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.

60 C H A P T E R 6 6.1 Specification Recommendations This section presents the development of a draft proposed AASHTO recommended practice that addresses the cause and magnitude of variability within and among the three speci- men types (i.e., LL, PL, and PF). Data collected in Task 4 were used to develop the specification recommendations. 6.1.1 Single-Operator Tolerance Among Specimen Types The individual data sets were combined to calculate the expected deviation among specimen types. The delta val- ues from the 10 mixtures were assumed to originate from the same population. Table 6-1 presents the average, mini- mum, and maximum differences observed from the mixtures evaluated. The confidence limit represents the 95% confi- dence band for the parameters measured. The confidence limit was determined by multiplying the standard deviation of the differences by the t-value associated with alpha = 0.05 (ta=0.05 = 1.96). Equation 21 represents the equation used to develop the 95% confidence intervals for each of the design vs. production parameters shown in Table 6-1. Tolerance, x Standard Deviation, x t (21) i,LL-PL i Delta, LL-PL 0.05,= × ( )α= ∞ Where xi = production parameters, viz., AV, VMA, VFA, AC, Gmm, Gsb, %Passing 0.075 mm Figures 6-1 through 6-4 compare the tolerances developed from the mixtures in this study and current state agencies’ tol- erance values. The figures show that many states allow devia- tions between the submitted JMF and values reported during production that are higher than the tolerances developed in this study. These findings indicate that, the within-laboratory test- ing (single-operator and same equipment) variation is less than the between-laboratory testing tolerance. Based on these find- ings, it may be reasonable for states to review their current toler- ance values and to determine if a reduction in tolerance from design to production is warranted where the design laboratory is also the QC laboratory. Many of states determine asphalt binder content by means of ignition oven. Therefore, the toler- ance developed by solvent extraction would need to be further evaluated for comparison to the ignition method. Typically, solvent extraction results in a lower standard deviation when compared to ignition. Consequently, the tolerance for ignition would be slightly higher. 6.1.2 Maximum Acceptable Difference In addition to single-operator tolerance values, the com- bined data were used to evaluate a range of acceptable differ- ences (d2s) for each parameter. The range was determined in accordance with ASTM C670, “Standard Practice for Prepar- ing Precision and Bias Statements for Test Methods for Con- struction Materials.” As stated in ASTM C670, the maximum acceptable range is a function of the standard deviation of the test parameter and the number of specimens tested. Table 6-2 presents the table in ASTM C670 that is used to determine the multiplier to compute the acceptable range. Equation 22 presents how the values from ASTM C670 were used to gen- erate acceptable ranges for the properties evaluated. Agen- cies may use these findings to evaluate current specifications. These values may be higher than specified agency maximum allowable differences because the data in this study were gen- erated from multiple regions of the country. State agency tol- erance values should be developed using local data. Maximum Acceptable Range, x Standard Deviation, x 3.3 (22) i,LL-PL * , iDelta, LL-PL* LL PF or PL PF = × − − Where xi = production parameters, viz., AV, VMA, VFA, AC, Gmm, Gsb, % Passing 0.075 mm Proposed Guidelines for Recommended Practice

61 Table 6-1. Single-operator tolerance. Comparison Property Avg. Min Max Confidence Limit (Tolerance) Design (LL) - Producon (PL) Air Voids,% 0.6 0.0 1.3 0.8 VMA,% 0.4 0.0 2.1 1.2 VFA,% 4.0 0.3 9.9 5.4 Asphalt Binder Content,% 0.2 0.0 0.4 0.2 Gmm 0.014 0.002 0.039 0.020 Gsb 0.011 0.002 0.025 0.014 Passing 0.075 mm, % 0.4 0.0 0.9 0.5 Design (LL) - Construc„on (PF) Asphalt Binder Content,% 0.2 0.0 0.3 0.2 Gmm 0.011 0.000 0.020 0.013 Gsb 0.010 0.001 0.033 0.019 Passing 0.075 mm, % 0.7 0.1 1.3 0.7 Produc„on (PL) - Construc„on (PF) Asphalt Binder Content,% 0.1 0.0 0.4 0.2 Gmm 0.009 0.001 0.027 0.018 Gsb 0.008 0.000 0.031 0.017 Passing 0.075 mm, % 0.5 0.1 0.8 0.5 Figure 6-1. Tolerance comparison—asphalt binder content. Above Tolerance At Tolerance Below Tolerance Not Specified Above Tolerance At Tolerance Below Tolerance Not Specified Figure 6-2. Tolerance comparison—air voids, N design.

62 Figures 6-5 through 6-8 compare the tolerances developed from the mixtures in this study with current state tolerance values. The figures show that many states allow deviations between the submitted JMF and values reported during pro- duction, which are within the maximum allowable deviations observed in this study. 6.1.3 Development of Conversion Factors for Mechanical Comparison As agencies move toward developing performance-related specifications (PRS), it will be beneficial to develop a relation- ship between mechanical tests among the different specimen types (design, production, and construction). To start this pro- cess, the average values of the mechanical property for each specimen type were divided by the average of the same prop- erty of another specimen type, as described by Equation 23 for the LWT rut depths for the 1WI mixture. The resulting conversion factor may be used to convert the data developed from a design specimen (LL) to produce results closer to those expected for the production (PL) or construction (PF) values. Average Rut Depth, Rut Depth, Rut Depth Conversion Factor, (23) LL, 1WI PL, 1WI LL PL,1WI= At Tolerance Below Tolerance Not Specified Above Tolerance Figure 6-3. Tolerance comparison—Gmm, design vs. production. At Tolerance Below Tolerance Not Specified Above Tolerance Figure 6-4. Tolerance comparison—VMA, design vs. production. Comparison Property Maximum Acceptable Range Design (LL) - Producon (PL) Air Voids,% ± 1.3 VMA,% ± 2.0 VFA,% ± 9.1 Design (LL) - Producon (PL) Asphalt Binder Content,% ± 0.30 Design (LL) - Construcon (PF) ± 0.30 Producon (PL) - Construcon (PF) ± 0.30 Design (LL) - Producon (PL) Gmm ± 0.034 Design (LL) - Construcon (PF) ± 0.022 Producon (PL) - Construcon (PF) ± 0.030 Design (LL) - Producon (PL) Gsb ± 0.024 Design (LL) - Construcon (PF) ± 0.032 Producon (PL) - Construcon (PF) ± 0.029 Design (LL) - Producon (PL) Passing 0.075 mm, % ± 0.80 Design (LL) - Construcon (PF) ± 1.2 Producon (PL) - Construcon (PF) ± 0.80 Table 6-2. Maximum acceptable range.

63 Conversion factors were developed for each of the 10 mix- tures evaluated in this project. Table 6-3 presents the con- version factors developed from the LWT test data. The table shows that the average conversion factor between the design (LL) and production (PL) results is 1.0. Thus, on average, the rut depths observed from design specimens (mixed and compacted in the laboratory, LL) were similar to those of the production samples (those produced in the asphalt plant and compacted in the laboratory, PL). Conversely, the table shows an average conversion factor of 0.75 for design (LL) vs. construction (PF) and production (PL) vs. construction (PF). This indicates that, on average, the field-compacted (PF) specimens had a 33% higher rut depth than laboratory- compacted (LL and PL) samples. This relation ship is observed throughout the mechanical evaluation among the speci- men types and is attributed to differences in compaction effort between laboratory-compacted and field-compacted specimens. Table 6-4 presents the conversion factor analysis for axial dynamic modulus. Given specimen size constraints, only laboratory-compacted specimens (LL, PL) were available for evaluation in this case. The analysis shows that, on average, the conversion factor between design and production spec- imens is close to one at low and intermediate temperatures. As the temperature increases, the differences in dynamic modulus become more pronounced. The conversion factor in the high-temperature region indicates that the LL speci- mens have a lower modulus value than that of PL speci- mens. This may be attributed to binder oxidation during production. At Tolerance Below Tolerance Not Specified Above Tolerance Figure 6-5. Maximum range comparison—asphalt binder content. Above Tolerance At Tolerance Below Tolerance Not Specified Figure 6-6. Maximum range comparison—air voids, N design. Above Tolerance At Tolerance Below Tolerance Not Specified Figure 6-7. Maximum range comparison—Gmm, LL vs. PL. Above Tolerance At Tolerance Below Tolerance Not Specified Figure 6-8. Maximum range comparison—VMA.

64 Table 6-5 presents the results of the conversion factor analysis for IDT dynamic modulus. As shown in this table, no conversion in the modulus data among the specimen types is required in the low-temperature region, nor is a conversion factor required at any temperature between laboratory-compacted specimens. However, a conversion factor is required between the modulus values of field- and laboratory-compacted specimens at intermediate and high temperatures. Comparison No. ofPasses Average Conversion Conversion Range Min Max Design (LL)/ Production (PL) 1000 1.0 0.5 1.6 5000 1.0 0.6 1.6 10000 1.0 0.6 1.9 15000 0.8 0.6 1.1 20000 0.8 0.5 1.2 Average 1.0 Design (LL)/ Construction (PF) 1000 0.8 0.4 1.2 5000 0.8 0.3 1.2 10000 0.7 0.3 1.1 15000 0.7 0.3 1.1 20000 0.7 0.2 1.1 Average 0.75 Production (PL)/ Construction (PF) 1000 0.8 0.3 1.0 5000 0.7 0.3 1.1 10000 0.7 0.3 1.1 15000 0.7 0.2 1.2 20000 0.9 0.2 1.4 Average 0.75 Table 6-3. LWT conversion factor. Table 6-4. Axial dynamic modulus conversion factor. Comparison Temperature, °C Average Conversion Conversion Range Min Max Design (LL)/ Produc
on (PL) -10.0 1.0 0.7 1.1 4.4 1.0 0.7 1.1 25.0 0.9 0.6 1.1 37.8 0.8 0.5 1.1 54.4 0.8 0.5 1.2 Figure 6-9 compares master curves developed from design and construction specimens tested using IDT dynamic modulus. The figure shows that the curves are similar at low temperature and then diverge in the intermediate- and high- temperature regions. Figure 6-10 presents the results of the converted master curve. The conversion factors presented in Table 6-5 were applied to the intermediate- and high-temperature modulus values prior to the development of the master curve. As shown in the figure, the resulting converted construction master curve closely matches the design master curve. This conver- sion may be useful predicting distresses with programs such as Pavement ME Design. 6.2 Effect of Variability on Performance Effect of construction variability on predicted performance was quantified. Results of dynamic modulus testing from LL, PL, and PF specimens were used as the material input into mechanistic-empirical (ME) design models to evaluate the effect of specimen type on the predicted performance of pavement structures for varying traffic conditions (i.e., low, medium, and high). Pavement ME Design was used as a tool Temperature, °C Comparison Average Conversion Conversion Range Min MAx -10 Design (LL)/Producon (PL) 1.0 0.8 1.1 Design (LL)/Construcon (PF) 1.0 0.9 1.3 Producon (PL)/ Construcon (PF) 1.1 0.9 1.4 10 Design (LL)/Producon (PL) 0.9 0.8 1.1 Design (LL)/Construcon (PF) 1.2 0.8 1.5 Producon (PL)/ Construcon (PF) 1.3 0.9 1.7 25-35 Design (LL)/Producon (PL) 1.0 0.6 1.4 Design (LL)/Construcon (PF) 1.4 0.9 2.1 Producon (PL)/ Construcon (PF) 1.5 0.8 2.2 Table 6-5. IDT dynamic modulus conversion factor.

65 1000 10000 100000 1000000 10000000 1E-08 0.000001 0.0001 0.01 1 100 10000 D yn am ic M od ul us , P SI Reduced Frequency, Hz Design (LL) Construction (PF) Figure 6-9. Master curve comparison. 1000 10000 100000 1000000 10000000 1E-08 0.000001 0.0001 0.01 1 100 10000 D yn am ic M od ul us , P SI Reduced Frequency, Hz Design (LL) Construction (PF) Converted Construction (PF) Figure 6-10. Converted master curve comparison. to predict pavement performance. Previous research shows variability in the dynamic complex modulus of 10% or less resulted in a change in the predicted level of performance of 10% or less. However, variability in the dynamic modulus of 20% changed the design life of the pavement structures by up to 42%, and the design HMA thickness was affected by as much as 19% (Mohammad et al. 2012). Figure 6-11 presents the results of the effects of specimen type on the Pavement ME Design predictions of common pavement distresses. The figure shows that performance pre- diction was affected by specimen type. In general, the largest difference observed was for production versus construction specimens. Design versus production comparisons resulted in the least difference. These findings further illustrate how laboratory compaction results in a particle orientation differ- ent from that of field compaction. Rutting in the asphalt layer was the most influenced distress. This was expected given the differences observed in the modulus of the specimens at high temperature. Total rutting was less affected than AC rutting due to the common influences of base and subgrade rutting. Alligator cracking showed a difference as high as 60% between production and construction specimens. The predicted IRI was the performance parameter least influenced by the change in specimen type. Table 6-6 summarizes the percentage difference of distress predictions among specimen types. The range of percent- ages was developed by determining the percentage difference among the specimen types for each mixture and evaluating the minimum and maximum difference for each distress. As shown in Table 6-6, the use of design (LL) or production (PL) moduli would result in significant differences in pavement performance prediction as compared to construction (PF)

66 (a) Low Traffic 0 5 10 15 20 25 30 35 40 45 Alligator Cracking (%) Total Rutting (in) AC Rutting (in) IRI (in/mi) Pe rc en ta ge C ha ng e Performance Prediction LL vs. PF LL vs. PL PL vs. PF (b) Low (NC) 0 20 40 60 80 100 120 Alligator Cracking (%) Total Rutting (in) AC Rutting (in) IRI (in/mi) Pe rc en ta ge C ha ng e Performance Prediction LL vs. PF LL vs. PL PL vs. PF (c) Medium Traffic 0 10 20 30 40 50 60 70 80 90 Alligator Cracking (%) Total Rutting (in) AC Rutting (in) IRI (in/mi) Pe rc en ta ge C ha ng e Performance Prediction LL vs. PF LL vs. PL PL vs. PF (d) High Traffic 0 10 20 30 40 50 60 70 Alligator Cracking (%) Total Rutting (in) AC Rutting (in) IRI (in/mi) Pe rc en ta ge C ha ng e Performance Prediction LL vs. PF LL vs. PL PL vs. PF Figure 6-11. Average performance impact. moduli. The “true” in-service prediction should be based on plant-produced field-compacted specimens (i.e., core) because they represent the final product after production and compaction. However, regular extraction of cores from the installed pavement may be challenging. Results of this analysis indicate that pavement performance predictions obtained from dynamic moduli measured for dif- ferent specimen types would not be equivalent without the use of proper conversion factors to account for differences in production and compaction between specimen types. Further evaluation of these factors is needed before using the devel- oped conversion factors in the design process. The current Pavement ME Design prediction models were largely cali- brated with the properties of plant-produced specimens from Distress Comparison Range of Percent Difference Alligator Cracking LL vs. PL 9 - 44 LL vs. PF 11 - 30 PL vs. PF 13- 67 Asphalt Layer Rutng LL vs. PL 21 - 63 LL vs. PF 27 - 62 PL vs. PF 42 - 114 IRI LL vs. PL 2 - 5 LL vs. PF 3 - 8 PL vs. PF 4 - 11 Table 6-6. Effect of specimen types on pavement prediction.

67 LTPP General Pavement Study (GPS) sections, which would account for these differences. The following findings reflect the results of the perfor- mance prediction analysis: • Specimens prepared in the field and in the laboratory exhibited large and significant differences in performance prediction, especially between laboratory-compacted and field-compacted specimens. This finding is attributed to the differences in the compaction efforts and procedures between the field and the laboratory. Current Pavement ME Design prediction models were largely calibrated with the properties of plant-produced specimens from LTPP GPS sections, which would account for these differences. • Results of the Pavement ME Design analysis showed that the performance predictions are affected by specimen type. Rutting in the asphalt layer was the most influenced distress. Further, alligator cracking showed a difference as high as 60% between production and construction specimens.