Comparing the Volumetric and Mechanical Properties of Laboratory and Field Specimens of Asphalt Concrete (2016)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

62.1 Phase I: Levels of Variability in Volumetric and Mechanical Properties of Asphalt Mixtures As part of the literature review for this study, data were col- lected from projects around the country which could be used to meet the objectives of this study. This research effort is referred to as Phase I throughout this document. 2.1.1 Overview of Data Sets Analyzed Figure 2-1 presents a map identifying the states that con- tributed data to this initial project phase. In addition to the state DOTs presented in Figure 2-1, the researchers obtained volumetric measurements collected in the Netherlands and from the FHWA mobile laboratory. Tables 2-1 and 2-2 present the data sets analyzed and the properties available in each data set. As shown in these tables, most of the data sets included PL and PF samples; only two included LL samples. Statistical analysis of the individual data sets and meta-analysis of the combined data sets were conducted to quantify levels of variability for the three specimen types considered in this project (i.e., LL, PL, and PF). The follow- ing sections provide details and information about the anal- ysis conducted on each data set as well as the results of the meta-analysis. 2.1.2 Summary of the Statistical Analysis Tables 2-3 and 2-4 present the levels of variability for each of the volumetric, gradation, and mechanical properties eval- uated in Phase I. The data set from the Netherlands was not considered in this summary, because testing and construc- tion practices in Europe are different from those in the United States. The data received from Texas and Oregon were not suf- ficient for the analysis (e.g., mixtures only contained one speci- men type). In general, contractor and state measurements were similar and were shown to be statistically equivalent for most of the data sets. In addition, levels of variability presented in Table 2-3a and 2-3b were comparable for the state and the contractor measurements. Table 2-5 presents the average levels of variability for the volumetric and gradation proper- ties evaluated in Phase I. 2.2 Phase IA: Factors Causing Variability Between Specimen Types At the conclusion of Phase I, Phase IA was initiated to deter- mine the magnitude and factors causing pair-wise differences among the three specimen types [design (LL), production (PL), and construction (PF)]. With the guidance of NCHRP, the following projects were reviewed as possible additional sources of data: 1. WesTrack (WesTrack Database and NCHRP Web-Only Document 111); 2. NCAT test track; 3. NCHRP Project 09-47A, “Performance and Properties of Warm Mix Asphalt Technologies”; 4. California Heavy Vehicle Simulator data; 5. FHWA: Eastern, Central, and Western Federal Lands High- way Divisions; 6. Louisiana and Florida Accelerated Loading Facility (ALF) data; 7. Long-Term Pavement Performance (LTPP) data; 8. NCHRP Project 09-9(01), “Verification of Gyration Levels in the Ndesign Table”; 9. Arizona Department of Transportation (AZDOT) from NCHRP Project 09-22, “Beta Testing and Validation of HMA PRS,” and several AZDOT projects; 10. SHRP project reports and database; and 11. State planning and research reports. C H A P T E R 2 Phases I and IA

7 Aggregate gradation density was used as a quantitative method to identify mixes sensitive to minor changes in gra- dation and asphalt binder content and that may show greater variability between LL and PL specimens (D’Angelo and Fer- ragut 1991). Other mix types were identified in the literature as sensitive, including tender mixes, gap-graded mixes, and mixes that cross the maximum density line (MDL) multiple times. Tender mixes often exhibit a “hump” near the No. 30 sieve on the 0.45 power curve. However, none of the mixes in the Phase IA data sets was identified as tender, gap-graded, or as crossing the MDL multiple times. Aggregate gradation density was quantified through the sum of absolute differences (SAD) from the MDL (Anderson and Bahia 1997). The SAD was normalized (NSAD) to the num- ber of sieves reported in the data set, because not all projects reported the same number of sieves in the gradation analysis. As described by Equation 1, a mix with a high NSAD (i.e., above 8.25) indicates that its gradation deviates significantly from the MDL: % (1)NSAD P P n i MDL i n∑ = − where NSAD = normalized sum of absolute differences, n = number of sieves considered in the gradation analysis, Pi = % passing sieve i, and PMDL = % passing for the maximum density line at sieve i. Table 2-6 summarizes the data collected during Phase IA. Although the researchers successfully collected most of the aforementioned data sets, data from SHRP, FHWA Eastern and Central Federal Lands Highway Divisions, and the Cali- fornia Heavy Vehicle Simulators (HVS) were not available. In addition to the data sets suggested, the researchers success- fully collected data from the University of Nevada at Reno. Table 2-7 illustrates the factors identified to explain how construction processes may influence the magnitudes of the differences within and among the three specimen types (LL, PL, and PF). This project focused on process-based factors. Therefore, the influences of design-based factors [e.g., nom- inal maximum aggregate size (NMAS)] are not considered in this report. However, aggregate absorption and sensitive mixes were identified during the Phase I review and have been included in the analysis. Figure 2-1. State DOTs that provided data to the project.

Properes State Aggregate Gradaon Mixture Performed by Air Voids Asphalt Binder Content Gmb Density VMA VFA Gmm Specimen Type Test Method Specimen Type Test Method Specimen Type Test Method Specimen Type Test Method Specimen Type Specimen Type Specimen Type Contractor State Third- Party CA PL - - PL - - - PF - - - - Y Y - FHWA PL PL - PL IO NC PL - - - PL PL PL - - Y FL PL PL SSD - - - - PF SSD PL PL - Y Y Y IL - PL SSD PL IO - - PF NC - - - Y Y Y IN - LL, PL SSD PL IO - - PF SSD PL - - Y Y - IA - PL - - - PF SSD - - PL - Y - KS - PL - - - - - PF - - - - Y Y - KY PL PL SSD PL, PF IO, SE, NC, BC PT PL, PF SSD PF NC PL PL - Y Y - LA(1) - PL, PF SSD, VS PQI - - - - - - - - - - Y - LA(2) - PL SSD - - - - PF SSD - - - Y - NC PL PL - PL - PL - PF SSD PL PL PL Y Y - OK PL PL SSD PL - PL SSD - - PL - - - Y - WI - PL - PL - PL - PF - PL - PL - Y - OOMS PL PL, PF - PL - PL, PF - - - - - PL, PF Y - - Gmb: Mixture bulk specific gravity; Gmm: Mixture maximum specific gravity; SSD: Saturated Surface Dry; VS: Vacuum Sealing; PQI: Pavement Quality Indicator; IO: Ignion Oven; SE: Solvent Extracon; NC: Nuclear method; BC: Back Calculaon method; PT: Printed Ticket method; Y: ‘Yes.’; -: Not available. Table 2-1. Overview of the volumetric data sets.

IDT Tensile Strength Dynamic Modulus Flow Number Test Methods Source PL PF LL PL PF LL PL PF LL University of Arkansas X Louisiana X X MnROAD X X FHWA X X X X Table 2-2. Summary of data sets analyzed in task 2 (mechanical). (a) Volumetric Properes Prop es Performed By Specimen Type Variability Min. Max. Average AV, % Contractor PL 0.40 0.84 0.60 State PL 0.36 0.99 0.61 Third-Party PL 0.68 0.91 0.81 AC, % Contractor PL 0.17 0.22 0.19 State PL 0.17 0.24 0.20 Third-Party PL 0.18 0.21 0.20 VMA, % Contractor PL 0.37 0.58 0.49 State PL 0.38 0.65 0.53 Third-Party PL 0.51 0.64 0.58 VFA, % Contractor PL 3.40 4.08 3.73 State PL 4.01 4.93 4.34 Third-Party PL 4.20 5.16 4.68 Gmb Contractor PL 0.013 0.017 0.015 State PL 0.008 0.018 0.014 Third-Party PL 0.016 0.016 0.016 State PF 0.025 0.025 0.025 Gmm Contractor PL 0.012 0.012 0.010 State PL 0.008 0.012 0.009 Third-Party PL 0.011 0.011 0.011 Field Density, % Contractor PF 0.74 1.44 1.13 State PF 0.79 1.49 1.23 Third-Party PF 0.90 0.90 0.90 (b) Gradaon Properes Percent Passing Sieve Size Contractor State Third Party Range, % Avg., % Range, % Avg., % Range, % Avg., % Min Max Min Max Min Max 25.0 mm 1.70 2.66 2.12 1.74 1.79 1.77 0.68 0.68 0.68 19.0 mm 0.82 2.59 1.93 0.91 2.26 1.64 1.28 1.28 1.28 12.5 mm 0.91 3.54 2.14 1.08 2.54 1.79 0.89 2.15 1.52 9.5 mm 1.61 3.75 2.60 1.82 2.54 2.25 1.65 2.29 1.97 No. 4 1.87 3.48 2.71 2.19 3.08 2.66 2.37 2.56 2.47 No. 8 1.75 2.38 2.13 2.12 2.73 2.30 1.76 2.07 1.92 No. 16 1.56 2.05 1.81 1.70 1.76 1.73 NA NA NA No. 30 1.37 1.73 1.54 1.43 1.89 1.62 NA NA NA No. 50 1.12 1.28 1.18 1.07 1.27 1.17 NA NA NA No No . 100 0.64 0.99 0.78 0.76 0.83 0.80 NA NA NA . 200 0.34 0.84 0.60 0.39 0.66 0.52 0.40 0.40 0.40 Table 2-3. Summary of levels of variability (st. dev.) for volumetric and gradation properties.

10 2.2.1 Data Analysis This section presents results of the individual analyses con- ducted for Arizona DOT (AZDOT) and University of Nevada, Reno, data sets as typical data sets as well as a summary of the entire data analysis. Additional details for the other data sets were presented in the interim report for Phase IA (Mohammad et al. 2009), available at http://apps.trb.org/cmsfeed/TRBNet ProjectDisplay.asp?ProjectID=2503. Arizona DOT Data Analysis Data were collected from a research project conducted by the AZDOT. The primary objective of this research project was to formulate performance-based pay factor criteria using the concept of service life and remaining service life (Patni 2007). An increase or decrease in service life is a rational way to interpret the performance of in situ asphalt concrete mix- ture (field mix design) with respect to the laboratory mix design or the JMF. Table 2-8 describes the projects in the AZDOT data set. Table 2-9 summarizes the volumetric properties provided in the AZDOT data set. Each project had one mixture and several lots. Bulk material was sampled from each lot, out of which four samples were compacted in the laboratory. The asphalt binder content and gradation of the sampled bulk material were measured using the ignition furnace. Data analysis was conducted to determine the magnitude of the differences (D) between design values (LL) and as-produced mixtures (PL) as indicative of production variability and to identify possible effects of selected factors on the variability of mixture volumetric properties. The only process-based factor considered in the AZDOT data set was aggregate gradation density (i.e., NSAD). Table 2-10 summarizes the differences (D) between PL and LL volumetric properties for the AZDOT data set. These values represent the averages of ten mixtures. The gradation analysis was reduced to the four sieves shown in the table because the differences reported for all other sieves were negligible. The sieves analyzed are the sieves used for payment in Arizona. The absolute average differences shown Table 2-10 were calculated by taking the average of the absolute differences for all mixtures in the experiment. The positive and negative averages were cal- culated by taking the average of the sections in the experiment in which the difference was either positive or negative. Figures 2-2 and 2-3 show the differences between PL and LL volumetric properties grouped by NSAD. It appears from the results shown in Figure 2-3 that the differences between PL and LL samples increased as the mix gradation departed from the maximum density line (i.e., greater NSAD). How- ever, additional data are needed to verify this observation. University of Nevada (Reno) Data Analysis Data were collected from two projects conducted at the University of Nevada, Reno (UNR) (see Table 2-11). The objective of the first project, referred to as Experiment 1, was to compare the properties of a polymer-modified mixture (a) Volumetric Properes Property SpecimenType Range of St. Dev. Avg. Asphalt Binder Content, % PL 0.17 – 0.29 0.20 Air Voids, % PL 0.33 – 0.99 0.62 VMA, % PL 0.38 – 0.64 0.54 VFA, % PL 3.40 – 4.92 4.03 Gmb PL 0.008 – 0.018 0.015 Gmb PF 0.008 – 0.033 0.019 Gmm PL 0.005 – 0.012 0.011 Field Density, % PF 0.74 – 1.49 1.11 (b) Gradaon Properes Percent Passing Sieve Size, % Specimen Type Range of St. Dev. Average 25.0 mm PL 1.55 – 2.66 1.86 19.0 mm PL 0.93 – 2.59 1.77 12.5 mm PL 0.99 – 3.54 2.17 9.5 mm PL 1.50 – 3.75 2.35 No. 4 PL 1.87 – 3.48 2.62 No. 8 PL 1.62 – 2.62 2.20 No. 16 PL 1.70 – 2.05 1.81 No. 30 PL 1.43 – 1.84 1.60 No No . 50 PL 1.07 – 1.22 1.16 . 100 PL 0.80 – 0.99 0.87 No. 200 PL 0.32 – 0.84 0.55 Table 2-5. Average levels of variability (st. dev.) for volumetric and gradation properties. Properes COV Range, % Average COV, % Min Max Dynamic modulus 10.0 23.8 13.9 Phase angle 3.9 15.4 7.1 Flow number 37.3 52.1 45.2 Indirect tensile strength 11.9 15.4 13.7 Table 2-4. Summary of levels of variability (COV) for mechanical properties.

11 Data ID Source Meta Designaon Specimen Type Status Comments 1 NCAT NT PL, LL Collected 2006 and 2009 Experiments 2 NCHRP Project 9-9 9.9 PL, PF Collected Data from 1999 to 2002 3 SPR – AZDOT AZ PL, LL, PF Collected Volumetric Properes 4 FDOT HVS FL PL, LL, PF Collected Experiments 5 and 6 FL PL and LL 5 Louisiana ALF LA PL, LL, PF Collected Experiments 1, 2, 3, and 4 6 WesTrack WS PL, LL, PF Collected Original and Rehabilitaon 7 LTPP LT PL, LL, PF Collected SPS 1 and SPS 91 8 WF Lands WF PF, PL, LL Collected Three projects 9 LA Gmm Study LA PF, PL, LL Collected Five projects 10 LA 98-1B Study LA PL, LL Collected Three projects 11 Un. Nevada ---- PL, LL Collected No process-based factors2 12 NCHRP 9-22 ---- PL, LL Collected No process-based factors2 13 California HVS ---- N/A Not Collected Requests were turned down 1: SPS 1 had limited data and SPS 9 did not contain multiple specimen types. 2: Collected data did not identify process-based factors. Table 2-6. Summary and description of collected data sets. ID Factors Details Re co m m en de d Fa ct or s 1 Compacon methods Difference between field and laboratory compacon methods 2 Silo storage Extended storage me at the plant may harden asphalt in the mix 3 Baghouse fines May affect mix gradaon and other volumetric properes 4 Reheang May affect binder properes and thus compacted specimens 5 Aggregate absorpon May differ between plant and lab and thus affect variability 6 Plant type and se
ngs May affect mixture properes and thus variability 7 Sampling locaon Sampling locaon (e.g., plant, behind paver) may affect variability Ad di o na l F ac to rs 8 Gradaon density Sensive mixes are more suscepble to mix proporons than non-sensive mixtures 9 Material transfer device Use of MTD may reduce material and thermal segregaon 10 Aggregate degradaon Mixture producon may increase the fines fracon for soŒ aggregates 11 Aggregate moisture Moisture in the stockpile may affect mix properes Table 2-7. Factors considered as sources of variability within and among the three specimen types. Project Name No. of Lots Binder Type Sampled From NMAS, mm Blake Ranch 11 PG 64-22 Binder Course 19 Cienega Creek 16 PG 64-16 Binder Course 19 Clifford Wash 7 PG 64-16 Binder Course 19 Detrital Wash NB 24 PG 76-16 Binder Course 25 Detrital Wash SB 5 PG 70-10 Binder Course 19 Kaiser Springs 30 PG 70-16 Binder Course 19 Penzance Curves 13 PG 64-22 Binder Course 19 Sells Wash 11 PG 70-10 Binder Course 12.5 Temple Bar Road 14 PG 70-10 Binder Course 19 Two Guns 20 PG 64-22 Binder Course 19 Signal Road 24 PG 76-16 Binder Course 25 Table 2-8. Descriptions of the AZDOT data set. No. of Mixtures Specimen Type Property Replicates 10 Lab-mixed–Lab- compacted Asphalt Binder Content, Air Voids, Aggregate grada­on 1 Plant-mixed– Plant-compacted Asphalt Binder Content, Air Voids, Aggregate grada­on 20-120 Plant-mixed– Plant-compacted and Plant-mixed– Field-compacted Air Voids N/A Table 2-9. Volumetric properties in the AZDOT data set.

12 Category Comparison Properes Average Differences Range AAD -Avg +Avg Min Max Volumetric PL-LL AC,% 0.19 -0.18 0.21 -0.51 0.65 PL-PF AV,% 0.64 -0.68 0.53 -0.98 0.70 Gradaon PL-LL 9.5 mm,% 2.49 -1.56 2.73 -2.43 5.33 2.36 mm,% 1.23 -1.24 1.22 -2.67 3.86 0.6 mm,% 1.88 -1.89 1.89 -5.00 4.57 0.075 mm,% 0.48 -0.61 0.40 -0.92 0.98 Table 2-10. AZDOT data set summary statistics. -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 CW TG KS SR SW DWS BR TB CC PC Mixture ID 3.75-5.25 >5.25<3.75 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 CW TG KS SR SW DWS BR TB CC PC Mixture ID 3.75-5.25 >5.25<3.75 -3 -2 -1 0 1 2 3 4 5 6 CW TG KS SR SW DWS BR TB CC PC Mixture ID 3.75-5.25 >5.25<3.75 -4 -3 -2 -1 0 1 2 3 4 5 CW TG KS SR SW DWS BR TB CC PC ∆ = 2. 36 m m PL - 2. 36 m m LL , % ∆ = 9. 5m m PL - 9. 56 m m LL , % ∆ = AV PL - AV LL , % ∆ = AC PL - AC LL , % Mixture ID 3.75-5.25 >5.25<3.75 Figure 2-2. PL-LL asphalt binder content and gradation properties (grouped by NSAD).

13 -6 -4 -2 0 2 4 6 CW TG KS SR SW DWS BR TB CC PC Mixture ID 3.75-5.25 >5.25<3.75 -1.5 -1 -0.5 0 0.5 1 1.5 CW TG KS SR SW DWS BR TB CC PC Mixture ID 3.75-5.25 >5.25<3.75 ∆ = 0. 07 5m m PL - 0. 07 5m m LL , % ∆ = 0. 3m m PL - 0. 3m m LL , % Figure 2-2. (Continued). 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 <3.75 3.75-5.25 >5.25 NSAD 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 <3.75 3.75-5.25 >5.25 NSAD ∆ = Av g. │A V P L - A V L L│ , % ∆ = Av g. │A C P L - A C L L│ , % Figure 2-3. Absolute average differences (PL-LL) for asphalt binder content and gradation properties (grouped by NSAD). (continued on next page)

14 0 1 2 3 4 5 6 <3.75 3.75-5.25 >5.25 NSAD 9.5 mm 2.36 mm 0.6 mm 0.075 mm ∆ = Av g. │P as sin g P L - P as sin g L L│ , % Figure 2-3. (Continued). (AC-20P) to the properties of a mixture prepared using a high-viscosity base binder (AC-30). This 1994 study included an extensive laboratory factorial evaluating three Hveem- designed mixtures with varying asphalt binder type and gra- dation (Farooq and Sebaaly 1994). Three LL mixtures were tested. However, only two of the PL mixtures were evaluated. Hence, only two comparisons were available for evaluation. The laboratory data collected included the Lottman tensile strength ratio, IDT strength, resilient modulus (Mr), perma- nent deformation, and thermal cracking tests (see Table 2-11). In addition, some volumetric properties were available: mix- ture bulk specific gravity (Gmb), mixture maximum specific gravity (Gmm), and air voids (AV). The second project, referred to as Experiment 2, compared mixtures designed using Superpave with mixtures designed using Hveem design methods (Sebaaly et al. 2005). Superpave mixture gradation satisfied the control points but did not con- sider the limits of the restricted zone. The experimental fac- torial consisted of testing both LL and PL specimens for the six mixtures. The mechanical tests evaluated included tensile strength ratio [TSR], Freeze Thaw (FrT), Asphalt Pavement Analyzer (APA), Repeated Load Test (RLT), Simple Shear Test (SST), and dynamic modulus (E*) (as shown in Table 2-11). Data analysis was conducted to determine the magnitude of the differences between design values (LL) and produc- tion mixtures (PL). However, the process-based factors in the experiment were not varied, which did not allow the assess- ment of the causes of the differences and variability among the three specimen types. To serve as a reference to the calculated values, differences are expressed in terms of the percentage differences from LL measurements. Tables 2-12 and 2-13 summarize the differ- ences (D) for the UNR data sets. The absolute average differ- ence (AAD) values shown in these tables are the average of all the absolute differences for all mixtures in the experiment. The positive and negative averages were calculated by taking the average of the sections in the experiment in which the difference was either positive or negative. The range shown is between the largest negative difference and the largest posi- tive difference observed in all sections. Differences between PL Year Designaon No. of Mixtures Specimen Type Mechanical Test Replicates 1994 Experiment 1 2 Lab-mixed–Lab- compacted ITS, Mr, TSR, Ec2 3 Plant-mixed–Plant- compacted ITS, Mr, TSR 3 2005 Experiment 2 3 Lab-mixed–Lab- compacted TSR, FrT, APA, RLT, SST, E* 2-31 Plant-mixed–Plant- compacted TSR, FrT, APA, RLT, SST, E* 2-31 1APA consisted of two replicates; 2Creep Modulus Table 2-11. Overview of UNR data set.

15 Table 2-12. UNR delta summary statistics (Experiment 1). Category Comparison Properes Average Differences Range AAD -Avg +Avg Min Max Mechanical (PL-LL)/LL Mr@0°C, % LL 13.3 N/A 13.3 12.0 14.6 (PL-LL)/LL Mr@34°C, % LL 39.3 N/A 39.3 23.4 55.2 (PL-LL)/LL Mr@77°C, % LL 10.5 -4.4 16.6 -4.4 16.6 (PL-LL)/LL Mr@104°C, % LL 16.0 N/A 16.0 9.3 22.6 (PL-LL)/LL TSR, % LL 12.6 N/A 12.6 4.7 20.5 Volumetric PL-LL Gmb .047 -0.047 N/A -0.088 -0.007 PL-LL Gmm .063 N/A 0.063 0.008 .118 Table 2-13. UNR delta summary statistics (Experiment 2). Category Comparison Properes Averages Range AAD -Avg +Avg Min Max Mechanical (PL-LL)/LL TSR, % LL 9.5 ---- 9.5 4.1 14.6 (PL-LL)/LL Rut Depth (APA), % LL 19.4 -29.7 14.2 -47.3 32.9 (PL-LL)/LL Accumulated Strain (RLT), % LL 22.4 -44.1 42.6 -72.2 42.6 (PL-LL)/LL Accumulated Strain (RLT), % LL 2.1 -2.1 ---- -3.5 ---- (PL-LL)/LL Accumulated Strain (SST), % LL 40.0 ---- 40.0 24.9 56.2 (PL-LL)/LL Dynamic Modulus @ 14°F, 25Hz,10Hz, 5Hz, 1Hz, 0.5Hz, 0.lHz, % LL 49.3 -26.3 65.7 -49.6 87.7 (PL-LL)/LL Dynamic Modulus @ 40°F, 25Hz,10Hz, 5Hz, 1Hz, 0.5Hz, 0.lHz, % LL 63.8 -19.6 88.1 -35.3 245.6 (PL-LL)/LL Dynamic Modulus @ 70°F, 25Hz,10Hz, 5Hz, 1Hz, 0.5Hz, 0.lHz, % LL 77.7 -10.2 91.1 -17.7 232.4 (PL-LL)/LL Dynamic Modulus @100°F, 25Hz,10Hz, 5Hz, 1Hz, 0.5Hz, 0.lHz, % LL 65.8 -10.6 97.7 -25.1 179.2 (PL-LL)/LL Dynamic Modulus @130°F, 25Hz,10Hz, 5Hz, 1Hz, 0.5Hz, 0.lHz, % LL 46.3 -11.9 54.9 -17.5 168.0 and LL complex modulus values did not appear to be influ- enced by the test temperatures. In addition, the TSR values of PL samples were greater than those of LL samples for both experiments. This may be attributed to asphalt binder oxida- tion during production. 2.2.2 Summary of the Data Analysis Tables 2-14 and 2-15 present the levels of variability for each of the volumetric and mechanical properties evaluated. The confidence intervals for the means shown in these tables were calculated based on Equation 2 (Law 2007): Confidence Limits ADD t stdev (2)n-1,1- /2 p= ± α where AAD = absolute average difference, tn-1, 1-a/2 = is the upper 1 - a/2 critical point for a t distribu- tion with n - 1 degrees of freedom (a is the level of significance set at 5%), and stdev = standard deviation of the AAD for the analyzed data sets. Confidence intervals are not presented for the properties in which only one data set was available. Other differences, indicated by N/A, were not available in the collected data sets. As shown in these tables, differences among the three speci- men types varied widely as evidenced by the high standard deviation and the confidence intervals computed for some of the properties. These wide variations are due to many factors,

16 Table 2-14. Summary of differences among the three specimen types for volumetric and gradation properties. Property Comparison Average Differences Confidence Intervals AAD St. Dev. Low Limit High Limit Asphalt Binder Content, % PF-PL 0.250 ---- ---- ---- PF-LL 0.237 0.021 0.170 0.303 PL-LL 0.277 0.131 -0.020 0.573 Air Voids, % PF-PL N/A N/A N/A N/A PF-LL 0.510 0.286 -0.400 1.420 PL-LL 0.806 0.295 0.137 1.474 VMA, % PF-PL N/A N/A N/A N/A PF-LL N/A N/A N/A N/A PL-LL 1.230 0.537 -1.082 3.542 Gmm PF-PL 0.038 ---- ---- ---- PF-LL 0.030 ---- ---- ---- PL-LL 0.032 0.022 -0.030 0.094 Gmb PF-PL 0.059 0.052 -0.166 0.284 PF-LL 0.054 ---- ---- ---- PL-LL 0.250 ---- ---- ---- Average Differences Confidence Intervals AAD St. Dev. Low Limit High Limit 25 mm PL-LL 1.56 2.133 -5.227 8.347 19 mm PF-LL 0.09 ---- ---- ---- PL-LL 0.779 0.935 -1.431 2.989 12.5 mm PF-LL 1.19 ---- ---- ---- PL-LL 1.367 0.645 -0.092 2.826 9.5 mm PF-LL 1.14 ---- ---- ---- PL-LL 2.246 1.257 -0.554 5.046 4.75 mm PF-LL 1.15 ---- ---- ---- PL-LL 2.079 1.202 -0.639 4.797 2.36 mm PF-LL 0.71 ---- ---- ---- PL-LL 1.829 1.216 -0.881 4.539 2.0 mm PL-LL 3 ---- ---- ---- (a) Volumetric Properes (b) Gradaon Properes Percent Passing Sieve Size, % Comparison

17 1.18 mm PF-LL 0.78 ---- ---- ---- PL-LL 1.538 1.078 -0.948 4.023 0.6 mm PF-LL 0.77 ---- ---- ---- PL-LL 1.721 1.25 -1.064 4.506 0.425 mm PL-LL 2.25 ---- ---- ---- 0.3 mm PF-LL 0.73 ---- ---- ---- PL-LL 1.653 1.516 -1.777 5.083 0.18 mm PL-LL 2.75 ---- ---- ---- 0.15 mm PF-LL 0.79 ---- ---- ---- PL-LL 0.855 0.541 -0.392 2.102 0.075 mm PF-LL 0.97 ---- ---- ---- PL-LL 0.617 0.388 -0.247 1.481 Percent Passing Sieve Size, % Comparison Average Differences Confidence Intervals AAD St. Dev. Low Limit High Limit Table 2-14. (Continued). Table 2-15. Summary of differences among the three specimen types for the mechanical properties. Category Property Comparison Average Differences Confidence Intervals AAD St. Dev. Low Limit High Limit Moisture Suscepbility Tensile Strength Rao, (TSR), % LL (PF-PL)/LL 10.55 2.47 -0.10 21.20 (PF-LL)/LL 19.00 2.12 9.87 28.13 (PL-LL)/LL 16.73 7.54 -4.21 37.66 Low- Temperature Cracking Temperature @ fracture (T), % LL (PF-PL)/LL 15.25 3.61 -0.27 30.77 (PF-LL)/LL 14.60 12.73 -40.17 69.37 (PL-LL)/LL 20.35 4.31 1.79 38.91 Stress @ fracture (s), % LL (PF-PL)/LL 33.20 0.99 28.94 37.46 (PF-LL)/LL 17.85 9.40 -22.62 58.32 (PL-LL)/LL 31.90 1.98 23.38 40.42 Beam Fague No. Cycles @ failure (Nf), % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL 42.50 N/A ---- ---- (PL-LL)/LL N/A N/A N/A N/A Phase Angle (PA), % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL 15.30 N/A ---- ---- (PL-LL)/LL N/A N/A N/A N/A Sffness (St), % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL 19.50 N/A ---- ---- (PL-LL)/LL N/A N/A N/A N/A APA Rut Depth (APA), % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 19.40 N/A ---- ---- RLT % Strain @ 12,000 Cycles, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 22.40 N/A ---- ---- Cycles to 3% Strain, %LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 2.10 N/A ---- ---- (continued on next page)

18 Dynamic Modulus @ 70°F, 25Hz, 10Hz, 5Hz, 1HZ, 0.5Hz, 0.1Hz, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 77.70 N/A ---- ---- Dynamic Modulus @ 100°F, 25Hz, 10Hz, 5Hz, 1HZ, 0.5Hz, 0.1Hz, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 65.80 N/A ---- ---- Dynamic Modulus @ 130°F, 25Hz, 10Hz, 5Hz, 1HZ, 0.5Hz, 0.1Hz, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 46.30 N/A ---- ---- Resilient Modulus (Mr) Mr @ 0°C, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 13.30 N/A ---- ---- Mr @ 34°C, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 39.30 N/A ---- ---- Mr @ 77°C, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 10.50 N/A ---- ---- Mr @ 104°C, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 16.00 N/A ---- ---- Category Property Comparison Average Differences Confidence Intervals AAD St. Dev. Low Limit High Limit Axial Dynamic Modulus SST Strain @ 5,000 Cycles, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 40.00 N/A ---- ---- Axial Dynamic Modulus Dynamic Modulus @ 14°F, 25Hz, 10Hz, 5Hz, 1HZ, 0.5Hz, 0.1Hz, % LL (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A (PL-LL)/LL 49.30 N/A ---- ---- Dynamic Modulus @ 40°F, 25Hz, 10Hz, 5Hz, 1HZ, 0.5Hz, (PF-PL)/LL N/A N/A N/A N/A (PF-LL)/LL N/A N/A N/A N/A Axial Dynamic Modulus 0.1Hz, % LL (PL-LL)/LL 63.80 N/A ---- ---- Table 2-15. (Continued). including differences in construction practices among agen- cies, differences in mix characteristics and designs among projects, and differences in variability between the different projects. Confidence intervals could not be developed for most of the mechanical properties because only two data sets were used. 2.3 Meta-Analysis The statistical analyses presented in the previous sections were conducted on a per-data-set basis, and a summary was compiled to provide an overall quantification of the levels of differences between the three specimen types. However, many of the process-based factors were not documented in the data collected and were not known. As a result of these limitations, the effects of the identified factors on the calculated variability could not be directly assessed because only one condition was used in most of the collected data sets. To address this limita- tion, the individual data were combined in a meta-analysis that made use of data obtained from different sources to identify the influences of process-based factors on volumetric proper- ties among the three specimen types (i.e., LL, PL, and PF). In the meta-analysis, calculated differences (PL–LL, PF–LL, and PF–PL) among the three specimen types were combined in “meta-data” sets, assumed to originate from the same popu- lation. Statistical t-tests were then conducted to test the null hypothesis that the means of the differences between two of the three specimens are equal when a change is made to only one

19 of the process-based factors. Some of the data grouped into the “meta-data” set did not originate from the same source, and the influences of unforeseen factors, such as mix design (i.e., NMAS), material properties, and construction practices in different states, may affect the validity of the comparison. In addition, conclusions from the t-test may be affected by the large difference in the mixtures of the meta-data set. Therefore, results of this analysis should only serve as a general indicator of future research needs. A total of 230 mixtures were included in the meta-analysis. The volumetric properties considered were asphalt binder con- tent, AV, VMA, Gmm, Gmb, and gradation. Mechanical proper- ties were not evaluated because process-based factors were unknown or not varied in the combined data sets. Evaluated factors included use of baghouse fines, reheating, aggregate absorption, plant type, sampling location, and use of a material transfer vehicle. Statistical analysis was performed using the statistical analysis software (SAS). The SAS T-test procedure was used to compare the means of the differences between two of the three specimens when a change was made to only one of the aforementioned process-based factors. An F-test was used to check the equality of variances, and the appropriate p-value is reported. The p-values are summarized in Table 2-16 for the PL–LL differences. Blank cells indicate that either no obser- vation was available for the factor or that the factor did not have two levels. The N values presented in Table 2-16 provide the sample sizes for the compared populations. The shaded cells are the statistical comparisons that were found signifi- cant, indicating that the evaluated factor may have an influ- ence on the difference between the two specimen types. As shown in Table 2-16, most of the comparisons did not show statistical influences of the evaluated factors. Due to lack of available data, the tables developed for the PF–LL and PF–PL differences were mostly empty and are reported in Appendix B (available on the project webpage). In addition to the results of the statistical analysis pre- sented in the previous section, attempts were made to iden- tify trends in the data by visual comparison using charts. The charts corresponding to statistically significant trends identi- fied in Table 2-16 are shown in this section. Other compara- tive charts are presented in Appendix A. Figure 2-4 shows the delta chart for asphalt binder content sorted by sampling location. The number of field-sampled mixtures was much larger than the number of mixtures sampled at the plant. This may have affected the finding of the t-test. The mean differ- ence for the field samples was -0.06% compared to the mean of +0.04% for the plant samples. This may be due to further aggregate absorption of asphalt binder during transportation. Figure 2-5 shows the delta chart for aggregate gradation percent passing 9.5 mm grouped by reheating. Only 6 mix- tures experienced reheating compared to 87 without reheat- ing. Conclusions from the t-test may be affected by the large difference in the number of mixtures for each grouping. The mean for the non-reheated group was 2.0% as compared to PL-LL Baghouse Reheating Aggregate Absorption Plant Type Sampling Location MTV Pvalue NYES NNO Pvalue NYES NNO Pvalue NLOW NHIGH Pvalue NDRUM NBATCH Pvalue NPLANT NFIELD Pvalue NYES NNO AC 0.23 25 6 0.09 7 133 0.72 24 3 0.97 103 9 0.01 17 52 0.65 75 17 AV 0.40 53 6 0.23 9 86 0.33 22 3 - - - 0.07 19 40 0.47 101 19 Gmm - - - 0.76 5 5 - - - 0.80 31 9 - - - - - - Gmb - - - - - - - - - 0.04 10 8 - - - - - - VMA 0.43 9 6 0.76 4 72 <0.01 12 3 - - - 0.43 9 6 0.39 67 9 25 - - - 0.88 3 68 - - - - - - - - - 0.83 67 4 19 0.30 49 6 0.98 6 86 0.55 19 2 - - - 0.84 15 40 0.10 105 11 12.5 0.91 50 6 0.30 6 87 0.87 19 3 - - - 0.59 16 40 0.34 101 16 9.5 0.12 50 6 0.01* 6 87 0.08 19 3 - - - 0.62 16 40 0.12 101 16 4.75 0.32 50 6 <0.01 6 87 0.51 19 3 - - - 0.49 16 40 <0.01 101 16 2.36 0.39 47 6 0.30 5 85 0.54 17 2 - - - 0.78 13 40 0.79 101 13 1.18 - - - 0.01 3 82 - - - - - - 0.59 14 32 0.97 101 8 0.6 0.98 48 6 0.09 6 85 0.89 18 2 - - - 0.32 14 40 0.45 101 14 0.3 0.86 48 6 0.10 6 85 0.81 18 2 - - - 0.75 14 40 0.83 101 14 0.15 - - - - - - - - - - - - <0.01 6 34 0.26 95 6 0.075 0.43 49 6 0.01 5 87 0.81 18 3 - - - 0.93 15 40 0.05 101 15 *: Pvalue < 0.05 indicates statistical significance Table 2-16. Meta-analysis PL-LL statistical results summary.

20 -1.9% for the reheated group. Figure 2-6 shows the delta chart for aggregate gradation percent passing 4.75 mm grouped by reheating. Similarly, for this sieve, only 6 mixtures experi- enced reheating compared to 87 without reheating. Conclu- sions from the t-test may be influenced by the large disparity in the number of mixtures for each grouping. The mean for the non-reheated group was 2.2% as compared to a mean of -1.6% for the reheated group. Figure 2-7 shows the delta chart for the aggregate gradation percent passing 4.75 mm grouped by material transfer vehicle (MTV) use. Similarly, results of the t-test may be affected by the large difference in the number of mixtures for each grouping. Furthermore, the sampling location was not known for about 50% of the data points in the MTV group, which could influ- ence the conclusions, because plant-sampled materials will not be affected by the use of MTV. About 60% of the data in the “No MTV” group came from Louisiana and the Florida accel- erated pavement test projects, which had low volumes of pro- duction. Figure 2-8 shows the delta chart for percent aggregate passing 1.18 mm grouped by reheating. For this comparison, only 3 mixtures experienced reheating compared to 82 without reheating. Conclusions from the t-test may be influenced by the large inequality in the number of mixtures for each group- ing. The mean difference for the reheated group was -1.7% compared to a mean of 1.0% for the non-reheated group with most of the differences in the ± 4% range. The highest data point in the non-reheated group comes from a mixture which was adjusted during production. If the post-adjustment peak is considered, this difference reduces substantially. Figure 2-9 shows the delta chart for percent aggregate pass- ing 0.15 mm grouped by sampling location. For this com- parison, only 6 mixtures were sampled at the plant compared to 34 sampled in the field. The mean difference for the field group was 0.2% as compared to 1.0% for the plant group. Most of the differences were within a range of ± 2%. The peak in the plant-sampled mixture came prior to an adjustment in the production of the mixture. If the post-adjustment sam- ples are considered only, the delta is reduced considerably. -1.000 -0.800 -0.600 -0.400 -0.200 0.000 0.200 0.400 0.600 0.800 1.000 9. 9- 15 9. 9- 18 9. 9- 22 9. 9- 24 9. 9- 26 9. 9- 01 9. 9- 03 9. 9- 05 9. 9- 08 9. 9- 11 9. 9- 13 9. 9- 16 9. 9- 21 9. 9- 29 9. 9- 31 9. 9- 34 9. 9- 36 9. 9- 38 9. 9- 10 W S- 21 W S- 23 W S- 25 W S- 29 W F- 01 W F- 03 W F- 02 FL -0 5 FL -0 1 FL -0 3 LA -0 6 LA -0 3 LA -0 7 LA -1 8 LA -1 6 LA -1 5 Mixture ID Field Plant ∆ = AC PL - AC LL , % Figure 2-4. Meta-analysis: PL-LL asphalt binder content (grouped by sampling location). -10.0 -8.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 LA -0 3 W F- 01 AZ -0 8 AZ -0 3 AZ -0 6 AZ -0 7 N T- 40 N T- 33 N T- 43 N T- 18 N T- 19 N T- 12 N T- 02 N T- 38 N T- 27 N T- 44 N T- 47 N T- 25 N T- 42 N T- 46 N T- 33 N T- 15 N T- 34 N T- 08 N T- 07 N T- 09 FL -0 4 FL -0 1 LA -1 7 LA -1 6 W F- 01 Mixture ID Reheated Not Reheated ∆ = 9. 5m m PL - 9. 56 m m LL , % Figure 2-5. Meta-analysis: PL-LL aggregate gradation percent passing 9.5 mm (grouped by reheating).

21 Figure 2-6. Meta-analysis: PL-LL aggregate gradation percent passing 4.75 mm (grouped by reheating). ∆ = 4. 75 m m PL - 4. 75 m m LL , % -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 LA -0 3 AZ -0 3 AZ -0 8 AZ -0 3 AZ -0 6 AZ -0 7 N T- 40 N T- 33 N T- 43 N T- 18 N T- 19 N T- 12 N T- 02 N T- 38 N T- 27 N T- 44 N T- 47 N T- 25 N T- 42 N T- 46 N T- 33 N T- 15 N T- 34 N T- 08 N T- 07 N T- 09 FL -0 4 FL -0 3 LA -1 7 LA -1 6 W F- 01 Mixture ID Not ReheatedReheated Figure 2-7. Meta-analysis: PL-LL aggregate gradation percent passing 4.75 mm (grouped by MTV). -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 W S- 01 W S- 27 W S- 34 W S- 05 W S- 09 W S- 14 W S- 18 W S- 23 W S- 29 N T- 40 N T- 36 N T- 01 N T- 19 N T- 09 N T- 23 N T- 27 N T- 22 N T- 32 N T- 42 N T- 13 N T- 36 N T- 34 N T- 06 N T- 12 LA -1 8 LA -1 5 FL -0 1 LA -0 2 W F- 02 W F- 03 Mixture ID NO MTVMTV ∆ = 4. 75 m m PL - 4. 75 m m LL , % Figure 2-8. Meta-analysis: PL-LL aggregate gradation percent passing 1.18 mm (grouped by reheating). -4.000 -2.000 0.000 2.000 4.000 6.000 8.000 LA -0 2 AZ -0 1 AZ -0 4 AZ -0 7 AZ -1 0 FL -0 3 LA -1 4 LA -1 7 N T- 01 N T- 03 N T- 06 N T- 09 N T- 11 N T- 13 N T- 16 N T- 18 N T- 20 N T- 22 N T- 25 N T- 27 N T- 29 N T- 32 N T- 34 N T- 36 N T- 38 N T- 41 N T- 43 N T- 46 N T- 49 Mixture ID Not ReheatedReheated ∆ = 1. 18 m m PL - 1. 18 m m LL , %

22 Conclusions from the t-test may be affected by the large dif- ference in the number of mixtures for each grouping. Figure 2-10 shows the delta chart for percent aggregate pass- ing 0.075 mm grouped by reheating. Conclusions from the t-test may be affected by the large inequality in the number of mixtures for each grouping. Figure 2-11 shows the delta chart for mixture bulk specific gravity grouped by plant type. For this comparison, sample sizes are similar, albeit small, between the two groups. However, the data in the analysis is all from a single dataset (LTPP). One may hypothesize from this com- parison that the difference between PL and LL for Gmb is greater for drum plant than for batch plant. However, the p-value from the t-test is nearly insignificant (0.04) at 95% confidence. Figure 2-12 shows the delta chart for voids in the min- eral aggregate (VMA) grouped by aggregate absorption. For this comparison, only 3 mixtures used highly absorptive aggregate as compared to 12 mixtures using non-absorptive aggregate. Conclusions from the t-test may be influenced by the large inequality in the number of mixtures for each grouping. Most the data points were in the ±1% range. 2.4 Conclusions and Findings of Phase IA The objective of Phase IA was to determine the cause and magnitude of the differences and variances in measured volu- metric and mechanical properties among three specimen types (i.e., laboratory-mixed–laboratory-compacted [LL], plant- mixed–laboratory-compacted [PL], and plant-mixed–field- compacted [PF]). In Phase IA of this project, specific highway and research agencies were contacted to collect existing vol- umetric and mechanical data in order to achieve the objec- tives of the project. Individual data analysis was conducted to quantify levels of differences among the three specimen types for volumetric and mechanical properties. The analysis found that the influence of NSAD on the different volumet- -1.000 -0.500 0.000 0.500 1.000 1.500 2.000 2.500 3.000 W S- 01 W S- 20 W S- 27 W S- 32 W S- 34 W S- 03 W S- 05 W S- 07 W S- 09 W S- 11 W S- 14 W S- 16 W S- 18 W S- 21 W S- 23 W S- 25 W S- 29 FL -0 5 FL -0 1 FL -0 3 Mixture ID Field Plant ∆ = 0. 15 m m PL - 0. 15 m m LL , % Figure 2-9. Meta-analysis: PL-LL aggregate gradation percent passing 0.15 mm (grouped by sampling location). ∆ = 0. 07 5m m PL - 0. 07 5m m LL , % -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000 4.000 LA -0 2 W F- 02 AZ -0 2 AZ -0 5 AZ -0 8 FL -0 1 FL -0 4 LA -0 7 LA -1 6 LA -1 9 N T- 02 N T- 05 N T- 08 N T- 10 N T- 12 N T- 15 N T- 18 N T- 19 N T- 21 N T- 24 N T- 26 N T- 28 N T- 31 N T- 33 N T- 35 N T- 37 N T- 40 N T- 43 N T- 45 N T- 48 W F- 02 Mixture ID Not ReheatedReheated Figure 2-10. Meta-analysis: PL-LL aggregate gradation percent passing 0.075 mm (grouped by reheating).

23 ric properties was mixed and was mostly inconclusive. Major limitations were encountered, because the collected data sets did not methodically vary most of the process-based factors identified as potential causes of variability. In addition, many of the process-based factors were not documented in the data collected and were not known. Because of these limitations, the effects of the identified factors on the calculated variability could not be directly assessed because only one condition was used in most of the collected data sets. To address this limitation, the collected data were combined in a meta-analysis that made use of data obtained from dif- ferent sources to identify causes and levels of variability for volumetric and mechanical properties among the three speci- men types (i.e., LL, PL, and PF). However, these data were not homogeneous and the influences of unforeseen factors, such as mixture design, were not considered. It is difficult to deter- mine whether or not the statistically significant differences determined by the meta-analysis were caused by sample size inequalities or if they were true representations of the effects of the process-based factors. Therefore, results of this analysis only served as a general indicator of the need for continued research, which is addressed in Phase II of this study. -0.04 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 LT -1 3 LT -1 4 LT -1 6 LT -1 1 LT -4 6 LT -4 7 LT -4 8 LT -6 7 LT -4 1 LT -4 2 LT -4 4 LT -4 5 LT -3 7 LT -3 8 LT -3 9 LT -3 5 LT -4 0 LT -5 0 Mixture ID Batch Drum ∆ = Gm b P L - G m b L L Figure 2-11. Meta-analysis: PL-LL mixture bulk specific gravity (grouped by plant type). ∆ = VM A P L - V M A L L, % -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 W F- 01 W F- 01 LA -0 1 LA -1 6 LA -1 8 LA -1 7 LA -1 4 LA -1 5 LA -1 9 W F- 02 W F- 03 W F- 02 W F- 03 LA -0 7 LA -0 8 Mixture ID High Low Figure 2-12. Meta-analysis: PL-LL voids in the mineral aggregate (grouped by aggregate absorption).