International Journal of Pavement Engineering ISSN: 1029-8436 (Print) 1477-268X (Online) Journal homepage: https://www.tandfonline.com/loi/gpav20 Risk and expected pay factor analysis for assessing gap and dense-graded Superpave mixture specifications Sahand Sasha Karimi, Dimitrios G. Goulias & Charles W. Schwartz To cite this article: Sahand Sasha Karimi, Dimitrios G. Goulias & Charles W. Schwartz (2015) Risk and expected pay factor analysis for assessing gap and dense-graded Superpave mixture specifications, International Journal of Pavement Engineering, 16:1, 69-79, DOI: 10.1080/10298436.2014.916404 To link to this article: https://doi.org/10.1080/10298436.2014.916404 Published online: 13 May 2014. Submit your article to this journal Article views: 170 View related articles View Crossmark data Citing articles: 2 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=gpav20 International Journal of Pavement Engineering, 2015 Vol. 16, No. 1, 69–79, http://dx.doi.org/10.1080/10298436.2014.916404 Risk and expected pay factor analysis for assessing gap and dense-graded Superpave mixture specifications Sahand Sasha Karimi1, Dimitrios G. Goulias* and Charles W. Schwartz2 Department of Civil and Environmental Engineering, University of Maryland, College Park, MD 20742, USA (Received 31 October 2013; accepted 4 November 2013) With the implementation of the Superpave mix design method, some state highway agencies have experienced significant problems in durability of hot mix asphalt (HMA) mixtures due to lower binder content. With the adoption of revised HMA specifications for the state of Maryland, it was desired to evaluate the potential risks to both agency and contractors. This was achieved by calculating the a and b risks through the construction of operating characteristic curves. Furthermore, simulation analyses were conducted for assessing the impact of the current HMA production quality on the specification tolerances percent within specification limits and mixture pay factor. The findings of this study are based on a large set of quality control and quality acceptance data since the analyses were based on 7 consecutive years of production and paving of Superpave mixtures in the state. The risk analysis indicated that the agency bears a significant risk on accepting low-quality material, and thus changes in the current State Highway Administration specification are needed and discussed in this paper. The simulation analysis illustrated that the correlation among mixture parameters had minimal impact on pay factors. Furthermore, it was concluded that the asphalt content has a more pronounced impact on the pay factor than any other mixture property. Overall, the analysis indicated that significant changes in the acceptance specifications are needed in order to reduce the risks of accepting lower quality, or rejecting good quality, HMA mixtures. Furthermore, changes of the order of 10%, 20% even 50% in specification tolerances represent feasible and currently achievable levels of production quality by the paving industry, and thus no increase in the production cost of HMA is expected. Such change in the specification tolerances is expected to promote an overall improvement in quality of Superpave mixtures in the state and the region. The approach used in this study can be used by highway agencies and the pavement industry to evaluate risks related to the acceptance of hot asphalt mixtures. The pay factor analysis can aid in adjusting pay schedules and specification tolerances for accepting higher quality materials. Keywords: risk analysis; Superpave specifications; operating characteristic curves; pay factors; asphalt mixtures Introduction With the implementation of Superpave mix design for hot mix asphalt (HMA), the Maryland State Highway Administration (MSHA) has experienced a reduction in binder content for asphalt mixtures. These drier mixtures present significant challenges in terms of compaction that may lead to lower density and higher permeability, lower film thickness around the aggregate particles and increased potential for premature failure (NCHRP Projects 9-09, 925, 9-31, 9-33). MSHA’s concern with the lower asphalt levels in Superpave mixes has led efforts to explore strategies to increase the asphalt content (AC) in Superpave mixes (Goulias et al. 2009). The first phase of the study had as an objective to examine HMA mixture properties at the plant and behind the paver and assess differences in the quality control (QC) data and the acceptance data (Karimi et al. 2011). This second phase of the study had as its objective the development of operating characteristic (OC) curves in order to evaluate the risks to both the contractor and the agency. There are generally two types of acceptance plans: *Corresponding author. Email: dgoulias@umd.edu q 2014 Taylor & Francis (1) accept/reject acceptance plans and (2) acceptance plans that include pay adjustment provisions. The development of OC curves and the definitions of risks are described next. Even though OC analysis has been used in the past in a variety of fields for evaluating risks, in highway construction the benefits of this approach for developing statistically sound and defensible specifications was only recently recognised (Burati et al. 2011). Furthermore, a national effort is currently ongoing on incorporating such analysis into the acceptance of a variety of structural materials (NCHRP 10-92). In this paper, the results from the premium MSHA stone matrix asphalt (SMA) mixtures (i.e. gap graded) are presented. Parameters related to the construction of OC curves and calculation of risks The following parameters are typically used in the construction of OC curves and the calculation of risks (Weed 1996, Burati 2005, 2006, Burati and Weed 2006, 2011, TRB Circular E-C137 2009): 70 S.S. Karimi et al. . Acceptable quality level (AQL) represents the . . . . minimum quality level for fully acceptable material. When quality is based on percent within specification limits (PWSL), the AQL is the PWSL value for which the material is considered fully acceptable. For example, a 90% PWSL is commonly used for AQL for HMA. The developed acceptance plans should in this case be designed so that AQL material will receive an expected pay (EP) of 100%. Rejectable quality level (RQL) represents the level of quality below which the material is considered unacceptable (rejectable). When quality is based on PWSL, the RQL is the PWSL value for which the material is rejectable. For example, a 40% PWSL is commonly used for RQL for HMA. OC curve represents the relationship between the actual quality of a lot (e.g. PWSL), and either (1) the probability of its acceptance (for accept/reject acceptance plans) or (2) the probability of its acceptance at various payment levels (for acceptance plans that include pay adjustment provisions). Seller’s risk, a (this risk is also referred to as Type I error) is the probability of rejecting an AQL material. It is the risk that a producer takes in having AQL material rejected. Buyer’s risk, b (this risk is also called the Type II error) is the probability of accepting a lower quality (RQL) material. It is the risk that the highway agency takes in having RQL material accepted. Construction of OC curves and calculation of types I and II errors Assessing the current conditions The following four mixture properties are used by MSHA for determining mixture acceptance and pay factors: aggregate passing 0.075 mm/No. 200 sieve (0.075 mm); aggregate passing 2.36 mm/No. 8 sieve (2.36 mm); aggregate passing 4.75 mm/No. 4 sieve (4.75 mm) and AC. HMA acceptance specifications typically include more than just the AC because the influence of the aggregate properties has an important role in the volumetric characteristics of the Superpave mixtures and their performance. The impact of AC and aggregate properties on the volumetric characteristics of Superpave mixtures (such as air voids and volume in the mineral aggregate) has been examined in other studies (NCHRP Projects 9-09, 9-25, 9-31, 9-33) and is beyond the scope of this study. The specifications define a lot as a maximum of 6000 tons and a sublot as a maximum of 1000 tons of production. A minimum of one set of quality acceptance tests and one set of QC tests is required for each sublot, where each set of tests consists of measurement of the above mixture properties. Contractors often elect to perform more than the minimum number of QC tests, either on plant samples or on supplementary samples obtained from behind the paver. The specification tolerances for these four mixture properties from the target value for each project are shown in Table 1. To conduct the OC curve analysis and identify the a and b risks, the population distribution for each mixture property (i.e. 0.075 mm, 2.36 mm, 4.75 mm and AC) was determined using the quality acceptance data. Based on these data, the ‘interquartile range’ normality test was used, and the results shown that all four mixture properties follow a normal distribution. As previously mentioned, the results for the gap-graded mixtures (SMA) are presented herein, since this mixture is considered the best mixture used in Maryland paving. Based on the population characteristics for these material properties, the OC curves shown in Figures 1– 4 were developed. The OC curves were plotted with varying sample size, n (with n representing the number of sublots considered). The OC curves were developed by using the standard error of the population in order to relate PWSL and probability of acceptance (Burati et al. 2011). As can be observed from the OC curves related to the 0.075 mm and 4.75 mm properties, the maximum PSWL that can be achieved given the specification tolerances and the current production mean and standard deviation is about 96% and 84%, respectively. The effects of changing the sample size n can be assessed from these OC figures. As shown in Table 1 and based on the current production quality, with an AQL ¼ 90% and RQL ¼ 40% and using six samples per lot, the agency has a probability of accepting poor quality material (b) ranging from 26.3% to 27.6% depending on the mixture property. On the other hand, the contractor has a very low (0.1%) or zero probability of having good quality material being rejected (a). Clearly, the probability of accepting low-quality material by the agency is significant, and changes in the current SHA specification are suggested. Similar effects have been reported in a past study (Villiers et al. 2003). Increasing the sample frequency for acceptance testing in order to adjust risks at acceptable levels may impact the financial resources of the agency and thus may not represent a feasible solution. Thus, the effects on risks have to be examined in conjunction to potential adjustments in Table 1. Risks based on AQL ¼ 90% and RQL ¼ 40% for n ¼ 6. Property Tolerance a at AQL ¼ 90% b at RQL ¼ 40% 0.075 mm 2.36 mm 4.75 mm AC ^2 ^5 ^5 ^ 0.5 0.0 0.1 0.0 0.1 26.5 26.8 27.6 26.3 71 International Journal of Pavement Engineering Gap Graded AC 100% 90% n=8 n=7 n=6 n=5 Probability of Acceptance 80% 70% 60% 50% 40% 30% 20% 10% 0% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% PWSL Figure 1. OC curve for AC of gap-graded mixtures. Gap Graded 0.075 mm 100% 90% n=9 n=8 n=7 n=6 Probability of Acceptance 80% 70% 60% 50% 40% 30% 20% 10% 0% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% PWSL Figure 2. OC curve for 0.075 mm of gap-graded mixtures. Gap Graded 2.36 mm 100% n=7 n=6 n=5 n=4 90% Probability of Acceptance 80% 70% 60% 50% 40% 30% 20% 10% 0% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% PWSL Figure 3. OC curve for 2.36 mm of gap-graded mixtures. improving the quality and homogeneity of the production process and eventually fine-tuning the specification tolerances. Both of these aspects were examined as well and presented next. Modifying AQL and RQL to adjust risks (a 5 1% and b 5 5%) Since the calculated a and b risks are far from the typical and recommended values of 1% and 5%, respectively, 72 S.S. Karimi et al. Gap Graded 4.75 mm 100% n=8 n=7 n=6 n=5 90% Probability of Acceptance 80% 70% 60% 50% 40% 30% 20% 10% 0% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% PWSL Figure 4. OC curve for 4.75 mm of gap-graded mixtures. used in practice (AASHTO R-9, 2009, Burati et al. 2011), new values of AQL and RQL may be identified. Table 2 provides the values of AQL and RQL that result in a and b risks of 1% and 5%, respectively, based on the OC curves presented previously. The calculated values of AQL and RQL shown in Table 2 indicate that if a ¼ 1% and b ¼ 5% levels of risks are desired, without any changes in production quality or specification tolerances, the agency would have to reduce the quality level considered for full acceptance (i.e. AQL) from 90% to 83.4% PWL, for example for the AC mixture property. Similarly, the level of quality that the material is considered rejectable (RQL) should be changed from 40% to 25.9% PWL for the AC property. Similar levels of AQL and RQL are shown in Table 2 for the remaining mixture properties. Alternatively, if the agency wants to keep the AQL and RQL to the typical values of 90% and 40% PWL, then the specification tolerances will need to be modified as discussed in the next section. Revising specification tolerances for a 5 1% and b 5 5% To identify the required change in acceptance tolerances for achieving a ¼ 1% and b ¼ 5% risks at AQL and RQL of 90% and 40%, respectively, the standard normal Z-values approach identified in the ‘FHWA Optimal Procedures for QA Specifications’ study (Burati et al. 2011) was used. The Z-values are calculated by considering the values of AQL and RQL of Table 2 from the target values of AQL Table 2. AQL and RQL for a ¼ 1% and b ¼ 5% (n ¼ 6). and RQL of 90% and 40%, and then multiplying them by the standard deviation of the distribution of each mixture parameter. Following this procedure, the new set of specification tolerances, shown in Table 3, was obtained. The important question is whether such tolerances represent realistic achievable levels of production by the paving industry. When comparing the current specification tolerances shown in Table 1 with the suggested tolerances of Table 3 for achieving a ¼ 1% and b ¼ 5%, it can be observed that overall HMA production quality must improve. Thus, contractors must produce mixtures with materials and properties that are closer to the target design values and with lower production variability (i.e. reducing the standard deviation). The tolerances should also consider the total testing variance which includes sampling and inherent material variability, and testing error. A review of the quality acceptance data indicated that such production quality is achievable by good contractors, usually performing work for large size contracts, while contractors bidding for smaller paving jobs usually deal with older plants and paving equipment, and thus have lower production uniformity. For example, the recorded values for production quality in the state of Maryland for several mid- to large-size projects (i.e. composed of 7– 25 sublots of 1000 tons per sublot) were in the order of ^ 0.05 deviation from target (i.e. delta mean) and 0.09 production standard deviation for AC, ^ 0.91 deviation from target (i.e. delta mean) and 1.81 production standard deviation for percent passing 4.75 mm, ^ 0.09 deviation from target (i.e. delta mean) and 0.83 production standard deviation Table 3. Revised specification tolerances for a ¼ 1% and b ¼ 5% at AQL of 90% and RQL of 40%. Property Tolerance AQL at a ¼ 1% RQL at b ¼ 5% Property Tolerance 0.075 mm 2.36 mm 4.75 mm AC ^2 ^5 ^5 ^ 0.5 82.9 82.9 75.6 83.4 25.0 25.1 25.0 25.9 0.075 mm 2.36 mm 4.75 mm AC ^ 0.9 ^ 1.2 ^ 2.9 ^ 0.15 73 International Journal of Pavement Engineering for percent passing 2.36 mm, and ^ 0.10 deviation from target (i.e. delta mean) and 0.31 production standard deviation for percent passing 0.075 mm. The standard deviation for the gap-graded mixture population versus the values reported for mid- to large-size projects for the four mixture properties is shown in Figure 5. Thus, it is expected that reducing the specification tolerances at the levels shown in Table 3 is feasible. Furthermore, such a change in specification tolerances is expected to actually promote an overall improvement in mixture quality for the state. For smaller size projects, higher production variability is expected since there is limited time to make any production quality adjustments during the short time the plant is up and running. Since the big cost implications and risks to the highway agencies are primarily from midto large-size projects, the revised tolerances could be applicable to these projects (i.e. with 7 –25 sublots or more and with 1000 tons/sublot), while the original tolerances can be used for smaller size projects. As mentioned earlier, increasing the acceptance testing sample frequency for reducing risks may represent a significant increase in financial resources for the highway agency. Thus, alternatively to just acting on the specification tolerances alone as discussed previously, the agency may reduce acceptance risks by acting on both the tolerances (i.e. improving production homogeneity by the contractors) and eventually prescribing alternative levels of AQL and RQL. The a and b risk analysis and OC calculations provided an initial assessment of the risks involved with the current specifications of these HMA mixtures. These analyses are based on the risks assessed for each individual mixture property rather than providing an assessment of the combined risk associated with all four mixture properties considered in the combined MSHA specification. However, calculating the risks associated with the combined effects of all four acceptance parameters together was not necessary since such approach is 7 Population Standard Deviation 6 Mid/Large Size Projects 5 4 primarily used for accept/reject plans. Since the MSHA specs include pavement adjustment provisions, the focus of the research was directed towards the EP calculations approach using simulation analysis. Simulation analysis and expected mixture pay factor The purpose of the simulation analysis was to (1) examine the impact of the current HMA production quality on the composite mixture PWSL (CMPWSL) and mixture pay factor (MF) and (2) assess the impact of alternative scenarios in terms of specification tolerances. The dense-graded HMA mixtures were considered in the simulation analysis because of the comparatively large amount of data available for these mixtures in the MSHA database. The simulation tool developed under this study considers the four HMA mixture parameters mentioned previously, as well as their correlations, for calculating the composite pay factor, CMPWSL, and the expected MF. These parameters are calculated in the MSHA specification according to Equations (1) and (2). As it can be observed from the MSHA specification, it is based on the percentage of asphalt and aggregate passing the 4.75 mm (#4), 2.36 mm (#8) and 0.075 mm (#200) sieves, and the contractor has the opportunity to achieve a 5% incentive, if CMPWSL exceeds 90%. ( MF ¼ 0:55 þ 0:5CMPWSL ð1Þ if CMPWSL , 40% MF ¼ 0 The CMPWSL is calculated by CMPWSL f 1PWSL1 þ f 2PWSL2 þ f 3PWSL3 þ f 4PWSL4 ð2Þ P ; ¼ f where PWSL1 is the asphalt content; PWSL2 is the aggregate passing 4.75 mm/# 4 sieve; PWSL3 the aggregate passing 2.36 mm/# 8 sieve and PWSL4 the aggregate passing 0.075 mm/# 200 sieve. The fi terms represent weighting factors: f1 is the asphalt content ¼ 62; f2 is the aggregate passing 4.75 mm/# 4 sieve ¼ 7; f3 is the aggregate passing 2.36 mm/# 8 sieve ¼ 7 and f4 the aggregate passing 0.075 mm/# 200 sieve ¼ 24. An example of the correlations between the four mix properties is shown in Table 4. Preliminary analyses have shown that the correlation effects of the four HMA mix 3 Table 4. Correlations between mix parameters for densegraded mixtures. 2 1 Property 0.075 mm 2.36 mm 4.75 mm AC 0.075 mm 2.36 mm 4.75 mm AC 1 0.338 0.208 0.242 0.338 1 0.562 0.261 0.208 0.562 1 0.305 0.242 0.261 0.305 1 0 0.075mm 2.36mm 4.75mm AC Figure 5. Standard deviations for gap-graded mixture population versus mid- to large-size (paving projects with 7 – 25 sublots of 1000 tons/sublot) projects. 74 S.S. Karimi et al. Table 5. Example of effect of correlation value on the average MF. Average CMPWSL Average MF Correlation 98.1 98.1 98.0 98.1 0.999 0.5 0.001 Population 0 –0.05 –0.1 –0.15 –0.2 –0.25 Delta_AC 86.2 86.2 86.0 86.2 –0.3 –0.35 Table 6. Population characteristics for dense-graded mixtures. Property Delta meana Standard deviation 0.075 mm 2.36 mm 4.75 mm AC 0.992 2 0.192 0.066 2 0.002 1.20 3.88 5.60 0.31 –0.4 –0.45 0 0.25 0.5 SD/SDpop 0.75 1 Figure 6. Effect of reduction in AC variability. 93 a Deviations from the target values. 92 Mean_CMPWSL 91 90 89 88 87 86 85 0 0.25 Figure 7. 0.5 SD/SDpop 0.75 1 Effect of reduction in AC variability on CMPWSL. 101.5 101 100.5 Mean_MF properties have little impact on the pay factor analysis. In fact, in the example of Table 5, the values of the correlations were varied from 0.001 to 0.999. As it can be seen, no effects on MF were observed, whether the material properties were highly correlated or not correlated at all (i.e. correlation between all material properties equal to 0.999 and 0.001, respectively), in relation to the observed material property correlations reported in Table 4. The analyses were based on Monte Carlo simulation algorithms developed for this study. Mean values and standard deviations for the specification variables were based on all dense-graded QA data, excluding projects with multiple target values. The statistical results for these data are shown in Table 6. The values in Table 6 represent the distribution characteristics of the deviations between the target (design) values for each project and the actual lot production values for these material properties. One of the benefits of such approach is that the distribution of the deviations is immediately evident. Furthermore, such approach takes into account the fact that paving projects may have difference target values for AC and aggregate gradation requirements. 100 99.5 99 98.5 98 97.5 0 0.25 0.5 0.75 1 SD/SDpop Effects of production variability and specifications tolerances Reducing AC variability The goal of this analysis was to examine how much a producer might be able to reduce the AC and still have an acceptable product, assuming that he/she can improve production control and thus reduce production variability (standard deviation). All the gradations (0.075, 2.36 and 4.75) were kept at the population characteristic values. The standard deviation of AC was progressively reduced to 75%, 50% and 25% of the population value. The results were plotted in Figure 6, for a constant MF of 97.5% representing the value obtained based on the current population characteristics. As shown in the figure, a contractor that is able to produce an HMA mixture with Figure 8. Effect of reduction in AC content variability on MF. Table 7. Effects of change in AC specification tolerance and impact on MF. % %Change %Change AC_Tol Change Mean_CM Mean_MF CM MF 1 0.75 0.6 0.55 0.5 0.45 0.4 0.25 100 50 20 10 0 2 10 2 20 2 50 Note: CM, CMPWSL. 92.4 91.5 89.3 88.0 86.2 83.6 80.7 66.8 101.2 100.8 99.7 99.0 98.1 96.8 95.4 88.4 7 6 4 2 0 23 26 2 23 3.1 2.7 1.6 0.9 0.0 2 1.3 2 2.8 2 10.0 75 International Journal of Pavement Engineering AC 10% Change in Mean_CMPWSL 5% 0% –5% –10% –15% –20% –25% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 50% 60% 70% 80% 90% 100% Change in Tolerance Figure 9. Effects of change in AC specification tolerance on CMPWSL. AC 4.0% Change in Mean_MF 2.0% 0.0% –2.0% –4.0% –6.0% –8.0% –10.0% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% Change in Tolerance Figure 10. Effects of change in AC specification tolerance on MF. 75% lower variability (0.25 SD/SDpop) than the current QA population variability can reduce the AC by 0.4% from the target value and receive the same MF. Considering that the current tolerance for AC is ^ 0.5%, this change in AC is significant. However, reductions in variability of 25% or perhaps 50% are more realistic, and Table 8. Effects of change in 0.075 specification tolerance and impact on MF. % % Change Change MF 0.075_Tol %Change Mean_CM Mean_MF CM 4 3 2.4 2.2 2 1.8 1.6 1 100 50 20 10 0 2 10 2 20 2 50 Note: CM, CMPWSL. 90.8 89.8 88.1 87.2 86.1 84.9 83.4 78.0 100.4 99.9 99.1 98.6 98.0 97.5 96.7 94.0 5.5 4.3 2.4 1.3 0.0 2 1.3 2 3.1 2 9.4 2.4 1.9 1.0 0.5 0.0 20.6 21.4 24.1 produce a reduction in AC of approximately 0.2% and 0.3%, respectively, from the target value. Next the effect of reducing production variability of AC on CMPWSL and MF was examined, and all remaining parameters (including population means for all mixture parameters and variances for the three gradation percent passing) were at the population characteristics. As shown in Figures 7 and 8, if a contractor reduces production variability by 75% (0.25 SD/SDpop) while aiming for the target AC content, it can increase its CMPWSL from 86% to about 93% and receive an MF of about 101% instead of 98% (corresponding at SD/SDpop ¼ 1). Modifying specification tolerances The next set of analyses examined the effects of specification limit (tolerance) changes on the average MF and CMPWSL. Based on the specification, the tolerance for AC is ^ 0.5%. All other tolerances were kept constant, and only the AC tolerance was varied. The results are shown in Table 7 and Figures 9 and 10. 76 S.S. Karimi et al. 0.075 8.0% Change in Mean_CMPWSL 6.0% 4.0% 2.0% 0.0% –2.0% –4.0% –6.0% –8.0% –10.0% –12.0% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 50% 60% 70% 80% 90% 100% Change in Tolerance Figure 11. Effects of change in 0.075 specification tolerance on CMPWSL. 0.075 3.0% 2.0% Change in Mean_MF 1.0% 0.0% –1.0% –2.0% –3.0% –4.0% –5.0% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% Change in Tolerance Figure 12. Effects of change in 0.075 specification tolerance on MF. A change in the tolerance of AC content of about 20% will result in a change of 4% in CMPWSL and 1.6% in MF. As discussed previously, the paving industry in the state is currently able to produce Superpave mixtures closer to the design ‘target’ values and with lower production variability (i.e. higher mixture homogeneity). Thus, changes of the order of 10%, 20% even 50% in tolerances represent feasible and currently achievable levels of quality by the paving industry, and thus no cost increase in production of HMA is expected. Furthermore, as pointed out previously, such change in tolerances is expected to promote an overall improvement in quality of Superpave mixtures in the state and the region. Similarly, the effects of changing the 0.075 mm percent passing specification tolerance were also examined. The current specification suggests a tolerance of ^ 2%. The results of varying the 0.075 mm percent passing tolerance while holding all other constant are shown in Table 8 and Figures 11 and 12. The effects of changing the 2.36 mm percent passing specification tolerance were then examined. The current specifications suggest a tolerance of ^ 5%. The results of varying the 2.36 mm percent passing tolerance while holding all other constant are shown in Table 9 and Figures 13 and 14. Table 9. Effects of change in 2.36 specification tolerance on MF. % % Change Change 2.36_Tol %Change Mean_CM Mean_MF CM MF 10 7.5 6 5.5 5 4.5 4 2.5 100 50 20 10 0 210 220 250 Note: CM, CMPWSL. 87.4 87.1 86.7 86.3 86.1 85.7 85.3 83.9 98.7 98.5 98.3 98.2 98.0 97.8 97.6 96.9 1.48 1.13 0.69 0.26 2 0.01 2 0.48 2 0.92 2 2.57 0.57 0.41 0.22 0.03 20.09 20.29 20.49 21.21 77 International Journal of Pavement Engineering 2.36 Change in Mean_CMPWSL 2% 1% 0% –1% –2% –3% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 70% 80% 90% 100% Change in Tolerance Figure 13. Effects of change in 2.36 specification tolerance on CMPWSL. 2.36 0.8% 0.6% Change in Mean_MF 0.4% 0.2% 0.0% –0.2% –0.4% –0.6% –0.8% –1.0% –1.2% –1.4% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% 50% 60% Change in Tolerance Figure 14. Effects of change in 2.36 specification tolerance on MF. Finally, the effects of changing the 4.75 mm percent passing specification tolerance were examined. The current specifications suggest a tolerance of ^ 7%. The results from varying the 4.75 mm percent passing tolerance while holding all others constant are shown in Table 10 and Figures 15 and 16. As it can be seen from these analyses, the change in AC content tolerance has the most significant effect on MF Table 10. MF. Effects of change in 4.75 specification tolerance on % % Change Change 4.75_Tol %Change Mean_CM Mean_MF CM MF 14 10.5 8.4 7.7 7 6.3 5.6 3.5 100 50 20 10 0 2 10 2 20 2 50 Note: CM, CMPWSL. 87.5 87.2 86.6 86.2 86.1 85.8 85.2 83.9 98.8 98.6 98.3 98.1 98.0 97.9 97.6 97.0 1.7 1.3 0.7 0.2 0.0 20.3 21.0 22.5 1 0.5 0.2 0.0 20.1 20.2 20.5 21 reflecting the heavy weight of the AC content in calculating the CMPWSL. It can also be observed that with the bonus provision of the specification, an MF more than 100% is achievable for certain conditions. Conclusions This study examined the potential implications on risks and pay factors of the current Superpave mixture specifications. The analysis of the study was based on a large set of QC and quality acceptance data since records from seven consecutive years of production and paving of Superpave mixtures were used. The research approach in this study can be used by highway agencies and the pavement industry to evaluate risks related to the acceptance of HMA. As shown in this study, the use of OC curves can aid in providing the desired levels of acceptance risks to the agency and producers by (1) selecting the required sample size to accept a lot, (2) setting appropriate AQL and RQL values and/or (3) modifying specification tolerances. Similarly, the pay factor analysis can assist in examining: (1) the impact of the current HMA production quality on the CMPWSL and MF and (2) assess the impact of alternative 78 S.S. Karimi et al. 4.75 2.0% Change in Mean_CMPWSL 1.5% 1.0% 0.5% 0.0% –0.5% –1.0% –1.5% –2.0% –2.5% –3.0% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 50% 60% 70% 80% 90% 100% Change in Tolerance Figure 15. Effects of Change in 4.75 specification tolerance on CMPWSL. 4.75 1.0% Change in Mean_MF 0.5% 0.0% –0.5% –1.0% –1.5% –50% –40% –30% –20% –10% 0% 10% 20% 30% 40% Change in Tolerance Figure 16. Effects of change in 4.75 specification tolerance on MF. scenarios in terms of specification tolerances. Thus, such analysis can aid in adjusting pay schedules for accepting higher quality materials and adjust specification tolerances based on PWSL and MF. The specific results and conclusions reflecting the HMA mixtures from the Maryland paving operations have shown that the probability of accepting low quality material by the agency is significant; thus, changes in the current SHA specification are needed. The risk can be reduced by increasing the sample size for accepting a production lot. However, such decision may have a significant impact in the overall quality assurance expenditures of the highway agency. Alternatively, the specification tolerances may be adjusted to reduce acceptance risks and provide reasonable pay factors. The pay factor analysis indicated that correlation among mixture parameters (i.e. AC content and percent passing on the specific sieves) had no effect on the pay factor. A contractor with tight control over the variability of mixture production can significantly reduce the AC content and still receive a reasonable pay factor. Also, due to the high weight of the AC content in the final composite pay factor equation, the effects of changing the AC tolerances has a more pronounced impact on the pay factor than any other mixture property. Overall, the analysis indicated that changes of the order of 10%, 20% even 50% in specification tolerances represent feasible and currently achievable levels of production quality by the paving industry; thus, no increase in the production cost of HMA is expected. Such change in the specification tolerances is expected to promote an overall improvement in the quality of Superpave mixtures in the state and the region. Notes 1. 2. Email: ssk@umd.edu Email: schwartz@umd.edu References AASHTO, 2009. Standard Recommended Practice for Acceptance Sampling Plans for Highway Construction. R9-09. Washington, DC: AASHTO. International Journal of Pavement Engineering Burati, J., 2005. Risks with multiple pay factor acceptance plans. Transportation Research Record No. 1907. Washington, DC: Transportation Research Board of the National Academies, 42 – 97. Burati, J., 2006. Evaluating specification limits. Transportation Research Record No. 1946. Washington, DC: Transportation Research Board of the National Academies, 92 – 98. Burati, J. and Weed, R., 2006. Accuracy and precision of typical quality measures. Transportation Research Record No. 1946. Washington, DC: Transportation Research Board of the National Academies, 82 – 91. Burati, J., Weed, R., Hughes, C., and Hill, H., 2011. Optimal procedures for quality assurance specifications. Report FHWA-RD-02-095. Washington, DC: Federal Highway Administration (FHWA). Goulias, D., Schwartz, C., Karimi, S., and Hughes, C., 2009. Increasing durability of hot mix asphalt pavements designed with the superpave system. Research Report MD-SP708B4E. Hanover, MD: Maryland Department of Transportation State Highway Administration. Office of Materials and Technology. Karimi, S., Goulias, D., and Schwartz, C., 2011. Evaluation of Superpave HMA mixture properties at the plant versus behind the paver: statistical comparison of QC and QA data. ASCE Journal of Transportation Engineering, 138 (7), 924– 932. 79 NCHRP, 1998. Project 9-09: refinement of the Superpave gyratory compaction procedure. Washington, DC: Transportation Research Board. NCHRP, 2006. Project 9-25: requirements for voids in mineral aggregate for Superpave mixtures. Washington, DC: Transportation Research Board. Report 567. NCHRP, 2006. Project 9-31: air void requirements for Superpave mix design. Washington, DC: Transportation Research Board. Report 567. NCHRP, 2010. Project 9-33: a mix design manual for hot mix asphalt. Washington, DC: Transportation Research Board. Report 673. NCHRP, 2013. Project 10-92: optimizing the risk and cost of materials QA programs. Washington, DC: Transportation Research Board. TRB, 2009. Circular E-C137: glossary of highway quality assurance terms. Washington, DC: TRB. Villiers, C., Mehta, Y., Lopp, G., Tia, M., and Roque, R., 2003. Evaluation of percent-within-limits-construction specification parameters. International Journal of Pavement Engineering, 221– 228. Weed, R.M., 1996. Quality assurance software for the personal computer. Publication No. FHWA-SA-96-026. Washington, DC: Federal Highway Administration.