Effect of Aggregate Gradation on Measured Asphalt Content

Transcription

1 TRANSPORTATON RESEARCH RECORD Effet of Aggregate Gradation on Measured Asphalt Content PRTHV S. l<andhal AND STEPHEN A. CROSS t is neessary to losely ontrol the asphalt ontent in hot-mix asp?lt (HMA) mixes to obtain optimum servieability and dura_bihty. However, oarser mixes (binder and base ourses) made with larger maximum partile-sized aggregate tend to segregate. The resulting variation in the aggregate gradation of the sampled HMA mix an signifiantly affet the measured asphalt ontent. A study was done to evaluate the effet of aggregate gradation on the measured asphalt ontent. Atual mix omposition (asphat ontent and _gradation) data from a major nterstate paving projet were obtamed and analyzed. A total of 547 binder ourse and 147 wearing ourse mix samples were obtained behind the paver and subjeted to extration analysis. A substantial amount of segregation was observed in the binder ourse mix whih povided the opportunity to orrelate the aggregate grdation with the measured asphalt ontent. Some of the deviation in the measured asphalt ontent of the binder ourse mixes from the job mix formula (JMF) was determined to be the result of the hang in grdation of the mix from JMF. The perentages of matenal passmg the 4.75-mm (No. 4) and 2.36-mm (No. 8) sieves a:e orrelated wi_th measured asphalt ontents. For segregated bmder ourse mixes, equations were developed to adjust the measured asphalt ontent to aount for the hange in gradation from the JMF as measured on the 12.5-mm (Y2-in.) and either mm (No. 4) or 2.36-mm (No. 8) sieves. Asphalt ontent must be losely ontrolled in hot-mix asphalt (HMA) mixes to obtain optimum servieability and durability. An HMA pavement an ravel or rak if it is defiient in asphalt ontent by as little as Yi perent, whereas Yi perent exessive asphalt ontent an ause flushing and rutting. Quality ontrol and quality assurane (QA) of HMA pavements generally require the measurement of asphalt ontent in HMA mixes during prodution using either a standard extration test or a nulear asphalt ontent gauge. However, the measured value an vary from test to test beause of material, sampling, and testing variability. n reent years, the material variability has been redued substantially by the use of automated HMA failities. Testing profiieny an be improved through training. Obtaining a representative HMA sample for testing still remains a problem beause of either segregation or ineffetive sampling and splitting tehniques. When oarser mixes (binder and base ourses) made with larger maximum partile-sized aggregates are involved, the sampling variation an overshadow the material and testing variations. Coarse HMA mixes tend to segregate. The oarse aggregate fration in the HMA mix holds less asphalt ement by weight ompared with the fine aggregate fration. Segre- P. S. Kandhal, National Center for AsphaltTehnology, 211 Ramsay Hall, Auburn University, Auburn, Al S. A. Cross Civil Engineering Department, University of Kansas, 26 Learned Hall, Lawrene, Kans gation auses the proportions of oarse and fine aggregate partiles (therefore, the gradation) to vary in HMA samples and thus affet the measured asphalt ontents. There is a need to evaluate the effet of aggregate gradation on measured asphalt ontent so that an adjusted asphalt ontent that is loser to the asphalt ontent atually inorporated in the HMA mix an be asertained. PROJECT DETALS The test data for this study were obtained from a major fourlane nterstate paving projet in Pennsylvani This rehabilitation projet involved 5.8 mm (2 in.) of Pennsylvania D- 2 binder ourse (a dense-graded binder mix with a maximum aggregate size of 38.1 mm or lyi in.) and 38.1 mm (lyi in.) of Pennsylvania D-2 wearing ourse (a dense-graded wearing mix with a maximum aggregate size of 12.5 mm or Yi in.). The job mix formulas (JMFs) for the binder and wearing ourse mixtures are given in Tables 1 and 2, respetively. Northbound (NB) and southbound (SB) lanes were paved with separate pavers. Beause the mix aeptane or QA samples were obtained behind eah paver separately, the test data have been reported and analyzed separately for NB and SB lanes. Pennsylvania Department of Transportation (PennDOT) has a statistially based end result speifiation for HMA pavements that requires obtaining loose mix samples behind the paver at random loations. The entire loose mix is sraped out of a well-defined area (usually 229 x 229 mm or 9 x 9 in.) at the seleted random loation to minimize segregation as a result of sampling operation. Five loose mix sublot samples are obtained for eah lot onsisting of about 5 Mg ( 55 tons). These samples are sent to PennDOT entral laboratory for extration to detrmine the mix omposition. Roadway ores are also obtained after ompation and sent to the entral laboratory for determination of the pavement density. Prie adjustments for eah lot are alulated by the entral laboratory on the basis of three pay items: asphalt ontent, the perentage of material passing a 75-µm (No. 2) sieve, and the roadway density. A total of 547 binder mix samples (271 in NB lanes and 276 in SB lanes) and 147 wearing mix samples (67 in NB lanes and 8 in SB lanes) were obained behind the paver and tested by the entral laboratory. A substantial amount of segregation was observed in the ompated binder ourse mix of this projet apparently beause of mix handling and plaing operations. Obviously, the mix gradation of sublot samples obtained behind the paver varied onsiderably and affeted the extrated asphalt on-

2 22 TRANSPORTATON RESEARCH RECORD 1417 TABLE 1 Summary Statistis for Binder Mixes NB Lanes Test n=271 Parameter JMF Std. Mean Dev. Asphalt Content (%) Density (pf) N/A /2 nh(%} nh(%) /2 nh (%) No. 4 (o/o} No. 8 (%) No.16(%) No. 3 (o/o) No. 5 (%) No.1 (%) No. 2 (%) SB Lanes All n=276 n= 547 Std. Std. Mean Dev. Mean Dev tent. Beause a large number of binder mix samples were obtained at random loations behind the paver on this projet and were analyzed for mix omposition (asphalt ontent and gradation), a unique opportunity was available for evaluating the effet of aggregate gradation on the measured asphalt ontents. Material prodution variability was onsidered to be minimal on this projet beause an automated HMA faility was used, and the mix samples obtained at the faility were reasonably uniform in omposition. The testing variability is also onsidered to be minimal beause all extration testing was done in the PennDOT entral laboratory by essentially the same testing rew. ASTM D2172 (Method D) was used for extrating the asphalt ement from HMA mix samples. t is possible to ondut a similar study in a laboratory. A mix an be prepared with a known asphalt ontent, intentionally segregated, and then extrated. This would eliminate the inherent material variation. However, it is not possible to simulate the segregation that ours in the field. Also, it is not pratial to test a very large number of samples as was done in this study. TEST RESULTS Beause of spae restritions it is not possible to inlude in this paper the mix omposition test data for 547 binder mix samples and 147 wearing mix samples. However, Tables 1 TABLE 2 Summary Statistis for Wearing Mixes NB Lanes SB Lanes All Test n=67 n=8 n= 147 Parameter JMF Std. Std. Std. Mean Dev. Mean Dev. Mean Dev. Asphalt Content(%) Density (pf) N/A /2 nh (%) /8 nh (%) No. 4 (%) No. 8 (%) No. 16 (o/o) No. 3 (%) No. 5 (%) No.1 (%) No. 2 (o/o)

3 Kandhal and Cross and 2 give the summary statistis for binder mixes and wearing mixes, respetively. Figures 1 through 4 give the ontrol harts of the test data for asphalt ontent, the perent passing the 12.5-mm (Yz-in.), 4.75-mm (No. 4), and 2.36-mm (No. 8) sieves for 271 binder mix samples obtained from the NB lanes of the paving projet. The ontrol harts of the test data from the SB lanes are similar to those from the NB lanes and, therefore, are not inluded. ANALYSS OF TEST RESULTS The purpose of this study was to determine the effet of a hange in gradation on the orresponding measured asphalt ontent. f a strong orrelation exists between gradation and asphalt ontent, a part of the deviation from the JMF in the measured asphalt ontent ould be explained by the measured deviation in gradation. As mentioned earlier, the summary statistis of mean and standard deviation for the quality assurane data are shown in Table 1 for the binder mixes and Table 2 for the wearing mixes. For the binder mixes, the standard deviation is over 5 perent for perent passing the 25.4-mm (1-in.), 12.5-mm (Yz-in.), and 4.75-mm (No. 4) sieves, and.42 perent for asphalt ontent. Table 2 shows lower standard deviations for the wearing mixes for most sieve sizes; none of the sieve sizes had a standard deviation over 3.9 perent. The standard deviation for asphalt ontent was.31 perent for the wearing mixes. However, a review of the ontrol harts showed that the standard deviation for asphalt ontent might be artifiially high beause of an apparent hange in the JMF asphalt ontent by the ontrator that did not appear in the test reords (.)... "'ffi > (jj 4 v z. 35 C> "Ci) 3 en a FGURE 1 Control hart for asphalt ontent in binder mixes (NB lanes) FGURE 3 Control hart for passing No. 4 sieve in binder mixes (NB lanes). 9-8 > 7 (jj.. u. 6 C\ -y- C> 5 "Ci) en a > 35 (jj o z C> "Ci) 3 25 en a FGURE 2 Control hart for passing %-in. sieve in binder mixes (NB lanes) FGURE 4 Control hart for passing No. 8 sieve in binder mixes (NB lanes).

4 24 TRANSPORTATON RESEARCH RECORD 1417 Control harts of the test data for asphalt ontent and the perent passing the 12.5-mm (Yz-in.), 4.75-mm (No. 4), and 2.36-mm (No. 8) sieves for the binder mixes (NB lanes) are shown in Figures 1through4. The permissible tolerane limits for these four test parameters were ±.5, ± 8, ± 8, and ± 6 perent, respetively. Tables 3 and 4 show the frequeny with whih the above test parameters were within and outside the speifiation tolerane limits for the binder and wearing mixes, respetively. For the binder mix, asphalt ontent was outside the speifiation limits 23.4 perent of the time and the perent passing the 12.5-mm (Yz-in.), 4.75-mm (No. 4), and mm (No. 8) sieves 28.9, 2.3, and 17.5 perent of the time, respetively. From the ontrol harts and the data in Table 3, it is obvious that the binder mix sampled from the roadway was not uniform. Review of the test data and visual observations showed segregation of the mix to be a major problem on both the NB and SB lanes. Table 4 shows the frequeny with whih the wearing mix test parameters of asphalt ontent and the perent passing the 4.75-mm (No. 4) and 2.36-mm (No. 8) sieves were within speifiation tolerane limits. The permissible tolerane limits for these three test parameters were ±.4, ± 8, and ± 6 perent, respetively. Asphalt ontent was outside the speifi- ation limits 21.8 perent of the time, and the perent passing the 4.75-mm (No. 4) and 2.36-mm (No. 8) sieves 2.1and5.5 perent of the time, respetively. Review of the ontrol harts and test data showed that the gradation of the mix was within projet limits 95 perent of the time. Some of the satter in asphalt ontent ourred when the ontrator lowered the asphalt ontent on the NB lanes from 6.6 perent to approximately 6.2 perent after 35 tests. However, the available test data did not show a orresponding hange in the JMF asphalt ontent. f the JMF had been hanged to 6.2 perent, as the data indiate, and the appliable tolerane of.4 perent applied, the perent of the asphalt ontent tests within speifiation limits would have hanged from 76.1 to 97. perent for the NB lanes and from 78.2 to 87.8 perent for all of the dat Correlation analysis was performed to determine whether the mat density or the perentages passing various sieve sizes orrelate with asphalt ontent. Table 5 shows the results of _ the orrelation analysis for the binder mixes, by lane, and with all of the dat The results show that all of the parameters exept unit weight (density of the ore samples) and perent passing the 38.1-mm (1 Y2-in.) sieve have a high probability of a true orrelation (a =.1) with asphalt ontent. The best TABLE 3 Frequeny Distribution of Test Data for Binder Mixes NB Lanes SB Lanes All n Spe Asphalt Out - Low Content Out - High Perent n Spe /2 nh Out - Low Sieve Out - High Perent n Spe No.4 Out - Low Sieve Out - High Perent n Spe No. 8 Out Low Sieve Out - High TABLE 4 Frequeny Distribution of Test Data for Wearing Mixes NB Lanes SB Lanes All n Spe Asphalt Out - Low Content Out - High Perent n. Spe No. 4 Out - Low Sieve Out - High Perent n Spe No. 8 Out - Low Sieve Out - High

5 Kandhal and Cross 25 TABLE 5 Summary of Correlation Coeffiients (R) with Asphalt Content for Binder Mixes NB Lanes SB Lanes All n=271 n=276 n= 547 Parameter R Alpha* R Alpha* R Alpha* Density /2 nh N/A N/A N/A N/A 1 nh /2 nh No No No No No No No Alpha = Probability orrelation oeffiient (R) not equal to. orrelations with asphalt ontent for the binder mixes were with the perent passing the 4.75-mm (No. 4) and 2.36-mm (No. 8) sieves. The results of the orrelation analysis for the wearing mixes are shown in Table 6. The analysis shows the highest probability of a true orrelation (a =.1) with asphalt ontent for the perent passing the 3-µm (No. 5), 15-µm (No. 1), and 75-µm (No. 2) sieves. However, the orrelation oeffiients (R) are not only too low to be useful, but they also indiate an unexpeted trend-that is, the asphalt ontent dereases with an inrease in the material passing these sieves. To further investigate the relationship between asphalt ontent and gradation, regression analysis was performed. The purpose of this study is to determine whether asphalt ontent ould be predited from measured gradation; therefore, asphalt ontent was seleted as the dependent variable and gradation as the independent variable. Table 7 is a summary of the best oeffiients of determination (R2) by lane and by mix type for the binder and wearing mixes. The data in Table 7 indiate that no orrelation exists between asphalt ontent and the perent passing the mm (No. 4) and 2.36-mm (No. 8) sieves for the wearing mix. There is very little spread in the gradation data, and no segregation was observed in the field. Therefore, all the satter appears to be aused by the normal variation in the material, sampling, and testing operations. Figures 5 and 6 show the relationship between asphalt ontent and the perent passing the 4.75-mm (No. 4) and mm (No. 8) sieves for the binder mix in both lanes, respetively. The results show that there is a relationship between hange in gradation and measured asphalt ontent. The re- TABLE 6 Summary of Correlation Coeffiients (R) with Asphalt Content for Wearing Mixes NB Lanes SB Lanes ALL n=67 n=8 n= 147 Parameter R Alpha* R Alpha* R Alpha* Density /2 nh N/A N/A N/A N/A N/A N/A 3/8 nh No No: No No No No No alpha = Probability orrelation oeffiient (R) not equal to.

6 26 TRANSPORTATON RESEARCH RECORD 1417 TABLE 7 Summary of Coeffiients of Determination (R 2 ) with Asphalt Content for D2 Mixes NB Lanes Number of Observations n=271 R2 ndependent Variable 1 /2 nh Sieve.422 No. 4 Sieve.676 No. 8 Sieve /2 nh & No. 4 Sieves /2 nh & No. 8 Sieves.685 SB Lanes All n=276 n= 547 R2 R2 12 Binder Mixes Wearing Mixes No. 4 Sieve.11 No. 8 Sieve lationships show that as the mix beomes finer for the given sieve size, the asphalt ontent inreases. The relationships have the following form: AC = (P4) R 2 =.64 AC = (P8) R 2 =.63 where AC = asphalt ontent, P4 = perent passing the 4.75-mm (No. 4) sieve, and PB = perent passing the 2.36-mm (No. 8) sieve. Equations 1 and 2 indiate that the measured asphalt ontents of the binder ourse mix in this study inrease by.6 and.8 perent (on the basis of slopes of the regression lines) with eah 1 perent inrease in the material passing mm (No. 4) and 2.36-mm (No. 8) sieves, respetively, from the JMF. Conversely, there will be a similar derease in the measured asphalt ontents if the sampled mix is oarser than the y = ()() A-SQUARE =.64 -(fl a; 4.5.s:: No. 4 Sieve(%) FGURE 5 Perentage passing No. 4 sieve versus asphalt ontent (binder mixes from both lanes). 8 (1) (2) JMF. These so-alled "orretion fators" an be used to orret the measured asphalt ontent for eah 1 perent deviation from the JMF. Some researhers (1-3) have developed the following "orretion fators" for binder ourse mixes (maximum aggregate size greater than 25.4 mm or 1 in.) on the basis of the material passing the 2.36-mm (No. 8) sieve after analyzing limited field dat Researher Customary in United Kingdom for rolled-asphalt mix before 197 (J) Goodsall and Mathews (J) Warden (2) Brown et al. (3) Kandhal and Cross (this paper) Corretion Fator (%) The "orretion fator" is expeted to be generally dependent on the fine aggregate gradation, the partile shape and surfae texture of the aggregates, and the atual asphalt ontent of the binder ourse mix. Further analysis was performed to determine whether a multivariable model would give a statistially stronger model y ()() A-SQUARE =.63 -(fl as 4.5.s:: No. 8 Sieve (%) FGURE 6 Perentage passing No. 8 sieve versus asphalt ontent (binder from both lanes).

7 Kandhal and Cross The best multivariable model was found by inluding the mm (Vz-in.) sieve with either the 4.75-mm (No. 4) or mm (No. 8) sieve. The relationship between the asphalt ontent and the perent passing the 12.5-mm (V2-in.) sieve is shown in Figure 7. The relationship has an R 2 of.52. By ombining the 12.5-mm (1/2-in.) sieve with the 4.75-mm (No. 4) sieve the model has an R 2 of.67. The relationship has the following form: AC= (Pl/2) +.45(P4) R 2 =.67 (3) where AC = asphalt ontent, P 1 / 2 = perent passing 12.5-mm (1/z-in.) sieve, and P4 = perent passillg mm (No. 4) sieve. A slightly stronger model was found by using th 12.5-mm (V2-in.) an 2.36-mm (No. 8) sieves. The relationship is shown in Figure 8 and has the following form: AC= O.Ol6(Pl/2) +.6l(P8) R =.68 (4) where AC = asphalt ontent, P 1 / 2 = perent passing 12.5-mm (Y2-in.) sieve, and PB = perent passing 2.36-mm (No. 8) sieve. The data shown in Figurs 6 through 8 show that the measured asphalt ontent is affeted by a hange in gradation. A hange in gradation will ause a orresponding hange in the measured asphalt ontent. By using any of the four models, the measured asphalt ontent an be adjusted for the amount aused by the hange in gradation. The adjusted asphalt ontent an then be heked against the tolerane limits for the JMF asphalt ontent to determine whether the variation in asphalt ontent is aused by the hange in gradation or segregation or by a true hange in the asphalt ontent. To hek the models developed, the measured asphalt ontents were adjusted for the measured hange in gradation using Equations 1, 2, and 4. The adjusted asphalt ontent y = (X) A-SQUARE =.52 - '#. 6 --: 5.5 -: lll 5 () -as 4.5.: : ll /2 nh Sieve(%) FGURE 7 Perentage passing 1/i-in. sieve versus asphalt ontent (binder mixes from both lanes). (AAC) is determined by adding the differene between the measured asphalt ontent (MAC) and the predited asphalt ontent (PAC) to the JMF. The AAC is then heked against the upper and lower tolerane limits of the JMF asphalt ontent. AAC = JMFAC + (MAC - where PAC) AAC = asphalt ontent adjusted for gradation, JMF AC = job mix formula asphalt ontent, MAC = measured asphalt on.tent, and PAC = predited asphalt ontent from Equation 1, 2, 3, or 4. Figure 9 shows the ontrol harts for the asphalt ontent adjusted using Equation 5 for binder mix samples from NB lanes. Table 8 shows the frequeny with whih AAC, adjusted 7 - '#. 6.5 Y = (P1/2) +.45(P8) A-SQUARE --: 6 - () (ij -: 5.5 =.67.: 5 en <( 4.5 "'C = 4 * =. '- 3.5 ' En '._,, Atual Asphalt Content(%) FGURE 8 Atual versus predited asphalt ontent (binder mixes from both lanes). l 'E 5.5 -: (.) 5 ro "C 1i) ::J 4 :- <( ,------,----..,.---r ; FGURE 9 Control hart for asphalt ontent adjusted for 1/i-in. and No. 8 sieves (binder mixes from NB lanes) (5)

8 28 TRANSPORTATON RESEARCH RECORD 1417 TABLE 8 Frequeny Distribution of Adjusted Asphalt Content for Binder Mixes Asphalt Content n Spe. Adjusted on No. 4 Sieve (eq. 1) Out - Low Out - High NB Lanes SB Lanes All Adjusted on No. 8 Sieve (eq. 2) Adjusted on 1 /2" & No. 8 Sieve (eq. 4) n Spe. Out - Low Out - High n Spe. Out - Low Out - High using Equations 1, 2, and 4, is within speifiation limits. Tlie results show 95 perent of the AA Cs within speifiation limits regardless of the equations used. The AACs for the binder mix show a ompliane perentage very similar to that obtained for the wearing mixes in whih segregation was not a problem. CONCLUSONS AND RECOMMENDATONS On the basis of the data obtained in this study the following onlusions are warranted. 1. n segregated HMA pavements, some of the deviation in asphalt ontent from the JMF is ontrolled by the hange in gradation of the mix from the JMF. 2. When segregated binder ourse mixes were sampled behind the paver, the perent passing the 4.75-mm (No. 4) and 2.36-mm (No. 8) sieves orrelated with MAC. 3. For segregated binder ourse mixes, the asphalt ontent an be adjusted to aount for the hange in gradation from the JMF as measured on the 12.5-mm (Vz-in.) and either the 4.75-mm (No. 4) or 2.36-mm (No. 8) sieves, as shown in Equations 3 and 4. However, these equations are valid for the aggregates and the JMF used in this study. Care should be taken in applying these formulas to other mixes. 4. Beause no signifiant segregation ourred during the laydown of the wearing ourse mix, gradation ould not be related to the measured asphalt ontent. REFERENCES 1. Goodsall, G. D., and D. H. Mathews. Sampling of Road Surfaing Materials. Journal of Applied Chemistry, Vol. 2, De Warden, W. B. Bitumen Extration Testing. Presented at 6th World Meeting of the nternational Road Federation, Montreal, Quebe, Canada, Ot Brown, E. R., R. Collins, and J. R. Brownfield. nvestigation of Segregation of Asphalt Mixtures in State of Georgi n Transportation Researh Reord 1217, TRB, National Researh Counil, Washington, D.C., The opinions, findings, and onlusions expressed here are those of the authors and not neessarily those of the National Center for Asphalt Tehnology or Auburn University or the University of Kansas. Publiation of this paper sponsored by Committee on Charateristis of Nonbituminous Components of Bituminous Paving Mixtures.