulland actual blending conditions was less with the higher qualityof RAP, following the quality rank mentioned previously (seeTable 5).

7. Summary of findings

The findings based on the analysis at the 20 year mark are asfollows.

� RAP source 1, 2, and 3 were mixed with virgin binder PG 70–28,and the DOB was 89.6%, 90.0%, and 84.7%, respectively.� RAP source 1 fractionated and had an overall well blended mix

compared to sources 2 and 3 which had much finer gradations.Ranking Mix-1 of the highest quality among the 3 mixes, fol-lowed by Mix-2 and the Mix-3.� Mix-1 had actual blending outperformed the full blending for

both fatigue cracking and rutting.� Mix-2 had the highest difference for fatigue cracking and rut-

ting but the full blending performed better.� Mix-3 followed a similar pattern to Mix-2 but had a small

decrease in difference between the two blending assumptions.� At larger thickness of surface layer, heavier traffic, and warmer

climate the difference in the fatigue performance predictionbetween actual and full blending condition increased.� Rutting difference was minimal for all the mixes and is overall

negligible.� The difference between blending conditions was minimized as

quality of RAP increased, in regards to gradation uniformity.

� DOB and Quality of RAP showed no significant trends betweenmixes.

8. Conclusions

The conclusions based on the summary of findings above are:

(1) The predicted rutting is not significantly impacted by theassumed degree of blending and quality of RAP

(2) The predicted fatigue cracking performance was signifi-cantly different between mixes with different qualities ofRAP.

(3) As the quality of RAP becomes better, the difference in per-formance between the full and actual blending conditions islower.

9. Recommendation

If the degree of blending is about 85% or higher, full blendingcould be assumed without compromising performance. This isespecially true as the RAP gradations are more uniform and theRAP is fractioned. Classifying RAP by gradation uniformity maybe a useful tool to determine if the DOB of a mix should be calcu-lated; however, other parameters should be considered to deter-mine the direct impact quality of RAP has on mix performance.Assumption of appropriate DOB could be critical for fatigue perfor-mance especially for heavier traffic and is the surface layer consist-ing of high RAP is thick.

References

[1] Coffey S. Mehta Y. Kehr D. Degree of blending and variability of recycled asphaltpavement. In: Proceedings of the conference for journal of solid wastetechnology and management. Philadelphia (PA); 2012.

[2] Mehta Y, Nolan A. High reclaimed asphalt pavement in hot mix asphalt.Technical Draft Report 2011.

[3] Al Qadi IL, Carpenter S, Roberts G, Ozer H, Aurangzeb Q. Investigation ofworking binder in hot-mix asphalt containing recycled asphalt pavements. TRB2009. 09–1262.

[4] AASHTO PP-62, 2010. Standard practice for developing dynamic modulusmaster curves for hot mix asphalt (HMA). Washington (DC): AASHTO; 2005.<www.transportation.org>.

[5] Shirodkar P, Sonpal K, Mehta Y, Nolan A, Norton A, Tomlinson C, et al. A study todetermine the degree of partial blending of reclaimed asphalt pavement (RAP)binder for high RAP hot mix asphalt. Washington DC: Proceedings of theTransportation Research Board, National Research Council; 2010.

[6] AASHTO T-342. Standard method of test for determining dynamic modulus ofhot-mix asphalt concrete mixtures. Washington (DC): AASHTO; 2011.<www.transportation.org>.

[7] AASHTO R 30. Standard practice for mixture conditioning of hot mix asphalt.Washington (DC): AASHTO; <www.transportation.org>.

[8] Mechanistic empirical design of new and rehabilitated pavement structuresguide. 403. NCHRP I-37 A. Web, 2012. <http://onlinepubs.trb.org/onlinepubs/archive/mepdg/home.htm>.

[9] AASHTO T319, 2008. Standard test method for the quantitative extraction andrecovery of asphalt binder from hot mix asphalt. Washington (DC) : AASHTO;2008. <www.transportation.org>.

Determining the impact of degree of blending and quality of reclaimed asphalt pavement on predicted...

Documents

Mechanistic-Empirical Pavement Design Procedure For

Blending Magnetic Resonance Imaging and Numerical Simulation

Dielectric Properties of Asphalt Pavement

Pavement Drainage and Pavement-shoulder Joint Evaluation \u0026 Rehabilitation

Asphalt Pavement Quality Indicator

PAVEMENT STRUCTURAL ANALYSIS, Lecture 2

Blending State and Federal Administrative Law - CORE

Optimal Multi-Time Period Gasoline Blending - MacSphere

Pavement Life Cycle Assessment Framework - Federal

Blending polygonal shapes with different topologies

Cognitive Poetics, Conceptual Metaphors and Blending in

Presentation on flexible pavement distress

Model Blending for Text Classification - arXiv

An Algorithm for image stitching and blending

ARLINGTON COUNTY PAVEMENT MARKING

Pavement Ruting problem in Viet nam

Recycling reclaimed road material in hydraulically bound layers

District Level Microcomputer Flexible Pavement Highway

Precast Prestressed Concrete Pavement ( PPCP )

Airport Pavement Health Robotic Inspection System