Report No. K-TRAN: KSU-09-7 P1 FINAL REPORT December Daba Gedafa, Ph.D., P.E. Mustaque Hossain, Ph.D., P.E. Lon Ingram, P.E.

Transcription

1 Report No. K-TRAN: KSU-09-7 P1 FINAL REPORT December 2012 Review of Data in Construction Management System (CMS) and Quality Control and Quality Assurance (QC/QA) Databases to Improve Current Specifications for Superpave and Concrete Pavements in Kansas: Part 1 Daba Gedafa, Ph.D., P.E. Mustaque Hossain, Ph.D., P.E. Lon Ingram, P.E. Kansas State University Transportation Center A cooperative transportation research program between Kansas Department of Transportation, Kansas State University Transportation Center, and The University of Kansas

2 This page intentionally left blank. i

3 1 Report No. 2 Government Accession No. 3 Recipient Catalog No. K-TRAN: KSU-09-7 Part 1 4 Title and Subtitle 5 Report Date Review of Data in Construction Management System (CMS) and Quality Control and December 2012 Quality Assurance (QC/QA) Databases to Improve Current Specifications for Superpave and Concrete Pavements in Kansas: Part 1 6 Performing Organization Code 7 Author(s) Daba Gedafa, Ph.D., P.E.; Mustaque Hossain, Ph.D., P.E.; and Lon Ingram, P.E. 9 Performing Organization Name and Address Department of Civil Engineering Kansas State University Transportation Center 2118 Fiedler Hall Manhattan, Kansas Sponsoring Agency Name and Address Kansas Department of Transportation Bureau of Materials and Research 700 SW Harrison Street Topeka, Kansas Performing Organization Report No. 10 Work Unit No. (TRAIS) 11 Contract or Grant No. C Type of Report and Period Covered Final Report February 2009 August Sponsoring Agency Code RE Supplementary Notes For more information write to address in block 9. See also K-TRAN: KSU-09-7 Part 2 16 Abstract Statistical specifications for highway construction are usually part of a statistical quality control process. These specifications provide the means to measure the important quality control attributes and ensure their compliance. The pay adjustments, part of these specifications, reflect the amount of deduction or bonus and the optimized risk distributed between the owner and the contractor. The Kansas Department of Transportation (KDOT) has built a comprehensive database of as-constructed properties of materials for Superpave pavements from the tests required as part of the Quality Control/Quality Assurance (QC/QA) program. Currently, KDOT pays incentives/disincentives for air voids and in-place density for Superpave pavements and thickness and strength for PCC pavements. A practical performance model and a composite index that include air voids, in-place density, asphalt content, and voids in mineral aggregate for Superpave pavements and thickness and strength for PCC pavements, respectively are needed to reflect the factors that affect their performance. The main objectives of this study were to investigate the effect of levels of significance and lot size, and to develop practical performance models and composite index for Superpave and PCC pavements in Kansas. Thirty-five Superpave pavements and 13 PCC projects from six administrative districts of KDOT were selected for this study. Lot-wise comparison showed that QC/QA means are significantly different in most cases. The number of cases with a significant difference in means increases with an increase in significance level. Practical performance models and composite index values from multiple quality characteristics have been proposed as integral parts of performance-related specifications (PRS) for Superpave and PCC pavements in Kansas. 17 Key Words QC/QA, Quality Control, Quality Assurance, Superpave, PCCP, Portland Cement Concrete Pavement 18 Distribution Statement No restrictions. This document is available to the public through the National Technical Information Service 19 Security Classification (of this report) Unclassified Form DOT F (8-72) 20 Security Classification (of this page) Unclassified 21 No. of pages Price ii

4 Review of Data in Construction Management System (CMS) and Quality Control and Quality Assurance (QC/QA) Databases to Improve Current Specifications for Superpave and Concrete Pavements in Kansas: Part 1 Final Report Prepared by Daba Gedafa, Ph.D., P.E. Mustaque Hossain, Ph.D., P.E. Lon Ingram, P.E. Kansas State University Transportation Center A Report on Research Sponsored by THE KANSAS DEPARTMENT OF TRANSPORTATION TOPEKA, KANSAS and KANSAS STATE UNIVERSITY TRANSPORTATION CENTER MANHATTAN, KANSAS December2012 Copyright 2012, Kansas Department of Transportation iii

5 PREFACE The Kansas Department of Transportation s (KDOT) Kansas Transportation Research and New- Developments (K-TRAN) Research Program funded this research project. It is an ongoing, cooperative and comprehensive research program addressing transportation needs of the state of Kansas utilizing academic and research resources from KDOT, Kansas State University and the University of Kansas. Transportation professionals in KDOT and the universities jointly develop the projects included in the research program. NOTICE The authors and the state of Kansas do not endorse products or manufacturers. Trade and manufacturers names appear herein solely because they are considered essential to the object of this report. This information is available in alternative accessible formats. To obtain an alternative format, contact the Office of Transportation Information, Kansas Department of Transportation, 700 SW Harrison, Topeka, Kansas or phone (785) (Voice) (TDD). DISCLAIMER The contents of this report reflect the views of the authors who are responsible for the facts and accuracy of the data presented herein. The contents do not necessarily reflect the views or the policies of the state of Kansas. This report does not constitute a standard, specification or regulation. iv

6 Abstract Statistical specifications for highway construction are usually part of a statistical quality control process. These specifications provide the means to measure the important quality control attributes and ensure their compliance. The pay adjustments, part of these specifications, reflect the amount of deduction or bonus and the optimized risk distributed between the owner and the contractor. The Kansas Department of Transportation (KDOT) has built a comprehensive database of as-constructed properties of materials for Superpave pavements from the tests required as part of the Quality Control/Quality Assurance (QC/QA) program. Currently, KDOT pays incentives/disincentives for air voids and in-place density for Superpave pavements and thickness and strength for PCC pavements. A practical performance model and a composite index that include air voids, in-place density, asphalt content, and voids in mineral aggregate for Superpave pavements and thickness and strength for PCC pavements, respectively are needed to reflect the factors that affect their performance. The main objectives of this study were to investigate the effect of levels of significance and lot size, and to develop practical performance models and composite index for Superpave and PCC pavements in Kansas. Thirty-five Superpave pavements and 13 PCC projects from six administrative districts of KDOT were selected for this study. Lot-wise comparison showed that QC/QA means are significantly different in most cases. The number of cases with a significant difference in means increases with an increase in significance level. Practical performance models and composite index values from multiple quality characteristics have been proposed as integral parts of performance-related specifications (PRS) for Superpave and PCC pavements in Kansas. v

7 Acknowledgements The authors would like to acknowledge the Kansas Department of Transportation for sponsoring this study under its Kansas Transportation and New Developments (K-TRAN) Program. Special thanks are due to Mr. Rick Barezinsky, Mr. Stephen Morris, and Mr. Bill Parcells of KDOT for providing QC/QA data for this study. vi

8 Table of Contents Abstract... v Acknowledgements... vi Table of Contents... vii List of Tables... x List of Figures... xii Chapter 1: Introduction General Problem Statement Objectives of the Study Organization of the Report... 3 Chapter 2: Data Analysis General Project Selection Data Collection Superpave Pavements Air Voids Density Smoothness PCC Control Charts Superpave Pavements QC Air Voids QC Density QA Density PCC Pavements QC Strength QC Thickness Comparison of QC and QA Density Mean Density Comparison Maximum Density Comparison Minimum Density Comparison Standard Deviation (STD) for Density Comparison Coefficient of Variation (COV) for Density Comparison vii

9 2.6 Comparison of Means Fisher s Least Significant Difference (LSD) Test Tukey s Honestly Significant Difference (HSD) Test Student-Newman-Keuls (SNK) Test Scheffe s Test Lot-Wise Means Comparison Superpave Pavements PCC Pavements Sublot-Wise Means Comparison Superpave Pavements PCC Pavements Sublot-Wise Comparison of Means Superpave Pavements PCC Pavements Effect of Significance Level on Pay Adjustments QC/QA Air Voids QC/QA Density F & t Tests Using Superpave Pavement Density Data Comparison of Design and Actual Asphalt Content Chapter 3: Practical Performance Model and Composite Index General Practical Performance Model Superpave Pavements Two Quality Characteristics Three Quality Characteristics Four Quality Characteristics Five Quality Characteristics PCC Pavements Two Quality Characteristics Three Quality Characteristics Pay Schedule Composite Index Superpave Pavements Two Quality Characteristics without Cross-Product Two Quality Characteristics with Cross-Product viii

10 Three Quality Characteristics without Cross-Product Three Quality Characteristics with Cross-Product Four Quality Characteristics without Cross-Product Four Quality Characteristics with Cross-Product Five Quality Characteristics with Cross-Product PCC Pavements Two Quality Characteristics without Cross-Product Two Quality Characteristics with Cross-Product Three Quality Characteristics without Cross-Product Three Quality Characteristics with Cross-Product Chapter 4: Conclusions and Recommendations Conclusions Recommendations References ix

11 List of Tables TABLE 2.1 Superpave Pavements Test Sections... 5 TABLE 2.2 PCC Pavements Test Sections... 6 TABLE 2.3 Summary of Control Charts for Superpave Pavements TABLE 2.4 Summary of Control Chart in PCC Pavements TABLE 2.5 Pay Adjustment for QC/QA Air Voids TABLE 2.6 Pay Adjustment for QC/QA Density TABLE 3.1 Data for PPM for Superpave Pavements (Two Variables, V=2) TABLE 3.2 Test of Derivation of PPM for Superpave Pavements (V=2 and C=1) TABLE 3.3 Test of Extremes of PPM for Superpave Pavements (V=2 and C=1) TABLE 3.4 Test of Offsetting Property of PPM for Superpave Pavements (V=2 and C=1) TABLE 3.5 Test of Offsetting Property of PPM for Superpave Pavements (V=2 and C=0.5) TABLE 3.6 Test of Extremes of PPM for Superpave Pavements (V=2 and C=0.5) TABLE 3.7 Test of Progressively Poorer Quality for Superpave Pavements (V=2 and C=0.5).. 69 TABLE 3.8 Data for PPM for Superpave Pavements (Three Variables, V=3) TABLE 3.9 Test of Derivation of PPM for Superpave Pavements (V=3 and C=1) TABLE 3.10 Test of Extremes of PPM for Superpave Pavements (V=3 and C=1) TABLE 3.11 Test of Offsetting Property of PPM for Superpave Pavements (V=3 and C=1) TABLE 3.12 Test of Offsetting Property of PPM for Superpave Pavement (V=3 and C=0.5) TABLE 3.13 Test of Extremes of PPM for Superpave Pavements (V=3 and C=0.5) TABLE 3.14 Test of Progressively Poorer Quality for Superpave Pavements (V=3 and C=0.5) 74 TABLE 3.15 Data for PPM for Superpave Pavements (Four Variables, V=4) TABLE 3.16 Test of Derivation of PPM for Superpave Pavements (V=4 and C=1) TABLE 3.17 Test of Extremes of PPM for Superpave Pavements (V=4 and C=1) TABLE 3.18 Test of Offsetting Property of PPM for Superpave Pavements (V=4 and C=1) TABLE 3.19 Test of Offsetting Property of PPM for Superpave Pavements (V=4 and C=0.5).. 77 TABLE 3.20 Test of Extremes of PPM for Superpave Pavements (V=4 and C=0.5) TABLE 3.21 Test of Progressively Poorer Quality for Superpave Pavements (V=4 and C=0.5) 79 TABLE 3.22 Data for PPM for Superpave Pavements (Five Variables, V=5) TABLE 3.23 Test of Derivation of PPM for Superpave Pavements (V=5 and C=1) TABLE 3.24 Test of Extremes of PPM for Superpave Pavements (V=5 and C=1) x

12 TABLE 3.25 Test of Offsetting Property of PPM for Superpave Pavements (V=5 and C=0.5)..83 TABLE 3.26 Test of Offsetting Property of PPM for Superpave Pavements (V=5 and C=0.5) TABLE 3.27 Test of Extremes of PPM Models for Superpave Pavements (V=5 and C=0.5) TABLE 3.28 Test of Progressively Poorer Quality for Superpave Pavements (V=5 and C=0.5) TABLE 3.29 Data for PPM for PCC Pavements (Two Variables, V=2) TABLE 3.30 Test of Derivation of PPM for PCC Pavements (V=2 and C=1) TABLE 3.31 Test of Extremes of PPM for PCC Pavements (V=2 and C=1) TABLE 3.32 Test of Offsetting Property of PPM for PCC Pavements (V=2 and C=1) TABLE 3.33 Test of Offsetting Property of PPM for PCC Pavements (V=2 and C=0.5) TABLE 3.34 Test of Extremes of PPM for PCC Pavements (V=2 and C=0.5) TABLE 3.35 Test of Progressively Poorer Quality for PCC Pavements (V=2 and C=0.5) TABLE 3.36 Data for PPM for PCC Pavements (Three Variables, V=3) TABLE 3.37 Test of Derivation of PPM for PCC Pavements (V=3 and C=1) TABLE 3.38 Test of Extremes of PPM for PCC Pavements (V=3 and C=1) TABLE 3.39 Test of Offsetting Property of PPM for PCC Pavements (V=3 and C=1) TABLE 3.40 Test of Offsetting Property of PPM for PCC Pavements (V=3 and C=0.5) TABLE 3.41 Test of Extremes of PPM for PCC Pavements (V=3 and C=0.5) TABLE 3.42 Test of Progressively Poorer Quality for PCC Pavements (V=3 and C=0.5) TABLE 3.43 Data for Composite Index for Superpave Pavements (Variables, V=2) TABLE 3.44 Data for Composite Index for Superpave Pavements (Variables, V=3) TABLE 3.45 Data for Composite Index for PCC Pavements (Two Variables, V=2) TABLE 3.46 Data for Composite Index for PCC Pavements (Variables, V=3) xi

13 List of Figures FIGURE 2.1 Typical Moving Average Control Charts for QC Air Voids in District FIGURE 2.2 Typical Moving Average Control Charts for QC Air Voids in District FIGURE 2.3 Typical Moving Average Control Charts for QC Air Voids in District FIGURE 2.4 Typical Moving Average Control Charts for QC Air Voids in District FIGURE 2.5 Typical Moving Average Control Charts for QC Air Voids in District FIGURE 2.6 Typical Moving Average Control Charts for QC Air Voids in District FIGURE 2.7 Moving Average Control Charts for QC Density in District FIGURE 2.8 Moving Average Control Charts for QC Density in District FIGURE 2.9 Typical Moving Average Control Charts for QC Density in District FIGURE 2.10 Typical Moving Average Control Charts for QC Density in District FIGURE 2.11 Typical Moving Average Control Charts for QA Density in District FIGURE 2.12 Moving Average Control Charts for QA Density in District FIGURE 2.13 Typical Moving Average Control Charts for QA Density in District FIGURE 2.14 Typical Moving Average Control Charts for QA Density in District FIGURE 2.15 Typical Moving Average Control Charts for QC PCC Strength in District FIGURE 2.16 Typical Moving Average Control Charts for QC PCC Strength in District FIGURE 2.17 Typical Moving Average Control Charts for QC PCC Strength in District FIGURE 2.18 Typical Moving Average Control Charts for QC PCC Strength in District FIGURE 2.19 Typical Moving Average Control Charts for QC PCC Thickness in District FIGURE 2.20 Moving Average Control Charts for QC PCC Thickness in District FIGURE 2.21 Typical Moving Average Control Charts for QC PCC Thickness in District FIGURE 2.22 Typical Moving Average Control Charts for QC PCC Thickness in District FIGURE 2.23 QC and QA Mean Density Comparison for US-81 in District FIGURE 2.24 QC and QA Mean Density Comparison for K-383 in District FIGURE 2.25 QC and QA Mean Density Comparison for US-54 in District FIGURE 2.26 QC and QA Mean Density Comparison for US-54 in District FIGURE 2.27 Summary of QC and QA Mean Density Comparison FIGURE 2.28 QC and QA Maximum Density Comparison for US-81 in District FIGURE 2.29 QC and QA Maximum Density Comparison for K-383 in District FIGURE 2.30 QC and QA Maximum Density Comparison for US-54 in District FIGURE 2.31 QC and QA Maximum Density Comparison for US-54 in District xii

14 FIGURE 2.32 Summary of QC and QA Maximum Density Comparison FIGURE 2.33 QC and QA Minimum Density Comparison for US-81 in District FIGURE 2.34 QC and QA Minimum Density Comparison for K-383 in District FIGURE 2.35 QC and QA Minimum Density Comparison for US-54 in District FIGURE 2.36 QC and QA Minimum Density Comparison for US-54 in District FIGURE 2.37 Summary of QC and QA Minimum Density Comparison FIGURE 2.38 QC and QA STD Density Comparison for US-81 in District FIGURE 2.39 QC and QA STD Density Comparison for K-383 in District FIGURE 2.40 QC and QA STD Density Comparison for US-54 in District FIGURE 2.41 QC and QA STD Density Comparison for US-54 in District FIGURE 2.42 Summary of QC and QA STD Density Comparison FIGURE 2.43 QC and QA COV Density Comparison for US-81 in District FIGURE 2.44 QC and QA COV Density Comparison for K-383 in District FIGURE 2.45 QC and QA COV Density Comparison for US-54 in District FIGURE 2.46 QC and QA COV Density Comparison for US-54 in District FIGURE 2.47 Summary of QC and QA COV density comparison FIGURE 2.48 Lot-Wise Means Comparison for QC Air Voids FIGURE 2.49 Lot-Wise Mean Comparison for QC Smoothness FIGURE 2.50 Lot-Wise Means Comparison for QC Density FIGURE 2.51 Lot-Wise Means Comparison for QA Density FIGURE 2.52 Lot-Wise Means Comparison for PCC Strength FIGURE 2.53 Lot-Wise Means Comparison for PCC Thickness FIGURE 2.54 Sublot-Wise Means Comparison for QC/QA Air Voids FIGURE 2.55 Sublot-Wise Means Comparison for QC Density FIGURE 2.56 Sublot-Wise Means Comparison for QA Density FIGURE 2.57 Sublot-Wise Means Comparison for QC/QA Density FIGURE 2.58 Sublot-Wise Means Comparison for QC PCC Strength FIGURE 2.59 Sublot-Wise Means Comparison for QC PCC Thickness FIGURE 2.60 Sublot-Wise Comparison of Mean Magnitude for QC/QA Air Void FIGURE 2.61 Sublot-Wise Comparison of Mean Magnitude for QA Density FIGURE 2.62 Sublot-Wise Comparison of Mean Magnitude for QC PCC Strength FIGURE 2.63 Sublot-Wise Comparison of Mean Magnitude for QC PCC Thickness FIGURE 2.64 Pay Adjustment for QC/QA Air Voids xiii

15 FIGURE 2.65 Pay Adjustment for QC/QA Density FIGURE 2.66 QC/QA Density Significant Difference Test FIGURE 2.67 Discrepancy of QC/QA Significant Difference Test for Equal and Unequal Variance FIGURE 2.68 Typical Comparisons of Design and Actual AC When Actual Is Lower Than Design FIGURE 2.69 Typical Comparison of Design and Actual AC When Actual is Lower Than Design FIGURE 2.70 Typical Comparison of Design and Actual AC When Actual is Lower and Higher Than Design FIGURE 2.71 Summary of Actual and Design AC Content Comparison xiv

16 Chapter 1: Introduction 1.1 General The history of highway quality assurance has progressed from the early materials and methods specifications through statistical end-result specifications to the current trend toward performance-related specifications (PRS) based on mathematical models and statistical concepts (Weed 2000). The impetus for use of statistical methods in managing highway construction quality began in 1963 with an initiative led by the Bureau of Public Roads (Benson et al. 2000). This initial effort resulted in development and implementation of Portland Cement Concrete (PCC) specifications in 1973, followed by their evaluation in 1979 (Diwan et al. 2003). These developments led to research efforts by a number of states to obtain information relative to the variability and quality level of the construction practices. This information on variability was then translated into specifications using standard statistical procedures for quality control and quality acceptance. The initial effort was finally completed by implementing these statistical specifications for highway materials and construction (Shah 1976). Many states have adopted statistical quality control/quality assurance (QC/QA) programs. The properties controlled under statistical QC/QA programs should be either related to performance or desirable end-results. These end-result specifications are usually based on statistics from historical construction data (Schmitt et al. 1998, Parker and Hossain 2002). Many agencies now also include bonus provisions that award payment somewhat in excess of the contract price when the quality level substantially exceeds the level that has been specified (NCHRP 1995, Weed 2002, Weed and Tabrizi 2005). One of the advantages of statistical specifications is the production of accurate data from valid random sampling procedures. This data may be analyzed later to improve the specifications further (Afferton et al. 1992). Some agencies are moving in the direction of PRS that specifies the desired levels of key construction quality characteristics that have been found to correlate with fundamental engineering properties which predict performance. When there are different types of tests to be performed on a particular construction item, it can become a complex matter to design an acceptance procedure that is fair, effective, and free from inconsistencies. Composite index avoids certain inconsistencies in practice that may occur with other methods for dealing with 1

17 multiple quality characteristics. It leads to rational pay schedules in that it assures that all combinations of individual quality measures that predict the same level of expected life will receive the same amount of pay adjustment (Weed 2006). 1.2 Problem Statement The Kansas Department of Transportation (KDOT) has built an impressive database of as-constructed materials properties for Superpave and Portland Cement Concrete (PCC) pavements from the tests required as part of the QC/QA program. KDOT also has a Construction Management System (CMS) that captures data on selected attributes related to highway construction in Kansas. Burati et al. (2004) have argued that any specification must also be an evolutionary process. Since new information is constantly becoming available in the form of additional test results, and as new construction or testing processes are employed, the specification must be continually monitored to see if improvements are needed. Thus a review of the current QC/QA specifications of KDOT is needed to find the opportunities for improvement. This need has also been echoed by the recent Federal Highway Administration (FHWA) QA Stewardship Review of KDOT with respect to use of a different payment lot size, review of acceptance of contractor s test results, changing level of significance for statistical testing, developing composite index, and practical performance model. 1.3 Objectives of the Study The main objectives of this study were to: Investigate any systematic bias in KDOT QC/QA data using moving average control chart analysis; Compare lot-and sublot-wise means and to investigate the possibility of changing lot size; Analyze the consequences of changing the level of significance from 1% to 2.5%; Determine the consequences of using the F-test along with the t-test to determine whether or not to use the test results from the contractor for acceptance; 2

18 Analyze actual and design asphalt contents to see if including asphalt content in pay adjustment is justifiable; and Develop practical performance models and composite index. 1.4 Organization of the Report This report includes four chapters. The first chapter deals with brief literature review, problem statement, and objectives of the study. Data analysis for both Superpave and PCC pavements is described in the second chapter. The third chapter includes practical performance models and composite index that include various quality characteristics for both Superpave and PCC pavements. The last chapter deals with conclusions and recommendations based on this study. 3

19 Chapter 2: Data Analysis 2.1 General This chapter deals with data analysis for Superpave and PCC pavements in Kansas. Control chart analysis for different quality characteristics, QC and QA density comparison, lotand sublot-wise comparison of means using four comparison methods at three significance levels, effect of significant levels on pay adjustments, feasibility of using F-test instead of t-test, and finally asphalt content of Superpave pavements constructed have been described in this chapter. 2.2 Project Selection Both Superpave and PCC pavements have been selected in this study. Thirty five Superpave pavements, built between 2004 and 2007, were selected based on total tonnage as shown in Table 2.1. The selected projects are such that multiple lots of 3,000 tons were produced and placed on these projects. These projects are from all six administrative districts of KDOT. The length of the projects varies from 1.92 miles to miles. The PCC pavements have been selected based on size as well. Thirteen PCC pavements were selected from four KDOT districts as shown in Table 2.2. Most of the PCC pavements are on interstate highways. 2.3 Data Collection Random sampling procedures were used to collect QC/QA data. It is well established that random sampling procedures avoid biases and lead to a more reliable estimate of the as-built construction quality (Weed 1989) Superpave Pavements Air voids, in-place density, asphalt content, smoothness, and voids in mineral aggregates (VMA) data have been used in this study. The following sections describe the ways these data have been collected. 4

20 Route County Name TABLE 2.1 Superpave Pavements Test Sections KDOT Fiscal Let Dist. Year Date No. No. of Lanes Length (mi) 1 U075 Brown /20/ K007 Doniphan /14/ U036 Doniphan /17/ U040 Douglas /19/ I135 McPherson /19/ I070 Saline /17/ U077 Marion /15/ I135 McPherson /15/ U081 Ottawa /16/ K156 Ellsworth /18/ U036 Jewell /15/ U283 Graham /18/ U283 Norton /18/ U024 Rooks /17/ U283 Trego /16/ U283 Trego /16/ K027 Sherman /15/ K027 Sherman /15/ U083 Sheridan /15/ K383 Decatur /21/ U160 Crawford /19/ I035 Coffey /20/ U077 Butler /11/ U077 Cowley /19/ U050 Edwards /12/ U050 Reno /19/ U054 Kingman /15/ K096 Barton /18/ K027 Greeley /08/ U050 Finney /14/ U056 Stevens /16/ U054 Seward /14/ K027 Stanton /14/ K096 Wichita /18/ U054 Seward /15/

21 TABLE 2.2 PCC Pavements Test Sections KDOT Route County Name District No. Let Date 1 I035 Osage 1 07/16/03 2 I035 Wyandotte 1 12/17/03 3 I035 Wyandotte 1 01/14/04 4 I035 Johnson 1 06/16/04 5 I035 Leavenworth 1 07/21/04 6 I070 Wyandotte 1 06/15/05 7 I070 Dickinson 2 07/21/04 8 U054 Bourbon 4 04/16/03 9 I035 Coffey 4-10 U069 Miami 4 12/17/03 11 U069 Bourbon 4 06/16/04 12 U054 Sedgwick 5 01/08/03 13 I135 Sedgwick 5 07/16/ Air Voids The normal lot-size for air voids is 3,000 tons. The lot is divided into four subolts of uniform size. KDOT specifies roadway sampling. Roadway samples are obtained for each sublot from behind the paver before compaction. A three-sided template is pushed into the mat prior to compaction. A square shovel is then used to extract all asphalt mixtures from the selected locations. The sample is obtained from a minimum of three locations randomly selected by KDOT personnel throughout one truck load of placed material. The selection process involves one random number for the sampled tonnage (truck load) and two random numbers for transverse and longitudinal locations (Elseifi et al. 2009). The samples are transported to the test facility using a method to retain heat to facilitate sample quartering procedures. Air voids tests are performed on Superpave gyratory-compacted samples of a given mix design. A lot normally consists of results of four contiguous results of individual QC and one QA. 6

22 Density KDOT considers the day s placement as a lot for density measurements. This lot is also subdivided into five uniform sublots. Random test locations are selected by the Contractor or the Engineer. Mat density is typically measured with nuclear density gages but can also be measured from cores. Contractor makes two and KDOT makes one independent mat density measurement for each sublot (2 to 1 sampling ratio) (Turochy and Parker 2007) Smoothness A California-type profilograph or if approved by the Bureau of Materials and Research of KDOT, other types of profilographs (such as a Light Weight Profiler) that produce results comparable to the California-type profilograph may be used. A 0.1-mile long sublot size is used. Only QC data is collected. Pay adjustment is based on pavement smoothness in terms of average profile index of the pavement section before any corrective work is performed. A zero blanking band is used for profilogram analysis PCC In Kansas, pay adjustments for pavement thickness and concrete compressive strength are based on test results from cores taken from each lot. All cores for determining strength shall be taken at a minimum of 21 days after the pavement has been placed, and in time to determine 28- day compressive strengths. For mainline and other pavements subject to coring for pay adjustments for both thickness and strength, a lot is defined as the surface area of mainline pavement lane placed in a single day. Normally, a lot representing a day's production is divided into five sublots of approximately equal surface area. For high daily production rates (rates exceeding 6,000 square yards per day), the contractor may choose to divide the day s production into two approximately equal lots consisting of five sublots each. Normally one core is taken per sublot (Khanum et al. 2006). Cores are transported to the laboratory as soon as possible and the thickness is measured at three points at approximately 120º apart. Then the 4-inch diameter cores are cured to be tested for 28-day compressive strength. 7

23 Air Voids (%) 2.4 Control Charts Microsoft Excel was used to calculate moving averages, average, lower, and upper limits (minus/plus three times standard deviation) for different quality characteristics. Typical control charts for Superpave and PCC pavements are presented Superpave Pavements Typical control charts for different quality characteristics from each district have been presented. Control charts for QC air voids, and QC/QA density have been discussed QC Air Voids Figure 2.1 shows moving average control chart for US-75 route in Brown County. It is typical for QC air voids in District 1. The moving average values are sometimes lower and higher than average value though the difference is not significant. All moving average values are within 3 where σ is the standard deviation : U75 (Brown)- PG 70-28, SM-9.5A, 1.5in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.1 Typical Moving Average Control Charts for QC Air Voids in District 1 8

24 Air Voids (%) Figure 2.2 shows moving average control chart for US-81 route in Ottawa County. It is typical for QC air voids in District 2. The moving average values are sometimes lower and higher than average value though the difference is not significant. All moving average values are within 3. Figure 2.3 shows moving average control chart for K-383 route in Decatur County. It is typical for QC air voids in District 3. The moving average values are slightly lower than the average for about half of the points and slightly higher than the average for another half of the points though the difference is not significant since all moving average values are within : U81 (Ottawa)-PG 70-28, SM-19A, 2.5in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.2 Typical Moving Average Control Charts for QC Air Voids in District 2 9

25 Air Voids (%) Air Voids (%) : K383 (Decatur)- PG 64-22, SR-19A, 2in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.3 Typical Moving Average Control Charts for QC Air Voids in District 3 Figure 2.4 shows moving average control chart for US-160 route in Crawford County. It is typical for QC air voids in District 4. The moving average values are mostly lower than average value though the difference is not significant. All moving average values are within : U160 (Crawford)-PG 64-22, SM-19A Moving Average Lower Average Upper No. of Moving Average FIGURE 2.4 Typical Moving Average Control Charts for QC Air Voids in District 4 10

26 Air Voids (%) Figure 2.5 shows moving average control chart for US-77 route in Butler County. It is typical for QC air voids in District 5. The moving average values are sometimes lower and higher than average value. All moving average values are within 3. Figure 2.6 shows moving average control chart for US-50 route in Finney County. It is typical for QC air voids in District 6. The moving average values are very close to the average value except at few points where they are slightly higher or lower than the average. All moving average values are within : U77 (Butler)-SM-19A Moving Average Lower Average Upper No. of Moving Average FIGURE 2.5 Typical Moving Average Control Charts for QC Air Voids in District 5 11

27 Air Voids (%) : U50 (Finney)-SM-19A, 11in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.6 Typical Moving Average Control Charts for QC Air Voids in District QC Density Districts 1 and 4 do not have complete QC density data for moving average control chart analysis i.e. one or more sublot density data is missing. All moving average values are within 3 except for K-383 route in Decatur County. Typical moving average control charts will be described for Districts 2, 3, 5 and 6. Moving average control chart for US-81 route is shown in Figure 2.7. This is the only QA density moving average control chart in District 2. Most of the moving averages are less than the average for all of the first 25% of data points whereas the reverse is true for the next 30% of the data points. The moving average for the remaining data points is close to the average. 12

28 Density (%) : U81 (Ottawa)- PG 58-28, SR-19A, 3.5 in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.7 Moving Average Control Charts for QC Density in District 2 Figure 2.8 shows the moving average control charts for K-383 in Decatur County. This is the only QC density moving average control chart which lies outside 3. This shows density at the beginning of the tests was very low compared to the rest. Except for the first few readings, the other moving average values are mostly equal to or greater than the average density. Figure 2.9 shows moving average control charts for US-54 route in Kingman County. This is typical for QC density moving average control chart in District 5. The moving averages are lower and higher than the average at certain interval, respectively. The trend may be due to an action taken by the contractor when density is low or high to keep it close to the average. 13

29 Density (%) Density (%) : K383 (Decatur)-PG 64-22, SR-19A, 2in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.8 Moving Average Control Charts for QC Density in District : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.9 Typical Moving Average Control Charts for QC Density in District 5 Moving average control charts for US-54 route in Seward County is shown in Figure This is typical for QC density moving average control chart in District 6. The moving 14

30 Density (%) averages are higher than the average for some of the data points at the beginning, but they are equal to or less than the average for the majority of points : U54 (Seward)- PG 70-22, SM-12.5A, 2in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.10 Typical Moving Average Control Charts for QC Density in District QA Density Districts 1 and 4 do not have complete QA density data for moving average control chart analysis i.e. one or more sublot density data is missing. All moving average values are within 3 except for K-383 route in Decatur County. Typical moving average control charts will be described for each District. Moving average control chart for US-81 route is shown in Figure This is typical QA density moving average control chart in District 2. Most of the moving averages are less than the average for most of the first 50% of data points whereas the reverse is true for the rest of data points though all moving averages are close to average from a practical point of view. Figure 2.12 shows moving average control chart for K-383 in Decatur Country. This is the only QA density moving average control chart which lies outside 3. This shows density at the beginning of the tests was very low compared to the rest. Except for the first few readings, 15

31 Density (%) Density (%) the other moving average values closer to the average value. The same trend was observed for QC density control chart for the same route in this county : U81 (Ottawa)-PG 58-28, SR-19A, 3.5in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.11 Typical Moving Average Control Charts for QA Density in District : K383 (Decatur)-PG 64-22, SR-19A, 2in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.12 Moving Average Control Charts for QA Density in District 3 16

32 Density (%) Figure 2.13 indicates moving average control chart for US-54 route in Kingman County. It is typical QA density control chart in District 5. Moving averages are lower than average for the first 30% of data points and higher than the average for the rest of the data points. Moving average control chart for US-54 in Seward County is shown in Figure It is typical QA density control chart for District 6. Moving average is sometimes lower and higher for about 60% of data points whereas it is lower than the average for rest of the data. Table 2.3 shows the summary of control chart analysis for Superpave pavements in different districts. The moving averages for air voids for all 49 QC cases are inside 3 where σ is the standard deviation. Density in one sublot is outside results do not clearly show any systematic bias in QC and QA data. 3 for both QC and QA. These : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.13 Typical Moving Average Control Charts for QA Density in District 5 17

33 Density (%) : U54 (Seward), PG 70-22, SM-12.5A, 2in. Moving Average Lower Average Upper No. of Moving Average FIGURE 2.14 Typical Moving Average Control Charts for QA Density in District 6 District TABLE 2.3 Summary of Control Charts for Superpave Pavements Superpave Pavements Density Air Voids Quality Control Quality Control Quality Assurance Total PCC Pavements Typical control charts for different quality characteristics from each district have been presented. Control charts for QC PCC strength and thickness have been discussed. Districts 3 and 6 do not have complete PCC strength and thickness data for moving average control charts. Typical moving average control charts are presented for the rest of the districts. 18

34 PCC Strength (MPa) QC Strength Figure 2.15 indicates moving average control chart for Interstate 35 (I-35) route in Osage County. The moving averages are slightly higher or lower than the average for about 50% of data points whereas they are higher than the average for rest of the data points : I35 (Osage) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.15 Typical Moving Average Control Charts for QC PCC Strength in District 1 Figure 2.16 shows moving average control chart for I-70 in Dickinson County. This is typical for QC PCC strength moving average control charts in District 2. The moving averages are mostly higher than the average for about 60% of the data points whereas they are mostly lower than the average for the rest of the data points. Moving average control chart for US-69 in Bourbon County is given in Figure It is typical for QC PCC strength control chart analysis in District 4. Moving averages are mostly less than the average for the first 45% of the data points. The moving averages for the remaining data points are sometimes higher and lower than the average value. 19

35 PCC Strength (MPa) PCC Strength (MPa) : I70 (Dickinson) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.16 Typical Moving Average Control Charts for QC PCC Strength in District : U69 (Bourbon) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.17 Typical Moving Average Control Charts for QC PCC Strength in District 4 20

36 PCC Strength (MPa) Figure 2.18 shows moving average control chart for US-54 route in Sedgwick County. It is the typical QC PCC moving average control chart in District 5. The moving averages are slightly higher or lower than the average, but it is very close to the average for all practical purposes : U54 (Sedgwick) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.18 Typical Moving Average Control Charts for QC PCC Strength in District QC Thickness All PCC thickness moving averages are within 3 except moving average control chart for I-70 route in Dickinson County. Figure 2.19 shows moving average control chart for I- 35 route in Osage County. This is typical for QC PCC thickness control chart in District 1. Moving averages are lower than the average for some of the data points and higher than the average for the rest. The moving averages are close to the average for last 12.5% of data points. Figure 2.20 shows moving average control chart for I-70 route in Dickinson County. This is the only QC PCC thickness moving average control chart which lies slightly outside the lower limit at two points. The remaining moving averages are very close to the average value. 21

37 PCC Thickness (in) PCC Thickness (in) : I35 (Osage) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.19 Typical Moving Average Control Charts for QC PCC Thickness in District : I70 (Dickinson) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.20 Moving Average Control Charts for QC PCC Thickness in District 2 22

38 PCC Thickness (in) Figure 2.21 shows moving average control chart for US-69 route in Bourbon County. It is the typical QC PCC thickness moving average control chart in District 4. The moving averages are very close to the average value for most of the points : U69 (Bourbon) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.21 Typical Moving Average Control Charts for QC PCC Thickness in District 4 Moving average control chart for US-54 in Sedgwick County is shown in Figure It is typical QC PCC thickness moving average control chart in District 5. The moving averages at some points are very close to the upper limit. Most of the moving averages are very close to the average value. Table 2.4 shows the summary of control chart analysis for PCC pavements in different districts. All moving averages for PCC QC data for thickness are within sublot for thickness. These results do not clearly show any systematic bias in QC data. 3 except for one 23

39 PCC Thickness (in) : U54 (Sedgwick) Moving Average Lower Average Upper No. of Moving Average FIGURE 2.22 Typical Moving Average Control Charts for QC PCC Thickness in District 5 TABLE 2.4 Summary of Control Chart in PCC Pavements District Quality Characteristics Strength Thickness Total Comparison of QC and QA Density QC and QA density comparison in terms of mean, minimum, maximum, standard deviation (STD), and coefficient of variation (COV) has been carried out in order to investigate whether QC density is consistently higher than the QA density. Districts 1 and 4 do not have complete data for QC and QA comparison. All the statistics were based on lots that include 10 QC data points and five QA data points. Projects with the highest number of lots were selected from the remaining districts. 24

40 Mean Density (%) Mean Density Comparison Figure 2.23 shows QC/QA mean density comparison for US-81 route in Ottawa County. It has the highest number of lots of all projects in District 2 in this study. QC mean density is higher than QA mean density in nine out of 12 lots, which is 75%. The mean difference is the highest in lot 5 and the smallest in lot 2, respectively. The mean difference may not be significant statistically : U81 (Ottawa)-PG 58-28, SR-19A, 3.5 in QC QA Lot Number FIGURE 2.23 QC and QA Mean Density Comparison for US-81 in District 2 Figure 2.24 shows QC/QA mean density comparison for K-383 route in Decatur County. It has the highest number of lots of all projects in District 3 in this study. QC mean density is higher than QA mean density in 13 out of 16 lots, which is about 81%. The mean difference is the highest in lot 3 and the smallest in lot 8, respectively. Lot 1 has the lowest QC/QA mean density. The mean difference may not be significant from a practical point of view. Figure 2.25 shows QC/QA mean density comparison for US-54 route in Kingman County. It has the highest number of lots of all projects in District 5 in this study. QC mean density is higher than QA mean density in six out of nine lots, which is about 67%. The mean difference is the highest in lot 2 and the smallest in lot 7, respectively. Lot 4 has the lowest QC 25

41 Mean Density (%) Mean Density (%) mean density whereas lot 1 has the lowest QA mean density. The mean difference may not be significant from a practical point of view : K383 (Decatur)-PG 64-22, SR-19A, 2 in QC QA Lot Number FIGURE 2.24 QC and QA Mean Density Comparison for K-383 in District : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5 in QC QA Lot Number FIGURE 2.25 QC and QA Mean Density Comparison for US-54 in District 5 26

42 Mean Density (%) Figure 2.26 shows QC/QA mean density comparison for US-54 route in Seward County. It has the highest number of lots out of all projects in District 6 in this study. There were no QA data for lots 1 and 2. QC mean density is higher than QA mean density in five out of 11 lots that had both QC and QA data, which is about 45%. This result is different from majority of the projects in which QC mean density is higher than QA mean density in most of the lots. The mean difference is the highest in lot 8 and the smallest in lot 9, respectively. Lot 11 has the lowest QC mean density whereas lot 5 has the lowest QA mean density : U54 (Seward)-PG 70-22, SM-12.5A, 2 in QC QA Lot Number FIGURE 2.26 QC and QA Mean Density Comparison for US-54 in District 6 Figure 2.27 shows summary of QC/QA mean density comparison for 12 projects in four KDOT districts. QC mean density was compared to QA mean density. QC mean density is higher than QA mean density for all projects except three. The three projects in which QC mean density is lower than QA mean density in most of the cases are projects 7, 11, and 12. Project 7 is US-83 route in Sheridan County, which is located in District 3. Projects 11 and 12 are located in District 6. Project 11 is K-27 route in Stanton County whereas project 12 is US-54 route in Seward County. In general, QC mean density is higher than the QA mean density. 27

43 Frequency (Number) 14.0 Higher Lower Project Number FIGURE 2.27 Summary of QC and QA Mean Density Comparison Maximum Density Comparison Figure 2.28 shows QC/QA maximum density comparison for US-81 route in Ottawa County. It has the highest number of lots out of all projects in District 2 in this study. QC maximum density is higher than QA maximum density in 10 out of 12 lots, which is about 83%. QC/QA maximum density difference is the highest in lot 5 like mean density difference. QC/QA maximum density difference is the smallest in lot 6 unlike mean density difference. Figure 2.29 shows QC/QA maximum density comparison for K-383 route in Decatur County. It has the highest number of lots of all projects in District 3 in this study. QC maximum density is higher than QA maximum density in 12 out of 16 lots, which is 75%. QC/QA maximum density difference is the highest in lot 3 like mean density difference. The smallest QC/QA maximum density difference is observed in lot 6. Lot 1 has the lowest QC/QA maximum density like the mean density. QC/QA maximum density difference may not be significant from a practical point of view. Also, since the contractor is doing more tests than KDOT per lot, the contractor test results should be higher and lower 28

44 Maximum Density (%) Maximum Density (%) : U81 (Ottawa)-PG 58-28, SR-19A, 3.5 in QC QA Lot Number FIGURE 2.28 QC and QA Maximum Density Comparison for US-81 in District : K383 (Decatur)-PG 64-22, SR-19A, 2 in QC QA Lot Number FIGURE 2.29 QC and QA Maximum Density Comparison for K-383 in District 3 Figure 2.30 shows QC/QA maximum density comparison for US-54 route in Kingman County. It has the highest number of lots of all projects in District 5 in this study. QC maximum 29

45 Maximum Density (%) density is higher than QA maximum density in six out of 9 lots, which is about 67%. QC/QA maximum density difference is the highest in lot 1 unlike mean density difference. The smallest QC/QA maximum density difference has been observed in lot 5 unlike mean density difference. Lot 4 has the smallest QC maximum density whereas lot 1 has the smallest QA maximum density : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5 in QC QA Lot Number FIGURE 2.30 QC and QA Maximum Density Comparison for US-54 in District 5 Figure 2.31 shows QC/QA maximum density comparison for US-54 route in Seward County. It has the highest number of lots of all projects in District 6 in this study. There were no QA data for lots 1 and 2. QC maximum density is higher than QA maximum density in one out of 11 lots that had both QC and QA data, which is about 9%. This result is different from majority of the projects in which QC maximum density is higher than QA maximum density in most of the lots. The maximum density difference is the highest in lot 10 and the smallest in lot 13, respectively. Lot 11 has the smallest QC maximum density whereas lot 10 has the smallest QA maximum density. 30

46 Maximum Density (%) : U54 (Seward)-PG 70-22, SM-12.5A, 2 in QC QA Lot Number FIGURE 2.31 QC and QA Maximum Density Comparison for US-54 in District 6 Figure 2.32 shows summary of QC/QA maximum density comparison for 12 projects in four KDOT districts. QC maximum density was compared to QA maximum density in a lot. QC maximum density is higher than QA maximum density for all projects except four. The four projects in which QC maximum density is lower than QA maximum density in most of the cases are projects 5, 7, 11, and 12. Projects 5 and 7 are located in District 3. Project 5 is US-283 in Norton County. Project 7 is US-83 route in Sheridan County. Projects 11 and 12 are located in District 6. Project 11 is K-27 route in Stanton County whereas project 12 is US-54 route in Seward County. QA density in lot 1 is missing for K-27 route and QA density for lots 1 and 2 are missing for US-54 route. In general, lot-by-lot comparison shows that QC maximum density is higher than the QA maximum density. One of the reasons for this may be due to more data points taken by contractor, in which maximum values are expected Minimum Density Comparison Minimum (lowest) QC and QA density was selected for each lot for different projects. Figure 2.32 shows QC/QA minimum density comparison for US-81 route in Ottawa County. It has the highest number of lots of all projects in District 2 in this study. QC minimum density is 31

47 Minimum Density (%) Frequency (Number) higher than QA minimum density in eight out of 12 lots, which is about 67%. QC/QA minimum density difference is the highest in lot 6 unlike mean and maximum density difference. QC/QA minimum density difference is the smallest in lot 2 like mean density difference, but unlike maximum density difference Higher Lower Project Number FIGURE 2.32 Summary of QC and QA Maximum Density Comparison : U81 (Ottawa)-PG 58-28, SR-19A, 3.5 in QC QA Lot Number FIGURE 2.33 QC and QA Minimum Density Comparison for US-81 in District 2 32

48 Minimum Density (%) Figure 2.34 shows QC/QA minimum density comparison for K-383 route in Decatur County. It has the highest number of lots of all projects in District 3 in this study. QC minimum density is higher than QA mean density in 12 out of 16 lots, which is 75% like maximum density comparison. QC/QA minimum density difference is the highest in lot 8 unlike mean and maximum density difference. The smallest QC/QA minimum density difference is observed in lot 2. Lot 1 has the lowest QC/QA minimum density like the mean and maximum density comparison. QC/QA minimum density difference may not be significant from a practical point of view : K383 (Decatur)-PG 64-22, SR-19A, 2 in QC QA Lot Number FIGURE 2.34 QC and QA Minimum Density Comparison for K-383 in District 3 Figure 2.35 shows QC/QA minimum density comparison for US-54 route in Kingman County. It has the highest number of lots of all projects in District 5 in this study. QC minimum density is higher than QA minimum density in three out of 9 lots, which is about 33% unlike mean and maximum density comparison. QC/QA minimum density difference is the highest in lot 1 like maximum density difference comparison, but unlike mean difference comparison. The 33

49 Minimum Density (%) smallest QC/QA minimum density difference has been observed in lot 6 unlike maximum density difference. Lot 1 has the smallest QC/QA minimum density : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5 in QC QA Lot Number FIGURE 2.35 QC and QA Minimum Density Comparison for US-54 in District 5 Figure 2.36 shows QC/QA minimum density comparison for US-54 route in Seward County. It has the highest number of lots out of all projects in District 6 in this study. There were no QA data for lots 1 and 2. QC minimum density is higher than QA mean density in six out of 11 lots that had both QC and QA data, which is about 55%. The minimum density difference is the highest in lot 5 and the smallest in lot 13, respectively. Lot 9 has the smallest QC minimum density whereas lot 5 has the smallest QA minimum density. Figure 2.37 shows summary of QC/QA minimum density comparison for 12 projects in four KDOT districts. QC minimum density was compared to QA minimum density for a lot. Frequency distribution for each project based on a lot-by-lot minimum density comparison has been plotted. QC minimum density is higher than QA minimum density for all projects for the majority of lots except project 9. Project 9 is US-54 route in Kingman County, located in District 5. QC minimum density is higher than QA minimum density in all lots for project 8. Project 8 is 34

50 Frequency (Number) Minimum Density (%) US-50 route in Reno County, located in District 5. In general, lot-by-lot comparison shows that QC minimum density is higher than QA minimum density. One reason may be due to more data points taken by contractor, in which lower values can be expected : U54 (Seward)-PG 70-22, SM-12.5A, 2 in QC QA Lot Number FIGURE 2.36 QC and QA Minimum Density Comparison for US-54 in District Higher Lower Project Number FIGURE 2.37 Summary of QC and QA Minimum Density Comparison 35

51 STD (%) Standard Deviation (STD) for Density Comparison Figure 2.38 shows QC/QA density comparison for US-81 route in Ottawa County. It has the highest number of lots of all projects in District 2 in this study. QC STD density is higher than QA STD density in six out of 12 lots, which is 50%. The STD difference is the highest in lot 6 and the smallest in lot 2, respectively. Again, the contractor STD is expected to be larger than the KDOT STD because the number of tests done per lot is higher for the contractor : U81 (Ottawa)-PG 58-28, SR-19A, 3.5 in QC QA Lot Number FIGURE 2.38 QC and QA STD Density Comparison for US-81 in District 2 Figure 2.39 shows QC/QA STD density comparison for K-383 route in Decatur County. It has the highest number of lots out of all projects in District 3 in this study. QC STD density is higher than QA STD density in 3 out of 16 lots, which is about 19%. The STD difference is the highest in lot 8 and the smallest in lot 5, respectively. Lot 9 has the smallest QC STD density whereas lot 3 has the smallest QA STD density. Figure 2.40 shows QC/QA STD density comparison for US-54 route in Kingman County. It has the highest number of lots out of all projects in District 5 in this study. QC STD density is higher than QA STD density in five out of nine lots, which is about 56%. The STD difference is 36

52 STD (%) STD (%) the largest in lot 1 and the smallest in lot 3, respectively. Lot 8 has the smallest QC STD density whereas lot 4 has the smallest QA STD density : K383 (Decatur)-PG 64-22, SR-19A, 2 in QC QA Lot Number FIGURE 2.39 QC and QA STD Density Comparison for K-383 in District : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5 in QC QA Lot Number FIGURE 2.40 QC and QA STD Density Comparison for US-54 in District 5 37

53 STD (%) Figure 2.41 shows QC/QA STD density comparison for US-54 route in Seward County. It has the highest number of lots out of all projects in District 6 in this study. There were no QA data for lots 1 and 2. QC STD density is higher than QA STD density in one out of 11 lots that had both QC and QA data, which is about 9%. The STD difference is the largest in lot 6 and the smallest in lot 13, respectively. Lot 11 has the smallest QC STD density whereas lot 10 has the lowest QA STD density : U54 (Seward)-PG 70-22, SM-12.5A, 2 in QC QA Lot Number FIGURE 2.41 QC and QA STD Density Comparison for US-54 in District 6 Figure 2.42 shows summary of QC/QA STD density comparison for 12 projects in four KDOT districts. QC STD density was compared to QA STD density. QC STD density is higher than QA STD density for only three projects out of 12, which is 25%. The three projects in which QC STD density is higher than QA STD density in most of the cases are projects 2, 9, and 10. Project 2 is US-81 route in Ottawa County, which is located in District 2. Projects 9 and 10 are located in District 6. Project 9 is US-54 route in Seward County whereas project 10 is US-50 route in Finney County. In general, QC STD density is lower than QA STD density. 38

54 Frequency (Number) Coefficient of Variation (COV) for Density Comparison Figure 2.43 shows QC/QA COV density comparison for US-81 route in Ottawa County. It has the highest number of lots of all projects in District 2 in this study. QC COV density is higher than QA COV density in six out of 12 lots, which is 50% like STD comparison. The COV difference is the highest in lot 6 and the smallest in lot 2, respectively like STD Higher Lower Project Number FIGURE 2.42 Summary of QC and QA STD Density Comparison 39

55 COV (%) : U81 (Ottawa)-PG 58-28, SR-19A, 3.5 in QC QA Lot Number FIGURE 2.43 QC and QA COV Density Comparison for US-81 in District 2 Figure 2.44 shows QC/QA COV density comparison for K-383 route in Decatur County. It has the highest number of lots of all projects in District 3 in this study. QC COV density is higher than QA COV density in 2 out of 16 lots, which is about 13%. The COV difference is the highest in lot 8 and the smallest in lot 5, respectively like STD comparison. Lot 9 has the smallest QC COV density whereas lot 3 has the smallest QA COV density like STD comparison. 40

56 COV (%) : K383 (Decatur)-PG 64-22, SR-19A, 2 in QC QA Lot Number FIGURE 2.44 QC and QA COV Density Comparison for K-383 in District 3 Figure 2.45 shows QC/QA COV density comparison for US-54 route in Kingman County. It has the highest number of lots out of all projects in District 5 in this study. QC COV density is higher than QA COV density in five out of nine lots, which is about 56% like STD comparison. The COV difference is the largest in lot 1 and the smallest in lot 9, respectively. Lot 4 has the smallest QC COV density whereas lot 8 has the smallest QA COV density. Figure 2.46 shows QC/QA COV density comparison for US-54 route in Seward County. It has the highest number of lots of all projects in District 6 in this study. There were no QA data for lots 1 and 2. QC COV density is higher than QA COV density in one out of 11 lots that had both QC and QA data, which is about 9% like STD comparison. The COV difference is the largest in lot 6 and the smallest in lot 13, respectively. Lot 11 has the smallest QC COV density whereas lot 10 has the lowest QA COV density. 41

57 COV (%) COV (%) : U54 (Kingman)-PG 64-28, SR-12.5A, 2.5 in QC QA Lot Number FIGURE 2.45 QC and QA COV Density Comparison for US-54 in District : U54 (Seward)-PG 70-22, SM-12.5A, 2 in QC QA Lot Number FIGURE 2.46 QC and QA COV Density Comparison for US-54 in District 6 Figure 2.47 shows summary of QC/QA COV density comparison for 12 projects in four KDOT districts. QC COV density was compared to QA COV density. QC COV density is higher than QA COV density for only three projects out of 12, which is 25%. The three projects in 42

58 Frequency (Number) which QC COV density is higher than QA COV density in most of the cases are projects 2, 9, and 10. Project 2 is US-81 route in Ottawa County, which is located in District 2. Projects 9 and 10 are located in District 6. Project 9 is US-54 route in Seward County whereas project 10 is US- 50 route in Finney County. QC COV is lower than QA COV in all the lots for project 11. Project 11 is K-27 in Stanton County, located in District 6. In general, QC COV density is lower than QA COV density Higher Lower Project Number FIGURE 2.47 Summary of QC and QA COV density comparison. 2.6 Comparison of Means The FHWA technical advisory recommends using the F & t statistical procedures to compare both variance and means of two data sets. The F-test compares the variances of two data sets. The objective of this test is to determine whether the differences in the variability of the contractor s tests and the department tests are greater than what might be expected if they came from the same population. On the other hand, the t-test compares the means of two data sets to assess whether they are statistically different (Elseifi et al. 2009). KDOT uses F-test to determine equality of variance and then t-test to compare QC and QA means. 43

59 The term analysis of variance (ANOVA) describes a group of inferential statistical tests whereas a t-test is used in statistics to determine if the means of two groups differ significantly. ANOVA evaluates the null hypothesis that in a set of k independent samples (where k 2), all k samples are drawn from the same population, with the alternate hypothesis that at least two of the samples are drawn from populations with different mean values. The test statistic computed is based on the F-distribution. In the case of comparing two means, which is the case for QC and QA, the t- and the F-tests are equivalent when variances are equal. The F-test in ANOVA can signify that not all the means of the levels of the classification variable are the same, but it cannot indicate which means differ from which other means. Comparison methods for means provide more detailed information about the differences among the means. Four comparison methods for means have been used in this study Fisher s Least Significant Difference (LSD) Test Multiple t-tests are used to compare pairs of means. Fisher s LSD tests is the most powerful for finding differences between pairs of means since it does not adjust the significance level needed to achieve significance in order to account for multiple testing. As a result, it has the greatest chance of resulting in one or more Type I errors Tukey s Honestly Significant Difference (HSD) Test This test is generally recommended when a researcher plans to make all possible pairwise comparisons since it controls the Type I error rate so that it will not exceed the significance level value pre-specified in the analysis. It maintains an acceptable significance level without an excessive loss of power Student-Newman-Keuls (SNK) Test This test is similar to and/or more powerful than Tukey s HSD. However, it does not control experiment-wise error rate at significance level. 44

60 2.6.4 Scheffe s Test This test is extremely flexible, allowing for any type of comparison. This increased versatility results in less power to detect differences between pairs of groups. It is the most conservative of the unplanned comparison procedures. The test specifies a fixed value of significance level which does not depend on the number of comparisons conducted Lot-Wise Means Comparison Lot-wise means comparison was carried out using four means comparison methods at three different significant levels. Significance difference was summarized into frequency distribution for both Superpave and PCC pavements. Quality characteristics for Superpave pavements include QC air voids, QC smoothness, and QC and QA density. Lot-wise means comparison was carried out for PCC QC strength and thickness Superpave Pavements Figure 2.48 shows lot-wise comparison of QC air void means. Student-Newman-Keuls (SNK) and Tukey s Honestly Significant Difference (HSD) show the same results at all significance levels. There is no significant difference between lot means in most cases for QC air at all significance levels and for all methods except LSD. Figure 2.49 shows lot-wise comparison of QC smoothness means. Student-Newman- Keuls (SNK) and Tukey s Honestly Significant Difference (HSD) show the same results at all significance levels. There is significant difference between lot means in most cases at all significance levels and for all methods for QC smoothness except Scheffe method at 1% significance level. Figure 2.50 shows lot-wise comparison of QC density means. Student-Newman-Keuls (SNK) and Tukey s Honestly Significant Difference (HSD) show the same results at all significance levels. There is significant difference between lot means in most cases at all significance levels and for all methods. Scheffe method is the only method that does not show significant difference between means for some cases at 2.5% and 5%, respectively. Figure 2.51 shows lot-wise comparison of QA density means. Student-Newman-Keuls (SNK) and Tukey s Honestly Significant Difference (HSD) show the same results at all 45

61 Frequency (Number) Frequency (Number) significance levels. There is no significant difference between lot means in most cases at all significance levels and for all methods except LSD. All methods show the same result at 5% significant level except LSD. LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.48 Lot-Wise Means Comparison for QC Air Voids 25 LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.49 Lot-Wise Mean Comparison for QC Smoothness 46

62 Frequency (Number) Frequency (Number) LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.50 Lot-Wise Means Comparison for QC Density 12 LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.51 Lot-Wise Means Comparison for QA Density 47

63 PCC Pavements Figure 2.52 shows lot-wise means comparison for QC PCC strength data. SNK and HSD show the same results at all significance levels. There is a significant difference between lot means in most cases using all methods at all significance levels except Scheffe method. This confirms that Scheffe method is the weakest in detecting significant differences. Figure 2.53 shows lot-wise means comparison for QC PCC thickness data. SNK and HSD show the same results at all significant levels. There is significant difference between lot means in most cases using all methods at all significance levels except using Scheffe s method. This confirms that Scheffe method is the weakest in detecting significant differences. Lot-wise comparison shows that QC and QA means are significantly different in most cases. As a result, QC/QA comparison should be considered lot-wise instead of KDOT s current procedure that combines data from five successive lots for air voids of Superpave pavements and strength and thickness of PCC pavements. More sublot data may be taken for each lot so that enough data can be obtained for statistical analysis. Ten QC readings and five QA readings per lot, similar to current Superpave density data, will be enough for more robust statistical analysis. This result confirms the study by Benson (1995). It was suggested that within practical limitations of the type of job, lot size could be expanded tenfold to encompass an entire week s production. There would be considerable benefits in terms of reduced staff and equipment inventory if larger lot sizes are implemented. The increase in risk to buyers and sellers as a result of slightly higher within-lot variability are not unreasonable Sublot-Wise Means Comparison Superpave Pavements Figure 2.54 shows sublot-wise mean comparison for QC/QA air voids using four mean comparison methods at three different significance levels. Sublot-wise QC/QA comparison for air voids has been done using four QC sublot readings and QA reading as the fifth sub-lot reading in each lot. Sheffe method shows that there is no significant difference between the sublot means of QC/QA air voids at 1% significance level. The result shows that significant difference using LSD method clearly increases with an increase in significance level. 48

64 Frequency (Number) Frequency (Number) LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.52 Lot-Wise Means Comparison for PCC Strength LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.53 Lot-Wise Means Comparison for PCC Thickness 49

65 Frequency (Number) LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.54 Sublot-Wise Means Comparison for QC/QA Air Voids Figure 2.55 shows sublot-wise mean comparison for QC density using four mean comparison methods at three different significance levels. There is no significant difference using all methods except LSD for QC density. The result shows that significant difference using LSD method clearly increases with an increase in significance level. Figure 2.56 shows sublot-wise mean comparison for QA density using four mean comparison methods at three different significance levels. There is no significant difference using all methods at all significant levels. Figure 2.57 shows sublot-wise mean comparison for QC/QA using four mean comparison methods at three different significance levels. Sublot-wise QC/QA density analysis has been done using 10 QC sublot data and five QA sublot data in each lot. There is no significant difference using all methods except LSD for QC/QA density. The result shows that significant difference using LSD method clearly increases with an increase in significance level. 50

66 Frequency (Number) Frequency (Number) LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.55 Sublot-Wise Means Comparison for QC Density LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.56 Sublot-Wise Means Comparison for QA Density 51

67 Frequency (Number) LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.57 Sublot-Wise Means Comparison for QC/QA Density PCC Pavements Figure 2.58 shows sublot-wise mean comparison for QC PCC strength using four mean comparison methods at three different significance levels. The results show that significant difference using all methods increases with an increase in significance level for PCC strength. All methods, except LSD show no significant difference at 1% significant level. Figure 2.59 shows sublot-wise mean comparison for QC PCC thickness using four mean comparison methods at three different significance levels. There is no significant difference between sublot means using all methods at all significance levels for PCC thickness except LSD at 5% significance level. This shows that LSD is the most powerful method to detect significant differences and significant difference increases with an increase in significance level. 52

68 Frequency (Number) Frequency (Number) LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.58 Sublot-Wise Means Comparison for QC PCC Strength LSD SNK HSD Scheffe No Yes No Yes No Yes Similar at Significance Level (%)? FIGURE 2.59 Sublot-Wise Means Comparison for QC PCC Thickness 53

69 Frequency (Number) Sublot-Wise Comparison of Means The order of sublot means has been investigated for any trend using QC/QA air voids, QA density, QC PCC strength and thickness Superpave Pavements The rank varies from 1 (largest mean) to 5 (smallest mean) corresponding to five sublots in a lot for QC/QA air voids. QA has been taken as sublot 5 in this analysis. Quality assurance for air voids and QC sublot 1 show for most of the time, the largest and smallest mean, respectively, as shown in Figure QC Sublot 1 QC Sublot 2 QC Sublot 3 QC Sublot 4 QA Rank FIGURE 2.60 Sublot-Wise Comparison of Mean Magnitude for QC/QA Air Void The rank varies from 1 (largest mean) to 5 (smallest mean) corresponding to five sublots in a lot for QA density. Figure 2.61 shows in most cases, sublots 2 and 3 have the largest and 54

70 Frequency (Number) smallest mean, respectively, for QA density. The results show that sublots 1 and 5 are neither the smallest nor the largest consistently. 6 Sublot 1 Sublot 2 Sublot 3 Sublot 4 Sublot Rank FIGURE 2.61 Sublot-Wise Comparison of Mean Magnitude for QA Density PCC Pavements The rank varies from 1 (largest mean) to 4 (smallest mean) corresponding to 4 sublots in a lot for QC PCC strength. Sublots 1 and 4 have the largest mean in most and fewest cases, respectively, for QC PCC strength as shown in Figure The rank varies from 1 (largest mean) to 4 (smallest mean) corresponding to 4 sublots in a lot for QC PCC thickness. Figure 2.63 shows that sublots 3 and 4 have the largest mean with the same frequency. Sublot 3 has the smallest mean in most cases. These results do not show any specific trend that the first or the last sublot reading is the largest or the smallest. 55

71 Frequency (Number) Frequency (Number) Sublot 1 Sublot 2 Sublot 3 Sublot Rank FIGURE 2.62 Sublot-Wise Comparison of Mean Magnitude for QC PCC Strength 7 Sublot 1 Sublot 2 Sublot 3 Sublot Rank FIGURE 2.63 Sublot-Wise Comparison of Mean Magnitude for QC PCC Thickness 56

72 2.6.8 Effect of Significance Level on Pay Adjustments Effect of significance level on pay adjustment was investigated using QC/QA air voids and density data. One project was selected from each district for this investigation. The Excel spreadsheet of KDOT (2009 version) was used for three different significance levels. The results are presented for QC/QA air voids and density separately QC/QA Air Voids Table 2.5 shows pay adjustment for QC/QA air voids corresponding to the three different significance levels. One large project from each district was selected for the investigation. Project size in terms of tonnage varies from 185,329 in District 1 to 222,276 in District 6. TABLE 2.5 Pay Adjustment for QC/QA Air Voids District Size (ton) Significant Level 1% 2.50% 5% 1 185,329 67,305 67,305 62, ,229 65,743 58,073 58, , , , , ,750 4,800 4,800 4, ,581 39,400 39,400 39, , , , ,563 Figure 2.64 shows pay adjustment for QC/QA air voids at three different significance levels in each of KDOT s six districts. There is difference in pay adjustment only in Districts 1 and 2. Pay adjustments are the same at 1% and 2.5% significance levels for the project in District 1. Pay adjustments are equal at 2.5% and 5% significance levels for the project in District QC/QA Density Table 2.6 shows pay adjustment for QC/QA density corresponding to the three different significance levels. Some lots from a project from each district were selected for the investigation. 57

73 Air Voids Pay Adjustment (dollar) 1% 2.50% 5% District FIGURE 2.64 Pay Adjustment for QC/QA Air Voids TABLE 2.6 Pay Adjustment for QC/QA Density District Size (ton) Significant Level 1% 2.50% 5% Figure 2.65 shows pay adjustment for QC/QA density at three different significance levels in each of KDOT s six districts. There is difference in pay adjustment only in Districts 1 and 4. Pay adjustments are different at all significant levels in both districts. 58

74 Density Pay Adjustment (dollar) Even though the differences in pay adjustments at 1% and 2.5% are not significant for the selected projects or sample lots, it can be significant amount of money when many large projects are considered. Currently KDOT uses 1% significance level and it is difficult to find significant difference at this level. It is recommended that 2.5% significance level be used as a compromise between 1 and 5% at all significance levels for both contractors and KDOT % 2.50% 5% District FIGURE 2.65 Pay Adjustment for QC/QA Density 2.7 F & t Tests Using Superpave Pavement Density Data Superpave pavement density data was used to investigate whether F-test can be used to determine significant differences between QC and QA data instead of determining the equality of variance only. It is to be noted that F and t tests give the same result when two means are compared and equal variance assumption is valid. Since KDOT deals with two means i.e. QC and QA, the first part is fulfilled. The main emphasis is when equal variance assumption is not valid. Students t-test has been done for both equal variance and unequal variance cases. 59

75 Frequency (Number) Figure 2.66 shows the comparison of t-test results when the variances are equal. However, computation was done for both equal variances and assumed unequal variance cases. Again, this test was repeated when the variances were unequal. In this case, computation was done for unequal variance as well as assuming equal variances. For equal variance case, in 72 cases, there was no significant difference between QC and QA data. However, significant difference was observed in 15 cases. QC and QA Different QC and QA Similar Equal Assume Unequal Assume Equal Unequal Equal Unequal Variance FIGURE 2.66 QC/QA Density Significant Difference Test When variances are unequal, assumptions of equal and unequal variances yield the same result. Figure 2.67 shows the discrepancy in test results when equal and unequal variances are used. When the variances are unequal, the discrepancy in test results is the same which shows the risk to KDOT and the contractor is equal. Overall, the probability of getting wrong results while using F-test is about 2% (only 2 out of 104 cases were wrong). This shows that F-test can be used to determine significant difference in means. 60

76 Frequency (Number) 6 QC and QA Different QC and QA Similar Equal Assume Unequal Assume Equal Unequal Equal Unequal Variance FIGURE 2.67 Discrepancy of QC/QA Significant Difference Test for Equal and Unequal Variance 2.8 Comparison of Design and Actual Asphalt Content Figure 2.68 shows comparison of actual and design AC content for US-75 route in District 1. It is typical case in which actual asphalt content is lower than the design asphalt content in all sublots. Figure 2.69 shows comparison of actual and design AC content for US-36 route in District 3. It is typical case in which actual asphalt content is higher than the design asphalt content in all sublots. Figure 2.70 shows comparison of actual and design AC content for US-77 route in Cowley County. This is typical of the situation when actual AC content is lower or higher than design AC content. In almost all cases, actual asphalt content is lower than the design asphalt content. Figure 2.71 shows the summary of actual asphalt content as compared to the design asphalt content. Actual asphalt content is higher than the design asphalt content in very few 61

77 Binder Content (%) Binder Content (%) cases. Actual asphalt content is lower or higher than design in most of the cases, but it is lower than the actual most points as shown in Figure : US75 (Brown), PG 64-22, SM-19A Pb-Actual Pb-Design Number of Data Points FIGURE 2.68 Typical Comparisons of Design and Actual AC When Actual Is Lower Than Design : U36 (Norton), PG 64-28, SM-19A 6.00 Pb-Actual Pb-Design Number of Data Points FIGURE 2.69 Typical Comparison of Design and Actual AC When Actual is Lower Than Design 62

78 Frequency (Number) Binder Content (%) : U77 (Cowley), PG 64-22, SM-19A Pb-Actual Pb-Design Number of Data Points FIGURE 2.70 Typical Comparison of Design and Actual AC When Actual is Lower and Higher Than Design Lower Higher Lower and Higher Actual AC Content as Compared to Design AC Content FIGURE 2.71 Summary of Actual and Design AC Content Comparison 63

79 It is recommended that the contractor provide mixes that have the average asphalt content as that in the approved design. Performance tests used to determine mix suitability (Lottman test, etc.) are performed at the design asphalt content. Therefore, production should provide mixes with the same properties. Incentives and/or penalties could be applied accordingly. The second option is to increase minimum VMA if additional asphalt is needed based on the performance of the pavements. 64

80 Chapter 3: Practical Performance Model and Composite Index 3.1 General Burati et al. (2003) concluded that percent within limit (PWL) is well suited as a statistical measure of quality since it has been well studied, statistically unbiased, suitable for both normal and distribution-free (attributes) applications, and works equally well for singlesided or double-sided specifications. Percent within limit (PWL) has been used to develop composite and practical performance models. Acceptable quality level (AQL) has been taken as 90 percent within limit for different quality characteristics for both Superpave and PCC pavements whereas different rejectable quality levels (RQL) have been used partly to investigate the effect of different RQL on the models and partly based on the effect of each quality characteristics on the performance of the pavement. The expected life (EL) was taken as 10 and 20 years for Superpave and PCC pavements, respectively when PWL=90 for all quality characteristics whereas EL was taken as 5 and 10 years when one of the quality characteristics is at RQL level. These values can be updated based on actual performance data and experience of the agency. Practical performance models and composite index were developed for Superpave and PCC pavements using different quality characteristics. 3.2 Practical Performance Model One of the first steps in developing a mathematical model is the choice of model form. Since most quality characteristics have points of diminishing returns, a model with an S shape may be appropriate (Weed 2006). Practical performance model of the form shown by Equation 3.1 has been developed. Expected life (EL) was used as a measure of performance (dependent variable) whereas different quality characteristics (variables) for both Superpave and PCC pavements were used as independent variables. Different shape factors (C) were assumed and simultaneous equations were solved using Excel for the model coefficients. Equation

81 3.2.1 Superpave Pavements Practical performance models that include two, three, four, and five quality characteristics were developed. Microsoft excel was used to solve simultaneous equations in order get model coefficients Two Quality Characteristics Practical performance model (PPM) that includes air voids and in-place density was developed using data in Table 3.1. Different values of shape factors were tried. The model was checked whether it returns precisely the values used to develop it. It was also checked at extreme values (PWL=100 and PWL=0), and examined how extra quality in some variables can offset the deficient quality in other variables while still producing design life of 10 years. TABLE 3.1 Data for PPM for Superpave Pavements (Two Variables, V=2) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years) Checking the Model (Shape Factor, C=1) The model was checked to make sure that it returns precisely the values used to develop it. Table 3.2 shows that the model returns the values used to develop it, which is shown in Table 3.1. TABLE 3.2 Test of Derivation of PPM for Superpave Pavements (V=2 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) 66 Density (DEN) (years)

82 A second test is to check at the extremes, an area in which many models break down. The extremes in this case occur when the individual PWL values are all either 100 or zero percent. These results are presented in Table 3.3. When PWL is 100 in both quality characteristics, the model predicts that the typical expected life of 10 years will be extended to approximately 14 years. It certainly falls within the experience of many agencies. At the other extreme, the model predicts an expected life less than a year. The model predicts an expected life of 3.4 years when PWL is 100 and 0 for air voids and in-place density, respectively. When PWL is 0 and 100 for air voids and in-place density, respectively, the model predicts an expected life of 2.4 years. At this stage, there is nothing to indicate the model is unsatisfactory, but several additional tests are required. TABLE 3.3 Test of Extremes of PPM for Superpave Pavements (V=2 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years) The third test is designed to examine how extra quality in some characteristics can offset deficient quality in others while still producing the design life of 10 years. This is an inherent feature in most design methods, and is believed to be an appropriate feature in any model of multiple quality characteristics. However, there would be concern if the model produced a sufficiently low level of quality in any individual characteristic that did not seem consistent with achieving the intended design life, even though the other characteristics were at excellent levels. Table 3.4 may not suggest that the model may have such a shortcoming, but other shape factors: 0.15, 0.25, 0.5, 0.75, 1.5, and 2 were considered to for the consistency of the models with other models which include more than two quality characteristics. Results from shape factor 0.5 are presented since it was found more reasonable. 67

83 TABLE 3.4 Test of Offsetting Property of PPM for Superpave Pavements (V=2 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years) Checking the Model (Shape Factor, C=0.5) As before, the first test of this model is to check that it correctly returns the values of expected life that were used to derive it, which it does. The next check is to repeat the series of tests shown in Table 3.4. The equivalent results, obtained with the revised model, are presented in Table 3.5. The values in Table 3.5 are not that much different from those in Table 3.4, but shape factor of 0.5 was used in the final model for consistency with other models that include more than two variables. TABLE 3.5 Test of Offsetting Property of PPM for Superpave Pavements (V=2 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years) The next test is to revisit Table 3.3 to check the values obtained at the extremes of PWL = 100 and PWL = 0. These are presented in Table 3.6 where it is seen that the inclusion of the exponential C term has given the revised model a diminished returns effect by reducing the maximum predicted life from the previous value of about 14 years to a value of 13 years. Further test was conducted. Both quality measures decline together. Table 3.7 shows a very logical progression as the results range from the maximum expected life of about 13 years for excellent quality down to the minimum of less than a year for extremely poor quality. It is 68

84 believed that most pavement engineers would consider this to be reasonably representative of field experience. TABLE 3.6 Test of Extremes of PPM for Superpave Pavements (V=2 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years) TABLE 3.7 Test of Progressively Poorer Quality for Superpave Pavements (V=2 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years)

85 The final PPM is shown by Equation 3.2. The model is used to better understand the consequences of either exceeding or falling short of the desired quality levels, and to provide a logical and defensible basis for the adjusted pay schedules that are an integral part of PPM. This model can be validated and/or improved based on Hamburg wheel tests at Kansas State University or KDOT s experience or actual performance data or a combination. Equation Three Quality Characteristics Practical performance model that includes three quality characteristics, air voids, in-place density, and Smoothness, was developed using data in Table 3.8. Smoothness was included in the model development if in case KDOT wants to include smoothness in the future. However, previous studies have recommended not including smoothness with other quality characteristics since the effect of initial smoothness has been assumed to be independent of the effects of the other variables (Weed 2000). Different values of shape factors were tried. The model was checked whether it returns precisely the values used to develop it. It was also checked at extreme values (PWL=100 and PWL=0), and examined how extra quality in some variables can offset the deficient quality in other variables while still producing design life of 10 years. TABLE 3.8 Data for PPM for Superpave Pavements (Three Variables, V=3) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years)

86 Checking the Model (Shape Factor, C=1) The model was checked to make sure that it returns precisely the values used to develop it. Table 3.9 shows that the model returns the values used to develop it, which are shown in Table 3.8. A second test is to check at the extremes, an area in which many models break down. The extremes in this case occur when the individual PWL values are all either 100 or zero percent. These results are presented in Table When PWL= 100 in all three quality characteristic, the model predicts that the typical expected life of 10 years will be extended to 15.5 years. This is an appreciable increase, but it certainly falls within the experience of many agencies. At the other extreme, the model predicts an expected life less than a year. The model predicts about a year when one of the quality characteristics has PWL=100 and the rest have PWL=0. Although not a frequent occurrence, most highway agencies have experienced this result at one time or another. At this stage, there is nothing to indicate the model is unsatisfactory, but several additional tests are required. TABLE 3.9 Test of Derivation of PPM for Superpave Pavements (V=3 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years) TABLE 3.10 Test of Extremes of PPM for Superpave Pavements (V=3 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) 71 (years)

87 The third test is designed to examine how extra quality in some characteristics can offset deficient quality in others while still producing the design life of 10 years. This is an inherent feature in most design methods, and is believed to be an appropriate feature in any model of multiple quality characteristics. However, there would be concern if the model produced a sufficiently low level of quality in any individual characteristic that did not seem consistent with achieving the intended design life, even though the other characteristics were at excellent levels. Table 3.11 suggests that the model may have such a shortcoming. For example, if PWL VA =PWL DEN =100, and PWL SM =65.5, the model predicts a design life of 10 years. This finding has raised doubts about the efficacy of the model when shape factor is one. It is now appropriate to consider other shape factors. Shape factors: 0.15, 0.25, 0.5, 0.75, 1.5, and 2 were considered. Results from shape factor 0.5 are presented since it was found more reasonable. TABLE 3.11 Test of Offsetting Property of PPM for Superpave Pavements (V=3 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years) Checking the Model (Shape Factor, C=0.5) As before, the first test of this model is to check that it correctly returns the values of expected life that were used to derive it, which it does. The next check is to repeat the series of tests shown in Table 3.11 that led to the rejection of the simpler model. The equivalent results, obtained with the revised model, are presented in Table The values in Table 3.12 seem more reasonable than those obtained in Table 3.11 even though the values are not far apart. 72

88 TABLE 3.12 Test of Offsetting Property of PPM for Superpave Pavement (V=3 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years) The next test is to revisit Table 3.10 to check the values obtained at the extremes of PWL = 100 and PWL = 0. These are presented in Table 3.13 where it is seen that the inclusion of the exponential C term has given the revised model a diminished returns effect by reducing the maximum predicted life from the previous value of about 15.5 years to value of about 14 years. TABLE 3.13 Test of Extremes of PPM for Superpave Pavements (V=3 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years) Further test was conducted. All quality measures decline together. Table 3.14 shows a very logical progression as the results range from the maximum expected life of about 14 years for excellent quality down to the minimum of less than a year for extremely poor quality. It is believed that most pavement engineers would consider this to be reasonably representative of field experience. The final PPM including the three quality characteristics for superpave pavements is shown by Equation 3.3. The model is used to better understand the consequences of either exceeding or falling short of the desired quality levels, and to provide a logical and defensible 73

89 basis for the adjusted pay schedules that are an integral part of PPM. This model can be validated and/or improved based on Hamburg wheel tests at Kansas State University or KDOT s experience or actual performance data or a combination. Equation Four Quality Characteristics Practical performance model (PPM) for Superpave pavements that includes four quality characteristics, air voids, in-place density, asphalt content, and voids in mineral aggregate (VMA), was developed using data in Table Different values of shape factors were tried. The model was checked whether it returns precisely the values used to develop it. It was also checked at extreme values (PWL=100 and PWL=0), and examined how extra quality in some variables can offset the deficient quality in other variables while still producing design life of 10 years. TABLE 3.14 Test of Progressively Poorer Quality for Superpave Pavements (V=3 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years)

90 TABLE 3.15 Data for PPM for Superpave Pavements (Four Variables, V=4) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years) Checking the Model (Shape Factor, C=1) The model was checked to make sure that it returns precisely the values used to develop it. Table 3.16 shows that the model returns the values used to develop it, shown in Table TABLE 3.16 Test of Derivation of PPM for Superpave Pavements (V=4 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years) A second test is to check at the extremes, an area in which many models break down. The extremes in this case occur when the individual PWL values are all either 100 or zero percent. These results are presented in Table When percent within limit 100 in all four quality characteristic, the model predicts that the typical expected life of 10 years will be extended to approximately 17 years. This is an appreciable increase, but it certainly falls within the experience of many agencies. At the other extreme, the model predicts an expected life of less than a year. Although not a frequent occurrence, most highway agencies have experienced this result at one time or another. The model predicts expected life less than a year when one of the 75

91 quality characteristics has PWL=100 and the rest have PWL=0. At this stage, there is nothing to indicate the model is unsatisfactory, but several additional tests are required. The third test is designed to examine how extra quality in some characteristics can offset deficient quality in others while still producing the design life of 10 years. This is an inherent feature in most design methods, and is believed to be an appropriate feature in any model of multiple quality characteristics. However, there would be concern if the model produced a sufficiently low level of quality in any individual characteristic that did not seem consistent with achieving the intended design life, even though the other characteristics were at excellent levels. Table 3.18 suggests that the model may have such a shortcoming. All four characteristics may be suspect, but the worst is probably VMA. For example, if PWL VA =PWL DEN = PWL AC =100, and PWL VMA =47, the model predicts a design life of 10 years. This finding has raised doubts about the efficacy of the model when shape factor is one. It is now appropriate to consider other shape factors. Shape factors: 0.15, 0.25, 0.5, 0.75, 1.5, and 2 were considered. Results from shape factor 0.5 are presented since it was found more reasonable. TABLE 3.17 Test of Extremes of PPM for Superpave Pavements (V=4 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years)

92 TABLE 3.18 Test of Offsetting Property of PPM for Superpave Pavements (V=4 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years) Checking the Model (Shape Factor, C=0.5) As before, the first test of this model is to check that it correctly returns the values of expected life that were used to derive it, which it does. The next check is to repeat the series of tests shown in Table 3.18 that led to the rejection of the simpler model. The equivalent results, obtained with the revised model, are presented in Table The values in Table 3.19 seem more acceptable than those obtained in Table 3.18 even though the difference is not that significant from a practical point of view. TABLE 3.19 Test of Offsetting Property of PPM for Superpave Pavements (V=4 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years) The next test is to revisit Table 3.17 to check the values obtained at the extremes of PWL = 100 and PWL = 0. These are presented in Table 3.20 where it is seen that the inclusion of the exponential C term has given the revised model a diminished returns effect by reducing the 77

93 maximum predicted life from the previous value of about 17 years to a possibly more realistic value of about 15 years. TABLE 3.20 Test of Extremes of PPM for Superpave Pavements (V=4 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years) Further test was conducted. All quality measures decline together. Table 3.21 shows a very logical progression as the results range from the maximum expected life of about 15 years for excellent quality down to the minimum of less than a year for extremely poor quality. It is believed that most pavement engineers would consider this to be reasonably representative of field experience. The final performance model is shown by Equation 3.4. The model is used to better understand the consequences of either exceeding or falling short of the desired quality levels, and to provide a logical and defensible basis for the adjusted pay schedules that are an integral part of PPM. This model can be validated and/or improved based on Hamburg wheel tests at Kansas State University or KDOT s experience or actual performance data or a combination. Equation

94 TABLE 3.21 Test of Progressively Poorer Quality for Superpave Pavements (V=4 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Asphalt Content (AC) VMA (years) Five Quality Characteristics Practical performance model (PPM) that includes five variables, air voids, in-place density, smoothness, asphalt content, and voids in mineral aggregate (VMA), was developed using data in Table Different values of shape factors were tried. The model was checked whether it returns precisely the values used to develop it. It was also checked at extreme values (PWL=100 and PWL=0), and examined how extra quality in some variables can offset the deficient quality in other variables while still producing design life of 10 years. 79

95 Air Voids (VA) TABLE 3.22 Data for PPM for Superpave Pavements (Five Variables, V=5) Percent Within Limit (PWL) for Various Quality Measures Density (DEN) Smoothness (SM) Asphalt Content (AC) VMA Expected Life (years) Checking the Model (Shape Factor, C=1) The model was checked to make sure that it returns precisely the values used to develop it. Table 3.23 shows that the model returns the values used to develop it, Table A second test is to check at the extremes, an area in which many models break down. The extremes in this case occur when the individual PWL values are all either 100 or zero percent. These results are presented in Table When PWL=100 in all five quality characteristics, the model predicts that the typical expected life of 10 years will be extended to about 19 years. This is an appreciable increase. At the other extreme, the model predicts an expected life of less than a year. Although not a frequent occurrence, most highway agencies have experienced this result at one time or another. At this stage, there is nothing to indicate the model is unsatisfactory, but several additional tests are required. 80

96 TABLE 3.23 Test of Derivation of PPM for Superpave Pavements (V=5 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Air Voids Density Smoothness Asphalt Content VMA Life (years) (VA) (DEN) (SM) (AC) TABLE 3.24 Test of Extremes of PPM for Superpave Pavements (V=5 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Air Voids Density Smoothness Asphalt Content VMA Life (years) (VA) (DEN) (SM) (AC) The third test is designed to examine how extra quality in some characteristics can offset deficient quality in others while still producing the design life of 10 years. This is an inherent feature in most design methods, and is believed to be an appropriate feature in any model of multiple quality characteristics. However, there would be concern if the model produced a sufficiently low level of quality in any individual characteristic that did not seem consistent with achieving the intended design life, even though the other characteristics were at excellent levels. 81

97 Table 3.25 suggests that the model may have such a shortcoming. All five characteristics may be suspect, but the worst is probably VMA in which PWL VMA =34 and PWL=100 for the rest predicts a design life of 10 years. This finding has raised doubts about the efficacy of the model when shape factor is one. It is now appropriate to consider other shape factors. Shape factors: 0.15, 0.25, 0.5, 0.75, 1.5, and 2 were considered. Results from shape factor 0.5 are presented since it was found more reasonable. TABLE 2.35 Test of Offsetting Property of PPM for Superpave Pavements (V=5 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Air Voids Density Smoothness Asphalt Content VMA Life (years) (VA) (DEN) (SM) (AC) Checking the Model (Shape Factor, C=0.5) As before, the first test of this model is to check that it correctly returns the values of expected life that were used to derive it, which it does. The next check is to repeat the series of tests shown in Table 3.25 that led to the rejection of the simpler model. The equivalent results, obtained with the revised model, are presented in Table The values in Table 3.26 seem more acceptable than those obtained in Table The next test is to revisit Table 3.24 to check the values obtained at the extremes of PWL = 100 and PWL = 0. These are presented in Table 3.27 where it is seen that the inclusion of the exponential C term has given the revised model a diminished returns effect by reducing the maximum predicted life from the previous value of about 19 years to a value of about 17 years. Further test was conducted. All quality measures decline together. Table 3.28 shows a very logical progression as the results range from the maximum expected life of about 17 years 82

98 for excellent quality down to the minimum of less than a year for extremely poor quality. It is believed that most pavement engineers would consider this to be reasonably representative of field experience. TABLE 3.26 Test of Offsetting Property of PPM for Superpave Pavements (V=5 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Air Voids (VA) Density (DEN) Smoothness (SM) Asphalt Content (AC) VMA Life (years) TABLE 3.27 Test of Extremes of PPM Models for Superpave Pavements (V=5 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Air Voids (VA) Density (DEN) Smoothness (SM) Asphalt Content (AC) VMA Life (years) The final performance model is shown by Equation 3.5. The model is used to better understand the consequences of either exceeding or falling short of the desired quality levels, and to provide a logical and defensible basis for the adjusted pay schedules that are an integral part of 83

99 PPM. This model can be validated and/or improved based on Hamburg wheel tests at Kansas State University or KDOT s experience or actual performance data or a combination. Equation 3.5 TABLE 3.28 Test of Progressively Poorer Quality for Superpave Pavements (V=5 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Air Voids Density Smoothness Asphalt Content VMA Life (years) (VA) (DEN) (SM) (AC) PCC Pavements Not all individual quality measures are equally suitable for incorporation into a composite measure. Measures that are best suited are those that jointly affect performance in such a way that higher quality in one tends to offset deficiencies in the others, within practical limits. Another requirement is that they be convenient to measure in association with each acceptance 84

100 lot. The example involving strength and thickness of rigid pavement is well suited (Weed 2000). Practical performance models that include two and three quality characteristics were developed Two Quality Characteristics Practical performance model (PPM) for Portland cement concrete (PCC) pavements that includes thickness and strength was developed using data in Table Different values of shape factors were tried. The model was checked whether it returns precisely the values used to develop it. It was also checked at extreme values (PWL=100 and PWL=0), and examined how extra quality in one variable can offset the deficient quality in other variable while still producing design life of 20 years. TABLE 3.29 Data for PPM for PCC Pavements (Two Variables, V=2) Percent Within Limit (PWL) for Various Quality Measures Thickness (TH) Strength (ST) Expected Life (years) Checking the Model (Shape Factor, C=1) The model was checked to make sure that it returns precisely the values used to develop it. Table 3.30 shows that the model returns the values used to develop it, shown in Table TABLE 3.30 Test of Derivation of PPM for PCC Pavements (V=2 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years)

101 A second test is to check at the extremes, an area in which many models break down. The extremes in this case occur when the individual PWL values are all either 100 or zero percent. These results are presented in Table When PWL=100 in both quality characteristic, the model predicts that the typical expected life of 20 years will be extended to about 27 years. This is an appreciable increase, but it certainly falls within the experience of many agencies. At the other extreme, the model predicts an expected life of about a year. The model predicts an expected life of about 7 years when PWL for thickness is 100 and PWL for strength is 0. The model predicts an expected life of about 5 years when PWL for thickness is 0 and PWL for strength is 100. At this stage, there is nothing to indicate the model is unsatisfactory, but several additional tests are required. TABLE 3.31 Test of Extremes of PPM for PCC Pavements (V=2 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years) The third test is designed to examine how extra quality in one characteristic can offset deficient quality in the other while still producing the design life of 20 years. This is an inherent feature in most design methods, and is believed to be an appropriate feature in any model of multiple quality characteristics. However, there would be concern if the model produced a sufficiently low level of quality in any individual characteristic that did not seem consistent with achieving the intended design life, even though the other characteristic was at excellent levels. Table 3.32 may not suggest that the model may have such a shortcoming, but other shape factors were considered for consistency with other models. Shape factors: 0.15, 0.25, 0.5, 0.75, 1.5, and 2 were considered. Results from shape factor 0.5 are presented since it was found more reasonable. 86

102 TABLE 3.32 Test of Offsetting Property of PPM for PCC Pavements (V=2 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years) Checking the Model (Shape Factor, C=0.5) As before, the first test of this model is to check that it correctly returns the values of expected life that were used to derive it, which it does. The next check is to repeat the series of tests shown in Table The equivalent results, obtained with the revised model, are presented in Table The values in Table 3.33 do not seem to differ significantly from those in Table TABLE 3.33 Test of Offsetting Property of PPM for PCC Pavements (V=2 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years) The next test is to revisit Table 3.31 to check the values obtained at the extremes of PWL = 100 and PWL = 0. These are presented in Table 3.34 where it is seen that the inclusion of the exponential C term has given the revised model a diminished returns effect by reducing the maximum predicted life from the previous value of about 27 years to value of about 26 years. The values are not significantly different from a practical point of view. 87

103 TABLE 3.34 Test of Extremes of PPM for PCC Pavements (V=2 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years) Further test was conducted. Both quality measures decline together. Table 2.35 shows a very logical progression as the results range from the maximum expected life of about 26 years for excellent quality down to the minimum of less than a year for extremely poor quality. It is believed that most pavement engineers would consider this to be reasonably representative of field experience. TABLE 3.35 Test of Progressively Poorer Quality for PCC Pavements (V=2 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years)

104 The final PPM is shown by Equation 3.6. The model is used to better understand the consequences of either exceeding or falling short of the desired quality levels, and to provide a logical and defensible basis for the adjusted pay schedules that are an integral part of PPM. This model can be improved based on KDOT s experience or actual performance data. Equation Three Quality Characteristics Air content is the property that is traditionally measured, as screening tests, to determine the durability (Diwan et al. 2003, Schell and Konecny 2003). Practical performance model that includes three quality characteristics, thickness, strength, and air content, was developed using data in Table Air content was included in the model development if in case KDOT wants to include air content as a surrogate for durability of PCC pavements. Different values of shape factors were tried. The model was checked whether it returns precisely the values used to develop it. It was also checked at extreme values (PWL=100 and PWL=0), and examined how extra quality in some variables can offset the deficient quality in other variables while still producing design life of 20 years. TABLE 3.36 Data for PPM for PCC Pavements (Three Variables, V=3) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years)

105 Checking the Model (Shape Factor, C=1) The model was checked to make sure that it returns precisely the values used to develop it. Table 3.37 shows that the model returns the values used to develop it, which are shown in Table TABLE 3.37 Test of Derivation of PPM for PCC Pavements (V=3 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years) A second test is to check at the extremes, an area in which many models break down. The extremes in this case occur when the individual PWL values are all either 100 or zero percent. These results are presented in Table When PWL= 100 in all three quality characteristic, the model predicts that the typical expected life of 20 years will be extended to about 31 years. This is an appreciable increase, but it certainly falls within the experience of many agencies. At the other extreme, the model predicts an expected life less than a year. The model predicts about two years when one of the quality characteristics has PWL=100 and the rest have PWL=0. Although not a frequent occurrence, most highway agencies have experienced this result at one time or another. At this stage, there is nothing to indicate the model is unsatisfactory, but several additional tests are required. 90

106 TABLE 3.38 Test of Extremes of PPM for PCC Pavements (V=3 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years) The third test is designed to examine how extra quality in some characteristics can offset deficient quality in others while still producing the design life of 20 years. This is an inherent feature in most design methods, and is believed to be an appropriate feature in any model of multiple quality characteristics. However, there would be concern if the model produced a sufficiently low level of quality in any individual characteristic that did not seem consistent with achieving the intended design life, even though the other characteristics were at excellent levels. Table 3.39 suggests that the model may have such a shortcoming. For example, if PWL TH =PWL ST =100, and PWL VA =63, the model predicts a design life of 20 years. This finding has raised doubts about the efficacy of the model when shape factor is one. It is now appropriate to consider other shape factors. Shape factors: 0.15, 0.25, 0.5, 0.75, 1.5, and 2 were considered. Results from shape factor 0.5 are presented since it was found more reasonable Checking the Model (Shape Factor, C=0.5) As before, the first test of this model is to check that it correctly returns the values of expected life that were used to derive it, which it does. The next check is to repeat the series of tests shown in Table 3.39 that led to the rejection of the simpler model. The equivalent results, obtained with the revised model, are presented in Table The values in Table 3.40 seem more reasonable than those obtained in Table 3.39 even though the values are not far apart. 91

107 TABLE 3.39 Test of Offsetting Property of PPM for PCC Pavements (V=3 and C=1) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years) TABLE 3.40 Test of Offsetting Property of PPM for PCC Pavements (V=3 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years) The next test is to revisit Table 3.38 to check the values obtained at the extremes of PWL = 100 and PWL = 0. These are presented in Table 3.41 where it is seen that the inclusion of the exponential C term has given the revised model a diminished returns effect by reducing the maximum predicted life from the previous value of about 31 years to value of about 28 years. Further test was conducted. All quality measures decline together. Table 3.42 shows a very logical progression as the results range from the maximum expected life of about 28 years for excellent quality down to the minimum of less than a year for extremely poor quality. It is believed that most pavement engineers would consider this to be reasonably representative of field experience. 92

108 TABLE 3.41 Test of Extremes of PPM for PCC Pavements (V=3 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years) TABLE 3.42 Test of Progressively Poorer Quality for PCC Pavements (V=3 and C=0.5) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Content (VA) (years) The final PPM including the three quality characteristics for PCC pavements is shown by Equation 3.7. The model is used to better understand the consequences of either exceeding or falling short of the desired quality levels, and to provide a logical and defensible basis for the 93

109 adjusted pay schedules that are an integral part of PPM. This model can be validated and/or improved based on KDOT s experience or actual performance data or a combination. Equation Pay Schedule The performance model serves two purposes. One is to better understand the consequences of either exceeding or falling short of the desired quality levels, and the other is to provide a logical and defensible basis for the adjusted pay schedules that are an integral part of PRS. Ideally, the purpose of the pay schedule is to provide incentive to the contractor to produce the desired levels of quality. Majority of highway agencies often include an additional incentive in the form of small bonus payments to contractors whose extra attention to quality control has produced work that substantially exceeds the acceptable quality levels. At the other extreme, when the desired levels of quality are not achieved, it is the purpose of the pay schedule to recoup for the highway agency the anticipated future losses resulting from poor performance (Weed 2006). To justify such an approach, there must be a link between quality received and economic gain or loss to the highway agency. Perhaps the most logical and consistent way to establish this link is through the use of life-cycle-cost analysis (Weed 2006). Equation for pay adjustment based on life-cycle-cost analysis was published by previous researchers (Weed 2001, Burati et al. 2003). The authors assumed for the derivation of Equation 3.8 that moderate deficiencies of construction are not repaired but, instead, lead to premature failure and an earlier scheduling of the next overlay. KDOT could readily obtain values for the constant terms in this equation. Equation 3.8 in which PA = appropriate pay adjustment for pavement or overlay (same units as C), C = present total cost of resurfacing, 94

110 DL = design life of pavement or overlay, EL= expected life of pavement or overlay (independent variable), OL= expected life of successive overlays (typically 10 years), and R = (1 + INF) / (1 + INR) in which INF is the long-term annual inflation rate and INT is the long-term annual interest rate, both in decimal form. Equations ( ) link quality to performance. They are used to predict the expected life (EL) used in Equation 3.8. Equation 3.8 links performance to economic gain or loss. Combining the two equations to link quality to economic effect provide a solid analytical basis for the pay schedule (Weed 2003). 3.3 Composite Index To demonstrate the practicality of the composite quality measure, a complete acceptance procedure must be specified. This includes the acceptable quality level (AQL), the rejectable quality level (RQL), the retest provision, and the pay schedule. Composite index was developed for both superpave and PCC pavements. Microsoft excel was used to solve simultaneous equations. Composite index without and with cross-product of quality characteristics were considered separated Superpave Pavements Composite index that includes two, three, four, and five supepave quality characteristics was developed separately. Composite index with cross-product of the quality characteristics was not developed since it became so cumbersome Two Quality Characteristics without Cross-Product Composite index (PWL*) was developed in terms of air voids (VA), and in-place density (DEN). The coefficients were obtained using the data in Table 3.1. The magnitudes of the coefficients reflect the effect of the variables on the long term performance of the pavements. The coefficients may be modified based on Hamburg wheel test results at Kansas State University or field performance or agency s experience or a combination. Composite index varies from zero to 100%. The final model developed is shown in Equation

111 Equation 3.9 To determine comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN =90 are substituted into Equation 3.9 to obtain PWL* = 90 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 90. To determine the value of PWL* associated with the RQL, any combination of values that give 5- year-life can be substituted into Equation 3.2. For example, entering PWL VA =79 and PWL DEN =80 into Equation 3.9 produces PWL*=66. Similarly, any combination of values that gives the 7.5-yearlife gives retest provision. Using PWL VA =79 and PWL DEN =80 gives PWL*=79.5. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.10 was derived. Equation 3.10 in which PA = lot pay adjustment ($ / lane kilometer), and PWL* = composite quality measure. It can be seen that when PWL* is at the AQL value of 90, Equation 3.10 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 66, the pay reduction of - $24,000 / lane kilometer (-$38,640 / lane mile) is obtained. For truly excellent quality, PWL VA = PWL DEN =100, PWL* =100, the pay equation awards a maximum bonus of $10,000 / lane kilometer ($16,100/ lane mile). At the other extreme, when PWL VA = PWL DEN =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$90,000 / lane kilometer (-$ 144,900/ lane mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Two Quality Characteristics with Cross-Product Data in Table 3.43 was used to develop expected life for two quality characteristics. Cross-product was included to investigate the difference between only addition and the one 96

112 which has cross-product. Microsoft excel was used to solve four simultaneous equations. The final expected life model that includes air voids and in-place density is shown by Equation The model predicts an expected life of 0.25 year when PWL=0 for both quality characteristics. Equation 3.11 was converted into composite index in terms of PWL* as shown in Equation Equation 3.12 gives PWL* ranging from 0 to 100%. Equation 3.11 Equation 3.12 TABLE 3.43 Data for Composite Index for Superpave Pavements (Variables, V=2) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) (years) To determine comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN =90 are substituted into Equation 3.12 to obtain PWL* = 80.5 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = To determine the value of PWL* associated with the RQL, any combination of values that give 5- year-life can be substituted into Equation For example, entering PWL VA =PWL DEN =64 into Equation 3.12 produces PWL*=39.5. Similarly, any combination of values that gives the 7.5- year-life gives retest provision. Using PWL VA =80 and PWL DEN =75.5 gives PWL*=59.5. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.13 is derived. 97

113 Equation 3.13 in which PA = lot pay adjustment ($ / lane kilometer), and PWL* = composite index. It can be seen that when PWL* is at the AQL value of 80.5, Equation 3.13 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 39.5, the pay reduction of - $41,000 / lane kilometer (-$66,010 / lane mile) is obtained. For truly excellent quality, PWL VA = PWL DEN =100, PWL* =100, the pay equation awards a maximum bonus of $19,500 / lane kilometer ($31,395/ lane mile). At the other extreme, when PWL VA = PWL DEN =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$80,500 / lane kilometer (-$129,605 / lane mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Three Quality Characteristics without Cross-Product Composite index (PWL*) was developed in terms of air voids (VA), in-place density (DEN), and smoothness (SM). The coefficients were obtained using the data in Table 3.9. The magnitudes of the coefficients reflect the effect of the variables on the long term performance of the pavements. The coefficients may be modified based on field performance and/or agency s experience. Composite index varies from zero to 100%. The final model developed is shown in Equation Equation 3.14 To determine the comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN = PWL SM =90 are substituted into Equation 3.14 to obtain PWL* = 90 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 90. To determine the value of PWL* associated with the RQL, any combination of values that give 5-98

114 year-life can be substituted into Equation 3.3. For example, entering PWL VA =75 and PWL DEN = PWL SM =70 into Equation 3.14 produces PWL*=72. Similarly, any combination of values that gives the 7.5-year-life produces retest provision. Using PWL VA = PWL DEN =80, and PWL SM =89 gives PWL*=82.5. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.10 was derived. When PWL* equals the RQL value of 72, the pay reduction of -$18,000 / lane kilometer (-$ 28,908/ lane mile) is obtained. For truly excellent quality, PWL VA = PWL DEN = PWL SM =100, PWL* =100, the pay equation awards a maximum bonus of $10,000 / lane kilometer ($16,100/ lane mile). At the other extreme, when PWL VA = PWL DEN = PWL SM =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$90,000 / lane kilometer (-$144,900 / lane mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Three Quality Characteristics with Cross-Product Data in Table 3.44 was used to develop expected life for three variables. Cross-product was included to investigate the difference between only addition and the one which has crossproduct. Microsoft excel was used to solve eight simultaneous equations. The final expected life model that includes air voids (VA), in-place density (DEN), and smoothness (SM) is shown by Equation The model predicts an expected life of 0.60 year when PWL=0 for all three quality characteristics. Equation 3.15 was converted into composite index in terms of PWL* as shown in Equation Equation 3.16 gives PWL* ranging from 0 to 100%. Equation

115 100 Equation 3.16

116 TABLE 3.44 Data for Composite Index for Superpave Pavements (Variables, V=3) Percent Within Limit (PWL) for Various Quality Measures Expected Life Air Voids (VA) Density (DEN) Smoothness (SM) (years) To determine comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN = PWL SM =90 are substituted into Equation 3.16 to obtain PWL* = 72.5 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = To determine the value of PWL* associated with the RQL, any combination of values that give 5- year-life can be substituted into Equation For example, entering PWL VA =68 and PWL DEN = PWL SM =70 into Equation 3.16 produces PWL*=34.5. Similarly, any combination of values that gives the 7.5-year-life produces retest provision. Using PWL VA =83 and PWL DEN = PWL SM =80 gives PWL*=53. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.17 was derived. Equation 3.17 in which PA = lot pay adjustment ($ / lane kilometer), and PWL* = composite quality measure. It can be seen that when PWL* is at the AQL value of 72.5, Equation 3.17 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 34.5, the pay reduction of - 101

117 $38,000 / lane kilometer (-$ 61,180 / lane mile) is obtained. For truly excellent quality, PWL VA = PWL DEN = PWL SM =100, PWL* =100, the pay equation awards a maximum bonus of $27,500 / lane kilometer ($44,275/ lane mile). At the other extreme, when PWL VA = PWL DEN = PWL SM =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$72,500 / lane kilometer (- $116,725 / lane mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Four Quality Characteristics without Cross-Product Composite index (PWL*) was developed in terms of air voids (VA), in-place density (DEN), asphalt content (AC), and voids in mineral aggregates (VMA). The coefficients were obtained using the data in Table The magnitudes of the coefficients reflect the effect of the variables on the long term performance of the pavements. The coefficients may be modified based on field performance and/or agency s experience. Composite index varies from zero to 100%. The final model developed is shown in Equation Equation 3.18 To determine comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN = PWL AC =PWL VMA =90 are substituted into Equation 3.18 to obtain PWL* = 90 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 90. To determine the value of PWL* associated with the RQL, any combination of values that give 5- year-life can be substituted into Equation 3.4. For example, entering PWL VA =PWL DEN = PWL AC =60 and PWL AC =65 into Equation 3.18 produces PWL*=61. Similarly, any combination of values that gives the 7.5-year-life produces retest provision. Using PWL VA = PWL DEN = PWL AC =75, and PWL VMA =74 gives PWL*=75. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.10 is derived. It can be seen that when PWL* is at the AQL value of 90, Equation 3.9 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 61, the pay reduction of - $29,000 / lane kilometer (-$46,690 / lane mile) is obtained. For truly excellent quality, PWL VA = PWL DEN = PWL AC = PWL VMA =100, PWL* =100, the pay equation awards a maximum bonus of 102

118 $10,000 / lane kilometer ($16,100/ lane mile). At the other extreme, when PWL VA = PWL DEN = PWL AC = PWL VMA =100, PWL* =0, the pay equation assigns the maximum pay reduction of - $90,000 / lane kilometer (-$ 144,900/ lane mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Four Quality Characteristics with Cross-Product Expected life model for Superpave pavements that include four quality characteristics was developed. Cross-product was included to investigate the difference between only addition and the one which has cross-product. Microsoft excel was used to solve 15 simultaneous equations. The final expected life model that includes air voids (VA), in-place density (DEN), asphalt content (AC), and VMA predicts an expected life of 2.20 year when PWL=0 for all four quality characteristics. The equations for expected life and composite index were not included since they were cumbersome. To determine the comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN = PWL AC =PWL VMA =90 are substituted into PWL* equation to obtain PWL* = 63 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 63. To determine the value of PWL* associated with the RQL, any combination of values that give 5- year life can be substituted into expected life equation. For example, entering PWL VA =73 and PWL DEN = PWL AC =PWL AC =70 into PWL* equation produces PWL*=22.5. Similarly, any combination of values that gives the 7.5-year life produces retest provision. Using PWL VA =80, PWL DEN =81.5, and PWL AC =PWL VMA =85 gives PWL*=43. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.19 was derived. Equation 3.19 in which PA = lot pay adjustment ($ / lane kilometer), and PWL* = composite quality measure. 103

119 It can be seen that when PWL* is at the AQL value of 63, Equation 3.19 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 22.5, the pay reduction of - $40,500 / lane kilometer (-$65,205 / lane mile) is obtained. For truly excellent quality, PWL VA = PWL DEN = PWL AC = PWL VMA =100, PWL* =100, the pay equation awards a maximum bonus of $37,000 / lane kilometer ($59,570/ lane mile). At the other extreme, when PWL VA = PWL DEN = PWL AC = PWL VMA =100, PWL* =0, the pay equation assigns the maximum pay reduction of - $63,000 / lane kilometer (-$101,430 / lane mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Five Quality Characteristics with Cross-Product Composite index (PWL*) was developed in terms of air voids (VA), in-place density (DEN), smoothness (SM), asphalt content (AC), and voids in mineral aggregates (VMA). The coefficients were obtained using the data in Table The magnitudes of the coefficients reflect the effect of the variables on the long term performance of the pavements. The coefficients may be modified based on Hamburg wheel tests at Kansas State University, field performance, and/or KDOT s experience. Composite index varies from zero to 100%. The final model developed is shown in Equation Equation 3.20 To determine comparable value of PWL* associated with the AQL, the values of PWL VA =PWL DEN = PWL SM =PWL AC =PWL VMA =90 are substituted into Equation 3.20 to obtain PWL* = 90 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 90. To determine the value of PWL* associated with the RQL, any combination of values that give 5-year life can be substituted into Equation 3.5. For example, entering PWL VA =PWL DEN = PWL SM =PWL AC =65 and PWL VMA =70 into Equation 3.20 produces 104

120 PWL*=66. Similarly, any combination of values that gives the 7.5-year life produces retest provision. Using PWL VA = PWL DEN = PWL SM =PWL AC =85, and PWL VMA =83 gives PWL*=84.5. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.10 was derived. It can be seen that when PWL* is at the AQL value of 90, Equation 3.10 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 66, the pay reduction of - $24,000/lane-kilometer (-$38,640/lane-mile) is obtained. For truly excellent quality, PWL VA = PWL DEN = PWL SM =PWL AC = PWL VMA =100, PWL* =100, the pay equation awards a maximum bonus of $10,000 / lane kilometer ($16,100/ lane mile). At the other extreme, when PWL VA = PWL DEN = PWL SM =PWL AC = PWL VMA =100, PWL* =0, the pay equation assigns the maximum pay reduction of -$90,000/lane kilometer (-$144,900/ lane-mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment. Expected life model and composite equation that include cross-product of quality characteristics were not developed for five quality characteristics since it became too cumbersome PCC Pavements Composite index for PCC pavements that includes two and three quality characteristics was developed. Composite index without and with cross-product of quality characteristics was considered Two Quality Characteristics without Cross-Product Composite index (PWL*) was developed in terms of thickness (TH) and strength (ST). The coefficients were obtained using the data in Table The magnitudes of the coefficients reflect the effect of the variables on the long term performance of the pavements. The coefficients may be modified based on field performance and/or agency s experience. Composite index varies from zero to 100%. The final model developed is shown in Equation Equation 3.21

121 To determine the comparable value of PWL* associated with the AQL, the values of PWL TH =PWL ST =90 are substituted into Equation 3.21 to obtain PWL* = 90 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 90. To determine the value of PWL* associated with the RQL, any combination of values that give 10-year life can be substituted into Equation 3.6. For example, entering PWL TH =65 and PWL ST =67 into Equation 3.21 produces PWL*=66. Similarly, any combination of values that gives the 15-year life produces retest provision. Using PWL TH =79 and PWL ST =80 gives PWL*=79.5. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.10 was derived. It can be seen that when PD* is at the AQL value of 90, Equation * produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 66, the pay reduction of - $24,000/lane-kilometer ($-38,640/lane-mile) is obtained. For truly excellent quality, PWL TH = PWL ST =100, PWL* =100, the pay equation awards a maximum bonus of $10,000/lane-kilometer ($16,100/lane-mile). At the other extreme, when PWL TH = PWL ST =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$90,000/lane-kilometer (-$144,900/lane-mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Two Quality Characteristics with Cross-Product Data in Table 3.45 was used to develop expected life that includes two quality characteristics for PCC pavements. Cross-product was included to investigate the difference between only addition and the one which has cross-product. Microsoft excel was used to solve four simultaneous equations. The final expected life model that includes thickness (TH) and strength (ST) is shown by Equation The model predicts an expected life of 0.50 year when PWL=0 for both quality characteristics. Equation 3.22 was converted into composite index in terms of PWL* as shown in Equation Equation 3.23 gives PWL* ranging from 0 to 100%. Equation

122 Equation 3.23 TABLE 3.45 Data for Composite Index for PCC Pavements (Two Variables, V=2) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) (years) To determine comparable value of PWL* associated with the AQL, the values of PWL TH =PWL ST =90 are substituted into Equation 3.23 to obtain PWL* = 80.5 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = To determine the value of PWL* associated with the RQL, any combination of values that give 10- year-life can be substituted into Equation For example, entering PWL TH =64 and PWL ST =63.5 into Equation 3.23 produces PWL*=39.5. Similarly, any combination of values that gives the 15-year life produces retest provision. Using PWL TH =80 and PWL ST =75.5 gives PWL*=60. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.24 was derived. Figure 3.24 in which PA = lot pay adjustment ($ / lane kilometer), and PWL* = composite quality measure. It can be seen that when PWL* is at the AQL value of 80.5, Equation 3.24 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 39.5, the pay reduction of - $41,000/lane-kilometer (-$66,010/lane-mile) is obtained. For truly excellent quality, PWL TH = PWL ST =100, PWL* =100, the pay equation awards a maximum bonus of $19,500/lane-kilometer 107

123 ($31,395/lane-mile). At the other extreme, when PWL TH = PWL ST =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$80,500/lane-kilometer (-$129,605/lane-mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Three Quality Characteristics without Cross-Product Composite index (PWL*) was developed in terms of thickness (TH), strength (ST), and air content (VA). The coefficients were obtained using the data in Table The magnitudes of the coefficients reflect the effect of the variables on the long term performance of the pavements. The coefficients may be modified based on field performance and/or agency s experience. Composite index varies from zero to 100%. The final model developed is shown in Equation Equation 3.25 To determine the comparable value of PWL* associated with the AQL, the values of PWL TH =PWL ST = PWL VA =90 are substituted into Equation 3.25 to obtain PWL* = 90 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 90. To determine the value of PWL* associated with the RQL, any combination of values that give 10- year-life can be substituted into Equation 3.7. For example, entering PWL TH =72, PWL ST =72.5, and PWL VA =70 into Equation 3.25 produces PWL*=71.5. Similarly, any combination of values that gives the 15-year life produces retest provision. Using PWL ST = 82.5, PWL ST =83, and PWL SM =80 gives PWL*=82.0. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.10 was derived. When PWL* equals the RQL value of 71.5, the pay reduction of -$18,500/lane-kilometer (-$29,785/lane-mile) is obtained. For truly excellent quality, PWL TH = PWL ST = PWL VA =100, PWL* =100, the pay equation awards a maximum bonus of $10,000/lane-kilometer ($16,100/ lane-mile). At the other extreme, when PWL TH = PWL ST = PWL VA =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$90,000/lane-kilometer (-$144,900/lane-mile). 108

124 In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment Three Quality Characteristics with Cross-Product Data in Table 3.46 was used to develop expected life for three variables. Cross-product was included to investigate the difference between only addition and the one which has crossproduct. Microsoft excel was used to solve eight simultaneous equations. The final expected life model that includes thickness (TH), strength (ST), and air content (VA) is shown by Equation The model predicts an expected life of about 2.1 years when PWL=0 for all three quality characteristics. Equation 3.26 was converted into composite index in terms of PWL* as shown in Equation Equation 3.27 gives PWL* ranging from 0 to 100%. Equation 3.26 Equation

125 TABLE 3.46 Data for Composite Index for PCC Pavements (Variables, V=3) Percent Within Limit (PWL) for Various Quality Measures Expected Life Thickness (TH) Strength (ST) Air Voids (VA) (years) To determine the comparable value of PWL* associated with the AQL, the values of PWL TH =PWL ST = PWL VA =90 are substituted into Equation 3.27 to obtain PWL* = 72 as the AQL. Therefore, the pay equation must produce a pay adjustment of zero at PWL* = 72. To determine the value of PWL* associated with the RQL, any combination of values that give 10- year life can be substituted into Equation For example, entering PWL TH =PWL ST = 70, and PWL VA =64.5 into Equation 3.27 produces PWL*=32. Similarly, any combination of values that gives the 15-year life produces retest provision. Using PWL TH =PWL ST =81, and PWL VA =80.5 gives PWL*=52. Assuming a simple linear pay equation will be sufficient, the pay schedule given by Equation 3.28 was derived. Equation 3.28 in which PA = lot pay adjustment ($ / lane kilometer), and PWL* = composite quality measure. It can be seen that when PWL* is at the AQL value of 72, Equation 3.28 produces a pay adjustment of zero. Similarly, when PWL* equals the RQL value of 32, the pay reduction of - 110

126 $40,000/lane-kilometer (-$64,400/lane-mile) is obtained. For truly excellent quality, PWL TH = PWL ST = PWL VA =100, PWL* =100, the pay equation awards a maximum bonus of $28,000/lane-kilometer ($45,080/lane-mile). At the other extreme, when PWL TH = PWL ST = PWL VA =0, PWL* =0, the pay equation assigns the maximum pay reduction of -$72,000/lanekilometer (-$115,920/lane-mile). In between, all pay adjustments are related to performance in that all quality levels that give any particular life will receive the same level of payment. 111

127 4.1 Conclusions Chapter 4: Conclusions and Recommendations Based on this study, the following conclusions can be made: Moving average control chart analysis does not clearly show any systematic bias in QC/QA data for Superpave and PCC pavements in Kansas. QC mean, minimum and maximum density values are higher than QA mean, minimum, and maximum density whereas as QC Standard Deviation and COV for density are lower than QA Standard Deviation and COV for density. Lot-wise comparison shows that QC/QA means are significantly different in most cases. One reason for these differences could be due to varying number of tests per lot done by the contractor and KDOT (contractor does more tests). The number of cases with a significant difference in means increases with an increase in significance level. However, statistical analysis did not show any specific trend in sublot mean data as far as order of data collection is concerned. F-test can be used to determine significant difference in means. The consequence has equal impact on both KDOT and the contractor. Currently majority of Superpave projects in Kansas are built with asphalt content lower than design. Asphalt content can be included in pay adjustment. However, provisions should be made in which contractors may be penalized for too much asphalt to avoid flushing or too low asphalt to avoid dry mixes. The performance model and composite index for both Superpave and PCCP pavements can be derived with multiple quality characteristics based on percent within limits (PWL). However, composite index with cross-product of quality characteristics gives a more realistic pay adjustment. 4.2 Recommendations Based on this study, the following recommendations can be made: There was no QC smoothness data to compare with QA smoothness data in this study. It is recommended that KDOT collects some QA smoothness data to verify 112

128 QC smoothness data in the future. The frequency of QA smoothness data collection needs to be established too. Ten QC readings and five QA readings per lot will be sufficient for statistical analysis. Bigger lot size or smaller sublots can be used based on economy and convenience though smaller sublots are recommended from statistical point of view. Under current KDOT practices, QC mean, minimum and maximum density values are higher whereas Standard Deviation and COV for QC density are lower. It is recommended that a procedure be developed to collect QC and QA data that have better statistical agreement. It is also recommended that selected projects be required that pavement cores be the only basis for acceptance to see if better QC and QA data agreement can be obtained. As an alternative, nuclear devices that have GPS and continuous data recording capabilities be used to measure densities. It is recommended that KDOT encourage the contractors to produce Superpave mix at an asphalt content that equals or exceeds the asphalt content used in the approved design. Pay adjustment at 1% significance level is less than or equal to the pay adjustment at other significant levels. It is recommended that 2.5% be used as significance level as a compromise between 1 and 5% instead of current 1%. Khanum et al. (2006) concluded that current KDOT PWL specifications for PCC pavement construction are more sensitive to the concrete strength than to the PCC slab thickness. This shows that PCCP strength higher than the specified strength will result in large bonus payment whereas the gain in performance due to the higher PCC strength may not be significant. It is recommended to make adjustments to the current combined pay equation for PCCP that includes strength and thickness to deemphasize the strength component. It is recommended to validate and/or improve practical performance models and composite index based on laboratory tests, field performance, and/or experience before starting to use for pay adjustment. 113

129 References Afferton, K.C, Freidenrich, J., and Weed, R.M Managing Quality: Time for a National Policy. Transportation Research Record: Journal of the Transportation Research Board, No. 1340, Transportation Research Board of the National Academies, Washington, D.C., pp Benson, P.E Comparison of End Result and Method Specifications for Managing Quality. Transportation Research Record: Journal of the Transportation Research Board, No. 1491, Transportation Research Board of the National Academies, Washington, D.C., pp Benson, P.E., Chong, Y.S., and Samaniego, F.J Nonparameteric Approach to Managing Materials Quality. Transportation Research Record: Journal of the Transportation Research Board, No. 1712, Transportation Research Board of the National Academies, Washington, D.C., pp Burati, L., Weed, R.M., Hughes, C.S. and Hill, H.S Optimal Procedures for Quality Assurance Specifications. Final Report No. FHWA-RD , Federal Highway Administration (FHWA), McLean, VA. Diwan, R.M., Shah, S., and Eggers, J Statistical Quality Control and Quality Assurance Evaluation of Structural & Paving Concrete. Transportation Research Record: Journal of the Transportation Research Board, No. 1861, Transportation Research Board of the National Academies, Washington, D.C., pp Elseifi, M.A., Trepanier, J., Wakefield, H., Pine, W.J., and Dahhan, A Hot-Mix Asphalt Sampling Techniques and Methods of Acceptance State DOTs. Transportation Research Board 88 th Annual Meeting Compendium of Papers DVD, Transportation Research Board of the National Academies, Washington, D.C. Khanum, T., J. W. Neil, M. Hossain, and L. S. Ingram Effect of JPCP Performance on PCC Quality Control/Quality Assurance Specifications. Transportation Research Board 85 th Annual Meeting Compendium of Papers, CD-ROM, Transportation Research Board of the National Academies, Washington, D.C. NCHRP Performance-Related Specifications for Highway Construction and Rehabilitation, NCHRP Synthesis of Highway Practice 212, Transportation Research Board, National Research Council, Washington, D.C. 114

130 Parker, F. and Hossain, M.S Statistics for Superpave HMA Construction QC/QA Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1813, Transportation Research Board of the National Academies, Washington, D.C., pp Schell, H. and Konecny, J Development of an End-Result Specification for Air Void Parameters of Hardened Concrete in Ontario s Highway Structures. Transportation Research Board 82 nd Annual Meeting Compendium of Papers, CD-ROM, Transportation Research Board of the National Academies, Washington, D.C. Schmitt, R. L., Russell, J.S., Hanna, A.S., Bahia, H.U., and Jung, G.A Summary of Current Quality Control/Quality Assurance Practices for Hot-Mix Asphalt Construction. Transportation Research Record: Journal of the Transportation Research Board, No. 1632, Transportation Research Board of the National Academies, Washington, D.C., pp Shah, S.C An Integrated and Computerized Quality-Assurance System. Transportation Research Record: Journal of the Transportation Research Board, No. 613, Transportation Research Board of the National Academies, Washington, D.C., pp Turochy, R.E. and Parker, F Comparison of Contractor and State Transportation Agency Quality Assurance Test Results on Mat Density of Hot-Mix Asphalt Concrete: Findings of a Multi-State Analysis. Transportation Research Record: Journal of the Transportation Research Board, No. 2040, Transportation Research Board of the National Academies, Washington, D.C., pp Weed, R.M Statistical Specification Development. 2 nd Edition. Report No. FHWA/NJ New Jersey Department of Transportation, Trenton. Weed, R.M Development of Composite Quality Measures. Transportation Research Record: Journal of the Transportation Research Board, No. 1712, Transportation Research Board of the National Academies, Washington, D.C., pp Weed, R.M Derivation of Equation for Cost of Premature Pavement Failure. Transportation Research Record: Journal of the Transportation Research Board, No. 1761, Transportation Research Board of the National Academies, Washington, D.C., pp Weed, R.M Mathematical Modeling of Pavement Smoothness. Transportation Research Record: Journal of the Transportation Research Board, No.1813, Transportation Research Board of the National Academies, Washington, D.C., pp

131 Weed, R.M Multicharacteristic Performance-Related Specification for Hot-Mix Asphalt Pavement: Complete Development Process. Transportation Research Record: Journal of the Transportation Research Board, No. 1861, Transportation Research Board of the National Academies, Washington, D.C., pp Weed, R.M., and Tabrizi, K Conceptual Framework for Pavement Smoothness Specification. Transportation Research Board 84 th Annual Meeting Compendium of Papers, CD-ROM, National Research Council, Washington, D.C. Weed, R.M Mathematical Modeling Procedures for Performance-related Specifications. Transportation Research Record: Journal of the Transportation Research Board, No. 1946, Transportation Research Board of the National Academies, Washington, D.C., pp

132