Correlation Study on the Falling Weight Deflectometer and Light Weight Deflectometer. for the Local Pavement Systems. A thesis presented to

Transcription

1 Correlation Study on the Falling Weight Deflectometer and Light Weight Deflectometer for the Local Pavement Systems A thesis presented to the faculty of the Russ College of Engineering and Technology of Ohio University In partial fulfillment of the requirements for the degree Master of Science Ahmadudin Burhani August Ahmadudin Burhani. All Rights Reserved.

2 2 This thesis titled Correlation Study on the Falling Weight Deflectometer and Light Weight Deflectometer for the Local Pavement Systems by AHMADUDIN BURHANI has been approved for the Department of Civil Engineering and the Russ College of Engineering and Technology by Shad M. Sargand Russ Professor of Civil Engineering Dennis Irwin Dean, Russ College of Engineering and Technology

3 3 ABSTRACT BURHANI, AHMADUDIN, M.S., August 2016, Civil Engineering Correlation Study on the Falling Weight Deflectometer and Light Weight Deflectometer for the Local Pavement Systems Director ofthesis: Shad M. Sargand The Falling Weight Deflectometer (FWD) and Light Weight Deflectometer (LWD) are essential nondestructive devices used for structural evaluation and characterization of pavement layer systems. This study evaluated the performances of both devices in 99 different test sites grouped into five clusters located in eight counties in Ohio. The structural adequacy of the local roads in Ohio was assessed by conducting field tests using deflectometry and backcalculation techniques. A field research program consisting of a series of FWD and LWD tests was undertaken at the same locations to investigate local pavement performances. The deflection data obtained from test results corresponding to pavement material properties were used to estimate: in-situ stiffness layer moduli, effective structural numbers, and a range of structural coefficients for different materials utilized to widen, construct, and rehabilitate county roads in Ohio. AASHTO 1993 Guide for Design of Pavement Structures and computer software, Modulus 6.0, Evercalc 5.0 were chosen to perform the backcalculation analysis. Specifically, this study investigated the feasibility and potential use of the Prima 100 LWD as in-situ testing device on the local roads. Although the FWD device could be used for the evaluation of the county roads, the cost of the equipment is prohibitive for most local agencies. The Prima 100 LWD on the other hand proved to be reasonable and

4 4 effective alternative. However, the application of Prima 100 LWD requires a methodological correlation with respect to benchmark test. Comparisons were made through comprehensive regression analyses using the SPSS software. Center and radial offset sensor deflections as well as backcalculated layer moduli, layer coefficients, and the effective structural numbers were compared. The correlation results for the layer coefficients and subgrade modulus across all test sites were improved by the Rohde method. The results demonstrated consistent relationship between both devices on the evaluation for the asphalt and concrete surfaces. However, lower relationship for sensor deflections was reported for aggregate overlay, full depth grinding, and soft soil surfaces. In the course of this study, a modified relationship between deflection basin parameter and pavement response was devised. This promising relationship is the Area Under Pavement Profile (AUPP) which can be used to predict tensile strain at the bottom of the asphalt concrete layer. The statistical analyses showed the proposed procedure appears to be a new valid parameter for the pavement evaluation using LWD sensor deflections. In the final analysis, the Prima 100 LWD proved to be an effective and economically viable test procedure for asphalt and concrete surfaces for the evaluation of local pavement systems.

5 5 DEDICATION I dedicate this work to my family for giving me support and encouragement throughout my career

6 6 ACKNOWLEDGMENTS First of all, I would like to express sincere appreciation and gratitude to my academic advisor Professor Shad M. Sargand for his continuous support and guidance throughout my entire research. Your devotion, encouragement, and advice helped me in realizing my potential and I appreciated any single minute spent on this adventure. Next, I specifically would like to extend my appreciation and thanks to the rest of my thesis committee: Dr. Teruhisa Masada, Dr. Issam Khoury, and Dr. Tatiana Savin for agreeing to be my committee member and for their supportive comments. Also, I give my deepest thanks to Mr. Roger Green and Mr. Benjamin Jordan for their continuous cooperation and assistance during my research. Without their supports, this thesis may not be completed. Finally, I also would like to thank all my colleagues and civil engineering family in Ohio University for making my study a memorable adventure here in Athens. I further give my deepest gratitude and thanks to my family who always encouraged, supported and loved me. Without their help, I would be unable to accomplish my goals.

7 7 TABLE OF CONTENTS Page Abstract... 3 Dedication... 5 Acknowledgments... 6 List of Tables List of Figures Chapter 1 Introduction Overview Research Goal and Objectives Outline of Thesis...24 Chapter 2 Literature Review Introduction The Falling Weight Deflectometer (FWD) Dynatest Model 8000 FWD KUAB America Carl Bro FWD JILS FWD The Light Weight Deflectometer (LWD) Prima 100 LWD The LWD Principle of Operation Existing Correlations between FWD and LWD...38

8 8 2.5 Determination of Pavement Responses Using Deflection Basin Parameter Backcalculation of Layer Moduli Overview of Backcalculation Software Modulus Program Evercalc Program Chapter 3 Evaluation of Pavement Condition Using FWD and LWD Measurements Field Testing Quantifying Pavement Condition Using FWD Deflections FWD Results Quantifying Pavement Condition Using LWD Deflections LWD Results Backcalculation Methodology and Pavement Layer Moduli AASHTO Method (Section 5.4.5, FWD) Determining Layer Coefficients from AASHTO Equations AASHTO Method (Section 2.3.5, LWD) Rohde s [1994] Method of Determination of Pavement Structural Number and Subgrade Modulus from FWD Testing Pavement Layer Moduli Chapter 4 Results and Discussion Introduction Regression Analysis Comparison FWD and LWD Sensor Deflections...96

9 Deflections at the Center of Loading plate, (D0) Deflections at Radial Offset Distance r = 300mm, (D1) Deflections at Radial Offset Distance r = 600mm, (D2) Area Under Pavement Profile (Deflection Basin Parameter) Comparison of Backcalculated Layer Moduli Comparison of Subgrade Moduli Comparison of Layer Coefficients Comparison of Effective Structural Numbers Chapter 5 Conclusion and Recommendations Summary Conclusion Recommendations References Appendix A: Pavement Layer Thicknesses and Material Properties by County Appendix B: Typical FWD and LWD Deflection Basins Appendix C: AASHTO Procedure Outputs Using FWD Sensor Deflections Appendix D: Summary of Backcalculated Layer Moduli from FWD and LWD Testing Appendix E: FWD and LWD Sensor Deflections Appendix F: Effective Structural Numbers of AASHTO Equations and The Rohde Method

10 10 LIST OF TABLES Page Table 2.1: Sensor Spacing of the FWD Device (FHWA, 2009 & Dynatest, 1995) Table 2.2: Physical Characteristics of Typical LWD Devices (Mooney & Miller, 2009) 35 Table 2.3: Regression Analysis Between FWD & LWD Moduli, (Shafiee, et al., 2013) 42 Table 2.4: Typical Poisson s Ratio Values, (ASTM D5858, 2003) Table 2.5: Existing Backcalculation Software (Adapted from Appea et al., 2003) Table 3.1: Ohio County Roads by Cluster and Construction Material Used Table 3.2: Prima 100 LWD Sensor Deflection Measurements for Cluster # Table 3.3: Representation of Backcalculation Procedure (Murillo & Bejarano, 2013) Table 3.4: Calculated Layer Coefficients Range Based on Material Types, AASHTO Table 3.5: Calculated Layer Coefficients Range Based on Material Types, AASHTO LWD Table 3.6: Coefficient for Structural Number versus SIP Relationships, (ROHDE, 1994) Table 3.7: Coefficient for E versus SIS Relationship, (Rohde, 1994) Table 3.8: Effective Structural Numbers and Subgrade Modulus from Rohde Procedure83 Table 3.9: Calculated Layer Coefficients Range Based on Material Types, Rohde [1994] Method Table 4.1: Statistical Analysis Model Summary of FWD vs. LWD Sensor Deflections (D2)

11 11 Table 4.2: Statistical Analysis, Model Summary of FWD & LWD Procedures Table 4.3: Summary of Regression Analysis of FWD versus LWD Generated from Developed Models Table A1: Layer Thicknesses and Material Properties, Defiance County Table A2: Layer Thicknesses and Material Properties, Harrison County Table A3: Layer Thicknesses and Material Properties, Carroll County Table A4: Layer Thicknesses and Material Properties, Auglaize County Table A5: Layer Thicknesses and Material Properties, Mercer County Table A6: Layer Thicknesses and Material Properties, Champaign County Table A7: Layer Thicknesses and Material Properties, Madison County Table A8: Layer Thicknesses and Material Properties, Muskingum County Table C1: AASHTO Equations Outputs Calculated from FWD Sensor Deflection Using 11.8-in. (300mm) Plate Table D1: Summary of Averaged Backcalculated Layer Moduli Computed from FWD Sensor Deflections Using 11.8-in. (300mm) Plate, Modulus 6.0 Software Table D2: Summary of Averaged Backcalculated Layer Moduli Computed from LWD Sensor Deflections Using 11.8-in. (300mm) Plate, Evercalc 5.0 Software Table E1: Normalized/Extrapolated to 9000 Pounds Sensor Deflections (D0, D1, and D2) at Radial Offset Distance 0, 12, 24 inches from the Center of the Load Table E2: Deleted Outliers/ Abnormal Sensor Deflections Obtained from FWD and LWD Testing

12 12 LIST OF FIGURES Page Figure 2.1: Haversine Loading Applied by FWD in Defiance, Section C146-Krouse Road Figure 2.2: Falling Weight Deflectometer Schematic (Ferne & Langdale, 2010) Figure 2.3: Dynatest Model 8000 FWD (LRTC 2000 & Nazzal, 2003) Figure 2.4: Schematic of Prima 100 with Additional Geophones, (Senseney, 2010) Figure 2.5: Typical Time History Data from LWD Test (Mooney & Miller, 2009) Figure 2.6: Best Fit Model of Fleming (2000) & Nazzal (2003) Figure 2.7: Area Under Pavement Profile (Adopted from Thompson, 1989) Figure 2.8: Backcalculation Flowchart (Lytton, 1989) Figure 2.9: Relationship Between Deflection and Modulus (Tawfiq, 2003) Figure 3.1: Ohio Counties Map (Adapted from ORIL, 2015) Figure 3.2: Typical Pavement Surface Deflection Basins Based on Load Levels, Champaign County, and Section Pisgah Road (C236-3) Figure 3.3: Coring and Obtaining Samples, form One of Tested Section Figure 3.4: FWD Deflection Basins, Various Loads, Cluster # 3, Section of Southland Road (Aug-C3-15), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate Figure 3.5: FWD Deflection Basins, Various Loads, Meter Road (CAR-T269-2), Aggregate Overlay Surface Layer, Carroll County, 11.8-in. (300-mm) Plate Figure 3.6: Conducting Tests on Pavement Surface Sections in Defiance... 63

13 13 Figure 3.7: Example of a LWD Output from Field Testing, Auglaize County, Section of Minster Fort Recovery Road, (Aug-C30-16) Figure 3.8: LWD Deflection Basins, Same Loads, Cluster # 3, Southland Road (Aug-C3-15), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate Figure 3.9: Box Plot of Layer Coefficients for Each Widening/Construction Treatment, Layer Type Based on AASHTO Figure 3.10: Chart for Estimating Structural Layer Coefficient of Asphalt Concrete (AASHTO, 1993) Figure 3.11: Used Chart for Cement-Treated Base Materials, (AASHTO, 1993) Figure 3.12: Box Plot of Layer Coefficients for Each Widening/Construction Treatment, Layer Type Based on AASHTO Figure 3.13: Box Plot Showing Layer Coefficients for Each Widening/Construction Treatment as Determined Using Rohde [1994] Procedure Figure 3.14: Evercalc 5.0 General File Data Entry Screen for Pisgah Road, Champaign County Figure 3.15: Evercalc 5.0 LWD Deflection File screen for Pisgah Road, Champaign County Figure 3.16: Evercalc 5.0 LWD Deflection Basin for Pisgah Road, Champaign County 89 Figure 3.17: Main Window of Modulus 6.0 (Liu and Scullion, 2001) Figure 3.18: Backcalculation Routine Window, Krouse Road, Defiance County Figure 3.19: Box Plot Showing Backcalculated Layer Moduli for Each Widening Treatment as Determined Using Modulus 6.0 Software, FWD Testing

14 14 Figure 3.20: Box Plot Showing Backcalculated Layer Moduli for Each Widening Treatment as Determined Using Evercalc 5.0 Software, LWD Testing Figure 4.1: Comparison Between FWD and LWD Deflections at the Center of Loading Plate, (D0) Figure 4.2: DFWD vs. dlwd Correlation, Comparison to, (Horak et al., 2008) Figure 4.3: Comparison of FWD and LWD Deflections at r = 300mm from the Center of Loading Plate, (D1) Figure 4.4: Comparison of FWD and LWD Deflections at r = 600mm from the Center of Loading Plate, (D2) Figure 4.5: AUPP (LWD 3 Sensors) Modified Deflection Basin Parameter Figure 4.6: AUPP Comparison of FWD and FWD across All Sites Figure 4.7: Backcalculated Layer Moduli of Pavement Layers Based on FWD and LWD Measurements Figure 4.8: Regression Analysis Fitting Linear Trendline to Data Points Figure 4.9: EFWD vs. ELWD, Comparison to Steinert et al. (2005), Nazzal (2003), and Fleming et al. (2000) Figure 4.10: FWD Measured Modulus of the Subgrade. Values Indicated are Minimum; Mean; and Maximum Respectively. (1 ksi = 6.89 MPa) Figure 4.11: LWD Measured Modulus of the Subgrade. Values Indicated are Minimum; Mean; and Maximum Respectively. (1 ksi = 6.89 MPa) Figure 4.12: Rohde Method Measured Modulus of the Subgrade. Values Indicated are Minimum; Mean; and Maximum Respectively. (1 ksi = 6.89 MPa)

15 15 Figure 4.13: Regression Analysis Fitting Linear Trendline to All Layer Coefficients Obtained from AASHTO FWD & AASHTO LWD Methods Figure 4.14: Regression Analysis Fitting Linear Trendline to All Layer Coefficients Obtained from AASHTO LWD and Rohde Method Figure 4.15: FWD vs LWD Layer Coefficient Models, Comparison to Rohde Method 116 Figure 4.16: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Defiance County Figure 4.17: Regression Model of Effective Structural Numbers Obtained from, the AASHTO Equations and the Rohde Method Figure B1: Deflection Basins for Three Loads, Cluster # 2, Section of Minster Recovery Road (Aug-C30-16), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate Figure B2: LWD Deflection Basins Same Loads, Meter Road (CAR-T269-2), Aggregate Overlay Surface Layer, Carroll County, 11.8-in. (300-mm) Plate Figure B3: FWD Deflection Basins for Three Loads, Cluster # 2, Section of East Shelby Road (Aug-C71-8), HMA Surface layer, Auglaize County, 11.8-in. (300-mm) Plate Figure B4: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Blank Pike Road (Aug-C160-12), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate Figure B5: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Kossuth Loop (Aug-C216A-3), Full depth Grindings layer, Auglaize County, and 11.8-in. (300-mm) Plate

16 16 Figure B6: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Fairground (Aug-FG-18), Full Depth Grindings Layer, Auglaize County, 11.8-in. (300-mm) Plate Figure B7: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Neptune Mendon Road (MER-C161C-7), HMA Surface Layer, Mercer County, 11.8-in. (300-mm) Plate Figure B8: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Harris Road (MER-C175B-8), HMA Surface Layer, Mercer County, 11.8-in. (300-mm) Plate Figure B9: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Dutton Road (MER-C230A-3), HMA Surface Layer, Mercer County, 11.8-in. (300-mm) Plate Figure B10: LWD Deflection Basins Same Loads, Cluster # 2, Kossuth Loop (Aug- C216A-3), Full Depth Grindings Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate Figure F1: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Auglaize County Figure F2: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Mercer County Figure F3: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Madison County Figure F4: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Champaign County

17 17 Figure F5: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Muskingum County Figure F6: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Carroll County Figure F7: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Harrison County

18 18 CHAPTER 1 INTRODUCTION 1.1 Overview A local road considered herein as low volume road which has approximately an average daily traffic (ADT) of less than 400 vehicles; design speed typically less than 50mph (80kph), and corresponding geometry (Keller & Sherar, 2003). A majority of local or low volume roads are experiencing growth in the annual average daily traffic due to increasing residential and commercial development (Sargand et al., 2016). Many county roads that fall under the low volume category still carry important levels of heavy vehicle traffic. As traffic grows, pavements have to be widened and/or strengthened in an effort to sustain the geometrics and structural integrity of the roadway. From a road way point of view there are numerous reasons such as economics, sustainability, and availability that many local engineers recommend and prefer to reuse the existing materials from the roadway or any available material such as recycled asphalt, recycled concrete, fly ash, and so forth. In addition, various construction methods such as full depth reclamation (which is an effective recycling procedure for low volume roads), white-topping, fabric reinforcement, and roller compacted concrete are used to strengthen or widen pavement. These methods are the keys to ensure that a local road meets the needs of the user, and is essential for community and infrastructure development. However, the load carrying capacity of these materials/methods techniques are unknown in Ohio (Sargand et al., 2016). Also, without structural inputs parameters, the thickness design of widening is not possible,

19 19 resulting in premature failure when placed too thin or an overly conservative design when placed too thick. Therefore, research was undertaken to develop structural input parameters for the pavement design/analysis based on AASHTO 1993 Guide for Design of Pavement Structures for the local road network, to ensure durability and adequately serve its users. The research evaluated structural condition of pavements using nondestructive test (NDT) technology. Also, evaluation of structural condition is one of the most important factors in pavements construction (AASHTO, 1993; Huang, 2004; Nazzal, 2003). Load carrying capacity for a pavement is highly related to pavement layer and pavement subgrade moduli. As a result, evaluating the local pavement conditions utilized to assess the structural adequacy of pavements and determining used materials properties must be considered significant in pavements construction. The current investigation of in-situ strength of various construction/widening methods utilized on local roads and evaluation of structural properties of pavements systems are based on field measurement using field tests to analyze and interpret structural properties of rural pavement performance. In 2015, a proposal for the Ohio Department of Transportation, Ohio Research Initiative for Locals (ORIL) program was tasked to establish and verify a low cost, nondestructive, repeatable methodology to characterize the load carrying capacity of materials used in road construction when established values are unavailable. The research was included field investigations to provide resilient moduli or a range of structural coefficients for different materials utilized to widen, rehabilitate, or construct roads on Ohio's lowvolume road pavement system. The results of the research can be used by local officials to

20 20 enhance their knowledge and understanding of the potential structural integrity of considered materials for use in roadway construction, maintenance, and improvement projects. This can also lead to a more efficient design and greater confidence in the load carrying capacity of the local roads. In addition, it can establish a rational basis for material selection to correlate with the readily available cost data, which will aid locals in managing budgets and ensuring the fiscal integrity of local pavement preservation programs, (ORIL, 2015). The ORIL (2015) was tasked to investigate a total of 99 different test sites grouped into five clusters, located in eight different counties (Defiance, Champaign, Mercer, Auglaize, Muskingum, Madison, Carroll, and Harrison) around the state of Ohio were used in the study. Field testing techniques for evaluation of paved and unpaved low volume roads were investigated. The field components included traveling across Ohio to perform site investigations, collecting deflection data, coring and measuring pavement layer thicknesses, and collecting samples for performing laboratory experiments. The following field tests were conducted to analyze and interpret local pavement performance: 1. Falling Weight Deflectometer (FWD) 2. Dynamic Cone Pentrometer (DCP) 3. Light Weight Deflectometer (LWD) 4. Portable Seismic Property Analyzer (PSPA) This thesis work investigated the use of the Falling Weight Deflectometer (FWD) and Light Weight Deflectometer (LWD) on the low-volume roads. In order to physically investigate low-volume roads layer system, the Dynamic Cone penetration (DCP) was

21 21 employed to determine material properties and layer thickness. The Portable Seismic Property Analyzer (PSPA) was used to evaluate low-volume road surface layers, but its results were covered in another thesis. Also, this study further documented the results from all the test sites. Both the FWD and LWD were employed to measure deflection at the same spot of each single location. A minimum of three (3) locations at each site were included in this evaluation in order to develop better widening, rehabilitation, and construction strategies for each county road based on material properties. The results was used to investigate the utilization of the LWD (a lower cost technique to evaluate pavement condition) with respect to conventional benchmark test, the FWD technology. The FWD test (a commercially available nondestructive technique) utilizes radial offset surface deflection measurements to evaluate pavement layer condition and backcalculate layer moduli (Mooney et al., 2015). It is significant to determine the relationships between FWD and LWD in order to provide the county engineers a low cost alternative to the FWD for pavement layer analysis. In selecting the best correlation, it is important to consider statistical analysis of the deflection data obtain from the sensors measurements. Herein, regression analyses were used to determine the best fitting trendline to the models corresponding to sensor deflection data. Also, the Statistical Package for the Social Sciences (SPSS) was performed to determine whether the LWD is a valid structural test for local pavement systems. Resultantly, statistical analyses demonstrate best correlations between FWD and LWD. Several site and material specific relationship of composite moduli between FWD and LWD have been conducted (Mooney et al., 2015).

22 22 However, Horak et al. (2008) and Mooney et al. 2015) are the only two studies that compares radial offset deflection data. Also, upon demonstration of close relationships between FWD and LWD sensor deflections, the author would be interested to investigate/modify the Area under Pavement Profile (AUPP), proposed by Hill and Thompson (1988). This modification at radial offset distances 0, 12, 24 inches (0, 300, and 600mm) from the load center, now appears to be a new valid parameter in the pavement evaluation using LWD investigation. The AASHTO (1993), a guide for design of pavement structures, allows the use of measured deflections to evaluate pavements conditions. AASHTO section equations were used to calculate effective structural numbers and layer coefficients using FWD measurement, and AASHTO section procedure were used for the LWD measurements. These procedures are further processed to confirm by the Rohde method (explained in chapter three) using FWD measurements. Also, the deflection data are then used to evaluate the pavement stiffness in terms of layer modulus. This layer modulus obtained from FWD and LWD measurements is termed as backcalculated layer modulus. Numerous commercial software are available in order to analysis nondestructive testing data to obtain backcalculated layer modulus. Two independent software applications, MODULUS 6.0 and EVERCALC 5.0, were used in this study. Due to feasibility and sensors adjustment capability of Evercalc 5.0 with LWD deflection data, the Evercalc 5.0 was used to analyze LWD data. Also, the Modulus 6.0 is capable of producing reliable results from FWD deflection data. Thus, Modulus 6.0 was chosen in this study to investigate pavement condition (Al-Jhayyish, 2014). Lastly, the

23 23 backcalculated layer moduli have a significant input in the determination of effective structural number (SNeff), and have also been used to determine the remaining life of the pavement performance, therefore, the role of layer modulus is highly important in the local pavement evaluation within this study. 1.2 Research Goal and Objectives This thesis has two main goals: The first goal is to determine the structural adequacy of the low-volume road pavement systems using nondestructive test (NDT) technology. This is achieved by conducting field tests on local pavement systems. To this end, the obtained deflection data from nondestructive tests conducted with the FWD and the LWD based on the material properties was used to estimate layer moduli, effective structural number. Thereafter, a range of structural coefficients for different materials utilized to widen/construct low-volume road pavement system was calculated. The second goal is to investigate the feasibility of employing the Light Weight Deflectometer (LWD) as an in-situ testing device for the low-volume road pavements which were earlier evaluated during the first goal activities. To accomplish this goal, a comprehensive regression analysis was conducted between FWD and LWD sensor deflections at various radial offset distances, developing a new method for evaluation of Area under Pavement Profile (AUPP), the in-situ stiffness moduli, layer coefficients, and effective structural numbers. The major objectives of this study are described below: 1. Evaluate low-volume road pavement conditions using non-destructive testing devices, namely the FWD and LWD.

24 24 2. Analyze the deflection data to backcalculate layer moduli using various backcalculation software. 3. Evaluate the analytical procedures to characterize the load carrying capacity of materials used in road construction when established values are unavailable. 4. Compare/correlate FWD and LWD results for a single spot for at least three different applied loading in each test location/section in order to find their consistency. 5. Document and explain the differences in the results of FWD and LWD on the local pavement evaluation methods. 6. Modifying a relationship (Area Under Pavement Profile) between deflection basin parameter and pavement response to determine the tensile strain at the bottom of an asphalt layer. 7. Perform statistical analysis to determine whether the LWD is a valid structural testing device for low-volume road pavement systems. 8. Using the Rohde method to improve the correlation between FWD and LWD. 1.3 Outline of Thesis This thesis is organized into five chapters and six appendices to effectively present the data and information in the following format. 1. Chapter One is a brief introduction to the evaluation of the structural pavement performance of low volume roads in Ohio. Also, this chapter further explains the principal objectives of the research.

25 25 2. Chapter Two provides a literature review on the Falling Weight Deflectometer and Light Weight Deflectometer. It also offers a short review of existing correlation study between FWD/LWD, backcalculation methodologies of layer moduli, and available commercial backcalculation programs. 3. Chapter Three includes the methodologies for the evaluation of pavement condition based on material properties from the FWD and LWD deflection data. This chapter considers in the AASHTO 1993 equations in order to determine effective structural numbers, layer coefficients, and backcalculated layer moduli. It also presents the Rohde method to determine effective structural numbers and subgrade modulus. 4. Chapter Four presents the results and discussions of the correlation study between FWD and LWD. This chapter also includes the statistical analysis (regression models), which were conducted to ascertain the best correlations. 5. Chapter Five draws and summarizes the conclusions from the results and provides recommendations for future studies on FWD and LWD.

26 26 CHAPTER 2 LITERATURE REVIEW 2.1 Introduction Nondestructive testing methods for pavement evaluation was developed by Waterways Experiment Station (WES) of the U.S Army Corps of Engineers in the mid s (Grau & Alexander, 1994). The use of non-destructive testing for the evaluation of pavement structural performance is increasing worldwide. Numerous studies have been conducted in the past years to determine pavement structural capacity. Since its inception in the 1960 s the Falling Weight Deflectometer (FWD) has become a nondestructive test that plays a significant role in the pavement engineering. The Light Weight Deflectometer (LWD), developed in early 1981, is another portable device for evaluating pavement layer system (Mooney et al., 2015; Mooney & Miller 2009; Fleming et al. 2009; Siekmeier et al. 2009; Vennapus & White 2009). Since then, various methods have been developed using FWD and LWD deflection data to investigate structural condition of pavement layers. This chapter focuses on background of nondestructive devices, and common methodologies that could be applied to their deflection data analyses. 2.2 The Falling Weight Deflectometer (FWD) The Falling Weight Deflectometer (FWD) is a non-destructive test device that can exert an impulse load into the pavement layer system. It mearues deflections at several distances from the applied load on the pavement surfaces. The FWD has been broadly used in pavement engineering to investigate pavement structural behaviors. It is a trailer or bed mounted truck system. The FWD is able to load asphalt pavement or concrete surfaces in a way that simulates real wheel loads in both magnitude and duration. As the name implies,

27 27 the FWD imparts a specified weight (usually 110 to 660 lbs (0.48 to 3.0 KN)) by raising the weight hydraulically and then dropped it with a buffer system into a standard 11.8 inches (300 mm) diameter rigid steel loading plate for about to 20 to 35 miliseconds almost the same load duration of a vehicle moving at 40 to 50 mph (see Figure 2.1 below (Ullidiz & Stubsad, 1985)). Typically, three drops of 6000 lb (27 kn), 9000lb (40 kn), and 12000lb (53 kn) were applied in the same location on an asphalt pavement surface to produce a peak dynamic force of about 1500 lb (6.67 kn) to lb (120.0 kn) in milliseconds, (Crovetti, J A Shahin & Touma, 2000). Load (lb) Time (milliseconds) Figure 2.1: Haversine Loading Applied by FWD in Defiance, Section C146-Krouse Road Deflections induced by the FWD equipment are collected at the center of the dropped weight and up to six other locations (a series of sensors each: -d1, d0, d1, d2, d3, d4, and d5; located along the centerline of the trailer). These deflection sensors are located in

28 various radial distances from the applied load as shown in Table 2.1. (FHWA, 2009 & Dynatest 1995). 28 Table 2.1: Sensor Spacing of the FWD Device (FHWA, 2009 & Dynatest, 1995) Sensor -D 1 D 0 D 1 D 2 D 3 D 4 D 5 Offset Load Center (inches) Based on the radial distances shown in Table 2.1, the deflection measurements are recorded by the data acquisition system typically located in the vehicle (Jordan, 2013). A typical test schematic of FWD device mounted in the trailer system together with deflection basin is indicated in Figure 2.2. The central sensor (d0), placed in the middle of plate measures maximum deflection during testing. At the same time, the first sensor (d1) offset 12 inches away from central sensor and the rmaining series of sensors, measure deflections at different points. Figure 2.2: Falling Weight Deflectometer Schematic (Ferne & Langdale, 2010)

29 29 Deflection data collected by a series of sensors indicated in Figure 2.2 are then processed to estimate the pavement stiffness in terms of layer resilient modulus. This layer modulus obtained from known FWD data is termed backcalculated modulus. A number of commercial and non-commercial software are available for the analysis of FWD data to determine this backcalculated layer modulus. The backcalculated modulus is not only used in design but also to determine the layer coefficient and/or remaining life of the pavement structures. Therefore, the role of this layer modulus is very important in pavement engineering. This study focuses on the evaluation of the backcalculated layer modulus using MODULUS 6.0 software and AASHTO 1993 guide for designing pavement structures. Moreover, FWD testing have several advantages. It can directly estimate the Modulus of Subgrade Reaction (MR), it can precisely simulate traffic loading, it is easy and can be operated by a single person, and it is quicke (can test up to 60 points per hour). Also, the dropping loads vary from 1,500 to 27,000 lb (6.67 to 120 KN (Dynatest, 2009)). Based upon available FWD device in the Ohio Department of Transportation (ODOT) and among several FWD systems described in the literature review, the Dynatest Model 8000 (a single-axle trailer-mounted FWD) was selected as the most applicable device for the evaluation of local pavements condition during this research Dynatest Model 8000 FWD The Dynatest FWD is a lightweight trailer-mounted device which has enjoyed the long service record in the United States (Crovetti, J A Shahin & Touma, 2000). Figure 2.3

30 30 in below shows a view of this equipment. The Dynatest FWD consists of three main components as describes below (Nazzal, 2003). 1. A Dynatest 8002E FWD Trailer. 2. A Dynatest System Processor. 3. A Hewlett-Packard HP-85B Laptop computer (Current system uses a windows based laptop). This device is equipped with a load cell to measure the applied force and seven to nine geophones (velocity transducers) to measure the deflections up to 2mm. The Dynatest FWD is further equipped with a standard or inches (300 or 450 mm) diameter rigid or segmented loading plates, a rubberized pad, and a buffer system to help distribute the load evenly (Dynatest 1995). The load is normally dropped from predetermined heights ranging 2 to 20 inches (50 to 510 mm), (Nazzal, 2003). The load cell and seismic deflection geophones (transducers) are both linked to sockets in a protective Trailer Connection Box on the trailer. The transducers and the trailer connection box are connected to a system processor (Dynatest, 1995).

31 31 Figure 2.3: Dynatest Model 8000 FWD (LRTC 2000 & Nazzal, 2003) Figure 2.3 illustrates a FWD type developed by the Dynatest which is the original commercial developer of the FWD technology, and is the world s larger supplier of FWD Equipment. The Dynatest FWD s dynamic load capacity goes up to 54,000 lb (240.2 KN), (Ahmed, 2010). A microprocessor based control and signal processing unit (the Dynatest system processor), links the FWD trailer with the computer system. Also, this system controls the FWD process, achieves scanning, modifying and further processing of the geophone signals and monitors the status of the FWD unit to assure precise measurements. The application of the loading is remotely controlled by the operator (Nazzal, 2003). In addition, many other manufacturers of impulse devices for the nondestructive testing of pavement structures are available. A brief list of those manufacturers were KUAB America, Carl Bro Group, and Foundation Mechanics Incorporated, who offers FWD equipment through its JILS sections (Ahmed, 2010).

32 KUAB America KUAB FWD manufactures a wide variety of FWDs which are capable of delivering dynamic loads up to 66 kips ( KN) and currently operates five FWD s types. The load is applied through a single or dual mass system, and the dynamic response of the pavement system is measured in term of vertical deformation, or deflection, over a seismometers area combined with LVDT s through a mass-spring reference system. A specific load plate is incorporated to produce a uniform pressure on the pavement surface (Ahmed, 2010) Carl Bro FWD Carl Bro is another producer of FWD devices. Dynamic load capacity of this type of FWD is about 56,200 lb (250 KN), (Alavi et al., 2008). A series of 9 to 12 velocity transducers are used to evaluate the load and dynamic response. A single mass is used and controlled hydraulically which reacts as rubber buffer system to supports the dropped weights JILS FWD Foundation Mechanics, based in California manufacture under its nameplate JILS FWD s that have seven to nine deflection sensors (velocity transducers) with a single integrated response to determine the deflection. This type of FWD generates a minimum load of 1,500 pounds (6.67 KN) to a maximum load capacity of 54,000 pounds (240.2 KN). Unlike the Dynatest FWD, the JILS FWD utilizes two adjustable air bags for controlling load direction, magnitude. The rise time is dependent on the mass, dropping height and arresting spring properties (Ahmed, 2010).

33 The Light Weight Deflectometer (LWD) A portable device, developed for in-situ testing by the Federal Highway Research Institute, the Light Weight Deflectometer (LWD) first appeared in 1981 at Magdeburg, Germany, (Amer, Elbaz, & Elhakim, 2014). The light weight deflectometer was invented to estimate the in-situ layer modulus of soils. This portable hand device can be used for structural evaluation of pavement layer systems. Resilient modulus, analogous to elastic modulus is the main parameter for characterizing base, subbase, and subgrade materials for pavement design in the United States, (Senseney, 2010). Additionally, the LWD consists of a circular plate ( typically varies in diameter 6, 8, and 12 inches (150, 200, and 300 mm)) resting on the ground to support an impulse load from a released weight, guide rode, sliding drop weight, a locking release mechanism, housing, geophone sensors, and urethane dampers. For safe operation, the sliding mass is supported with a transportation lock pin. During LWD testing a drop weight slides down from variable height (typically 33.5 inches (850 mm)) and applies a dynamic force impulse to the circular steel load plate, (Senseney, & Mooney, 2010). Three geophones, located at center underneath the plate and different offsets from loading point measure deflections. The one mounted in the center of the load plate measures a maximum deflections (d0) and two extra mounted on a support bar resting on the surface, measure deflection at two additional fixed locations. Force transducer mounted inside the housing measures the applied force (P) from the standard 22 lb (10 kg) or the optional 33 lb (15 kg) or 44 lb (20 kg) drop weight setups. In addition, the LWD transfers an average contact stress of 14 to 29 psi (100 to 200 Kpa) on the pavement surface, (a load pulse of 15 to 20 ms duration), (Tayabji, & E. Lukanen, 2000). According

34 to Senseney & Mooney (2010), the conventional LWD modulus (ELWD) is calculated in Equation 2.1 as follows: 34 E Equation (2.1) Where: ELWD = conventional modulus ν = Poisson s ratio of soil a = plate radius A = contact stress distribution parameter (A = 2 for a uniform stress distribution, A= π/2 for an inverse parabolic distribution, A = 8/3 for a parabolic distribution). Moreover, there are three main types of LWD, which have been used in previous research; the German Dynamic Plate (GDP), the Transport Research Laboratory (prototype) Foundation Tester (TFT), and the Prima 100 LFWD, (Nazzal, 2003). Table 2.2 provides a brief summary of the characteristics provided by five different LWD manufacturers. Each device is unique in terms of its dropping weight and height, impulse time, plate diameter and style, contact pressure, and sensors types, (Mooney & Miller, 2009).

35 35 Table 2.2: Physical Characteristics of Typical LWD Devices (Mooney & Miller, 2009) Manufacturer CSM Zorn Prima Loadman TFT Plate style Solid Solid Annulus Solid Annulus Plate diameter (mm) 200, , 200, , 200, , 200, , 150, 200,300 Plate mass (kg) 6.8, Variable Drop mass (kg) , 15, , 15, 20 Drop height (m) Variable 0.72 Variable 0.80 Variable Damper Urethane Steel spring Rubber Rubber Rubber Force measured Yes No Yes Yes Yes Plate response sensor Geophone Acceleromet er Geophone Accelerometer Geophone Impulse time (ms) ± Max load (KN) 8.8 a 7.07 a 1-15 a 20 a 1-15 a Contact stress User def. Uniform User def. Rigid User def. Poisson's ratio User def User def User def. (a) Dependent Upon Drop Height and Damper Table 2.2 demonstrates that although there are differences in the design and mode of operation which can cause variations in the field measurement output, there are many similarities in their mechanics of operation Prima 100 LWD The first LWD model used in this thesis was the Prima with its plate manufactured by Keros Technology and Carl Bro. both of Denmark, (Steinert et al., 2005). The Prima 100 made by Carl Bro. weighs, in total, approximately 57.2 lb (26 kg) and has varying falling mass between 22, 33, and 44 lb (10, 15, and 20 kg) along with a varying drop height 0.4 to 33.5 inches (10 to 850 mm). This device has a load impulse of between milliseconds and load range capacity of 225 to 3372 lb (1 to 15 KN) with its 11.8 inches (300mm) bearing plate diameter, (Fleming, et al., 2000). Also the Prima 100 allows

36 36 collection of up to two deflections at a specified radial distance of 12 to 24 inches (300 to 600mm) from the center geophone. It measures both the impact force (P) from the falling weight, and deflections as determined by integration from the velocity of the surface (Christensen, 2003). The Prima 100 is shown in Figure 2.4. Figure 2.4: Schematic of Prima 100 with Additional Geophones, (Senseney, 2010) Furthermore, a personal digital assistant (PDA) device connected to the LWD apparatus via wireless Bluetooth connection collects and saves measured load and deflections. The collected deflections create a deflection basin profile and combined surface modulus immediately after each reading The LWD Principle of Operation The Light Weight Deflectometer is a portable device for repeated testing which can be operated by a single person. It is a fast and less expensive test method. The relatively

37 37 small weight of LWD compared to FWD makes it more applicable for testing unbound pavement layers. The lower contact stress allows the apparatus to sometimes bounce and move immediately after impact of the weight (Von Quintus & Minchin, 2009). During operation, it requires a flat surface to function properly and three seating drops are performed to ensure close contact. Then another three drops were performed, and the deflection corresponding to each blow and the soil s dynamic modulus were calculated by the data acquisition system. An important insight into the soil property can be obtained by a typical output from acquisition system of LWD, which show time history data (see Figure 2.5 in below) Figure 2.5: Typical Time History Data from LWD Test (Mooney & Miller, 2009) The LWD is, however, not ideal for thicker pavements because of low contact stress and a limited depth of influence to the pavement layers. Also, it does not collect pavement

38 38 temperature in both thin and thick asphalt pavements; thus a further means of recording temperature is needed (Icenogle & Kabir, 2013). 2.4 Existing Correlations between FWD and LWD Numerical studies have been explored in the past to assess the FWD and LWD measurements and to evaluate the effect of some relevant parameters. However, little researches have been given to fully understand the correlation of LWD with different instrument configurations such as FWD. Only two published studies by Horak et al. (2008) and Mooney et al. (2015) addressed the relationship between the FWD and LWD with additional geophones/sensors (Mooney et al., 2015). The findings by Horak et al. (2008) on 3 to 4 inches (75 to 100mm) thick layers of sand treated with emulsion between FWD and LWD sensor deflections with various radial offset distances. His regression model at the center of loading plate yielded a nonlinear model (see Equation 2.2) with a low correlation (R 2 = 0.62). D d. Equation 2.2 However, high relationships (R 2 = 0.82 and R 2 = 0.67) were found between FWD and LWD sensor deflections at radial offset distance of r = 300 and r = 600 respectively. His regression models are describe in Equations 2.3 and 2.4, respectively. D d. Equation 2.3

39 39 D d. Equation 2.4 The results by Mooney et al. (2015) on full depth reclamation of asphalt layers using additives (Badlands, Carlsbad, and Mesa Verde), at the center of the loading plate, r = 300 mm, and r = 600 mm with R 2 = 0.71, R2 = 0.96, and R2 = 0.98 respectively were found to be: w 0.23d 0.26 Equation 2.5 w 0.22d 0.05 Equation 2.6 w 0.28d 0.01 Equation 2.7 Similarly, Fleming (2000) conducted a correlation study between three main types of LWD moduli with that of FWD, and the results of those tests proved that the evaluated moduli of the Prima 100 LWD was well correlated with resilient modulus of FWD. Equation 2.8 shows an example of well conducted results. MFWD = ELWD Equation (2.8) The next study was accomplished by Nazzal (2003), see Figure 2.6. His regression analysis for FWD and LWD results have proved that the best model to predict the FWD

40 40 backcalculated moduli, MFWD, in (MPa) from the LWD modulus, ELWD, in (MPa) is briefly described in Equation 2.9 in below with R 2 = 0.94, significance level < 99.9%, and standard error = 33.1: MFWD = 0.97 (ELWD) for 12.5 MPa < ELWD < 865 MPa Equation (2.9) Figure 2.6: Best Fit Model of Fleming (2000) & Nazzal (2003) As clearly indicated in Figure 2.6, Nazzal (2003) demonstrated a good correlation between FWD and LWD, which generally agreed with those of Fleming (2000). Moreover, FWD deflection normally correlate well with LWD deflections, but the back calculations shows variation (Saadeh & Rhagavendra; Zhang; Mohammad 2007). The correlation between LWD and FWD is known to vary with thickness. (Fleming and Lambert 2007).

41 41 Smaller contact stress, fewer geophones and shallow depth of influence of the LWD could be the reason of variations as well (Nazzal 2007). As back calculation procedure has an indispensable role in LWD modulus measurement, such a bad evaluation of inputs results in erroneous layer moduli. In addition, supporting layers can influence the surface layer (Von Quintus & Minchin 2009). Icenogle (2013) found that the FWD and LWD deflections correlated well. However, the backcalculated moduli of the surface layer between these two tests do not correlate. This is because of the variations of the back-calculation software and the number of geophones representing the deflection basin. Conventional FWD moduli were found to be 2.5 to 3.3 times larger than LWD moduli (Livneh & Goldberg, 2001). Variations of loading level/rate used in FWD and LWD is the author s reason. Furthermore, the LWD moduli depends on location, soil type, pavement thickness, gradation, and moisture content. The stiffness moduli ratio between the FWD and the LWD varied between 0.8 to1.21 with R 2 = 0.5 to 0.9 (Fleming et al, 2007). According to Rahimzadeh (2004), the correlation between FWD and LWD was found to be material thickness and type dependent. Table 2.3 shows a short summary of aforesaid correlation equations obtained to relate FWD with LWD moduli in various researches.

42 Table 2.3: Regression Analysis Between FWD & LWD Moduli, (Shafiee, et al., 2013) 42 Equation Layer Description R-Square (R 2 ) Value LWD Model Source LWD(MPa) = 0.97FWD(MPa) 450-mm granular capping over silt and caly 0.60 Prima 100 LWD(MPa) = 1.21FWD(MPa) 260-mm limecement treated caly subgrade 0.77 Prima 100 (Fleming et al., 2000) LWD(MPa) = 0.80 to 1.30FWD(MPa) 225-mm wellgraded crush stone granular subgrade 0.50 Prima 100 LWD(MPa) = 1.03FWD(MPa) Granular subgrade 0.97 Prima 100 (Nazzal et al., 2004) LWD(MPa) = 1.33FWD(MPa) LWD(MPa) = 0.75FWD(MPa) Thin asphalt layer ( 127mm) Thicker asphalt layer ( 178mm) 0.87 Prima Prima 100 (Steirent et al., 2006) As indicated in Table 2.3, the coefficient of determination, R 2, value is smaller for thick and soft materials. This demonstrates that the Prima 100 LWD is an applicable device for thin layer consisting of stiff materials (Shafiee et al., 2.13). 2.5 Determination of Pavement Responses Using Deflection Basin Parameter According to Garg et al. (1998) and Kim et al. (2000), several pavement responses were identified by the researchers as good performance indicators during the structural evaluation of pavements. These included: (1) horizontal strain (tensile strain) at the bottom of asphalt layer; (2) vertical compressive strain on the top of the base layer; and (3) vertical compressive strain on the top of the subgrade.

43 43 In addition, many researchers have explored the relationships between deflection basin parameters and pavement responses such as stresses and strains using FWD test (Kim & Park, 2002). Also, Thompson (1989, 1995) proposed a relationship for full depth pavements and aggregate base pavements using the Area under Pavement Profile (AUPP). The AUPP is a FWD deflection basin shaped parameter. This dimensionless deflection basin parameter definition is complimentary to the AREA parameter. Also it has been widely used as a measure of pavement stiffness which means higher AUPP corresponds to lower stiffness and vice versa (Gopalakrishnan & Kim, 2010; Rada et al., 2015; Tang et al., 2012). The AUPP is described by Thompson (1989, 1995) in Equation 2.10 and Figure 2.7 as follows: AUPP 5D 2D 2D D Equation (2.10) Where: D0 = FWD sensor deflection at the center of the loading plate, mils D1 = FWD sensor deflection 12 inches from the center of the loading plate, mils D2 = FWD sensor deflection 24 inches from the center of the loading plate, mils, D3 = FWD sensor deflection 36 inches from the center of the loading plate, mils

44 44 Figure 2.7: Area Under Pavement Profile (Adopted from Thompson, 1989) The tensile strain at the bottom of the asphalt layer (εac), for full-depth asphalt is computed from Equation Log ε Log AUPP Equation (2.11) For aggregate base pavements, the tensile strain can be predicted using Equation 2.12 as follows: Log ε Log AUPP Equation (2.12

45 45 It is worthy to mention that the geometric property of the deflection basin (AUPP) is a significant parameter, which can be used to predict the horizontal strain (tensile strain) at the bottom of the asphalt layer. The use of AUPP for predicting (εac) is not affected by the type of pavement and subgrade (Kim & Park, 2002). Specifically, this study further investigated Thompson s (1989, 1995) promising relationship (AUPP) for the determination of the horizontal strain (tensile strain) at the bottom of an asphalt layer using FWD and LWD sensor deflections at radial offset distance 0, 12, and 24, 36 inches and 0, 12, 24 inches from the center of the loading plate respectively. 2.6 Backcalculation of Layer Moduli Backcalculation is an analytical technique by which pavement layer moduli and other stiffness properties are calculated, corresponding to the measured load and deflections. The analysis may be conducted by the following methods: iteration, closed form solution, database-searching, and simultaneous equations (using nonlinear regression equations produced from layered elastic analysis output data), (ASTM D , 2003; Alavi et al., 2008). Backcalculations using iteration method for calculating pavement layer moduli and subgrade resilient modulus is the most widely accepted method based, on pavement deflection profile or basins generated by FWD and LWD (Muench, et al., 2003; Rahim & Geprge, 2003; Romanoschi & Metcalf, 1999). This method requires the initial inputs such as assumed layer moduli that is often called (seed) modulus for the pavement, number of layers, layer thicknesses, and Poisson s ratio (ASTM D5858, 2003). This value should be selected carefully for the subgrade layer. A typical range of Poisson s ratio values

46 based on ASTM D5858 (2003) which may be used if other values are not available, are describe in Table Table 2.4: Typical Poisson s Ratio Values, (ASTM D5858, 2003) Asphalt concrete 0.30 to 0.40 Portland cement concrete 0.10 to 0.20 Unbound granular bases 0.20 to 0.40* Cohesive soil 0.25 to 0.45* Cement-stabilized soil 0.10 to 0.30 Lime-stabilized soil 0.10 to 0.30 * Depending on Stress/Strain Level and Degree of Saturation. After assuming the initial layer moduli, the surface deflections at radial offsets (geophone location) can be computed by the mechanistic analysis based on seed modulus and layer geometry. The computed deflections are then compared to the field measured deflection values. The process is continued by changing or adjusting the layer moduli each time, until a good match (within some tolerable error) between the computed and theoretical (measured) deflections can be reached (FHWA, 1994). A basic schematic of the backcalculation technique is shown in Figure 2.8 as follows:

47 47 Figure 2.8: Backcalculation Flowchart (Lytton, 1989) The flowchart indicated above shows backcalculation technique. This flowchart is further explained briefly according to Qin (2010) below: 1. Layer thicknesses and loads: The first and second boxes in the left hand side, represent the layer thicknesses and applied load levels on the pavements surface respectively. These values are the inputs and should be known in advance. 2. Measured deflections: FWD field measured sensor deflections. 3. Seed moduli: Input of the initial modulus in order to calculate theoretical sensor deflections. 4. Deflection calculation: Use pavement response models such as stresses and strains which can be used to compute theoretical sensor deflections. 5. Error check: Correlate between computed and measured deflections.

48 48 6. Search for new moduli: Iteratively search until the computed and measured deflection are paired within tolerable error limit in order to find a new moduli of the pavement layers. 7. Controls on the range of moduli: A range of modulus which can be define for each pavement layer by backcalculation technique to avoid inconsistent pavement layer moduli Overview of Backcalculation Software Several well-known software for the evaluation of flexible pavements layer moduli are available. After a literature review MODULUS 6.0, and EVERCALC 5.0 were selected for this study. Since both are common and capable of producing reliable outputs. These programs are based on linear layered theory for the basic structural model of the pavement response. Table 2.5 shows a list of backcalculation software, (Appea, Brandon, & Jr, 2003).

49 49 Table 2.5: Existing Backcalculation Software (Adapted from Appea et al., 2003) Software EVERCAL C 5.0 BOUSDEF MODCOMP 5.0 PEDD MECHBAC K UMPED ELMOD MODULUS 6.0 Paveme nt Type Flexible Flexible and Rigid Flexible and Rigid Flexible and Rigid Flexible Flexible and Rigid Flexible and Rigid Flexible and Rigid Analysi s Method Static Static Static Static Static Static Static Static *Fixed or User Defines Positions Moduli Calculation Method Bowl Matching Bowl Matching Bowl Matching Determinin g Equations and Bowl Matching Bowl Matching Determinin g Equations and Bowl Matching Bowl Matching Bowl Matching Convergenc e Criteria Root Mean Square Error Absolute Sum Root Mean Square Error Minimum Absolute Difference Root Mean Square Error Minimum Absolute Difference Root Mean Square Error Root Mean Square Error Forward Analysis Method and Program Multilayered Linear Elastic, WESLEA Multilayered Linear Elastic, Boussinesq theory Multilayered Linear/Nonline ar Elastic, CHEVLAY2 Multilayered Linear Elastic, ELSYM5 Multilayered Linear Elastic, CHEVRON Multilayered Linear Elastic, CHEVRON Odemark- Boussinesq Method Multilayered Linear Elastic, WESLEA Stress and Strains* User defines position Does not Calculate Forward Calculatio n and User defines positions User Defines Position Does not Calculate User Defines Position Fixed No

50 50 From the ranges shown in Table 2.5, MODULUS 6.0 was selected based on the reliability of its results to analyze the FWD data, and EVERCALC 5.0 was selected due to its capability of sensors adjustments with LWD geophones to analyze LWD deflection data, (Al-Jhayyish, 2014 & Tawfiq, 2003). These two software s were used to estimate the pavement layer moduli. A comparison of their backcalculated layer moduli were conducted by the author in this thesis MODULUS Program Modulus developed by the Texas Transportation Institution is the most commonly used software for backcalculation pavement layer moduli, (Scullion et al., 1990; Uzan et al., 1989). It can be applied to a two, three, and four-layer system, and is based on the linear elastic theory. WESLEA, a layered elastic solution platform developed by US Army Corps of Engineers covered in Modulus as a subroutine to perform the forward calculation for building a database of calculated deflection basin (Tutumluer, Investigator, Pekcan, & Ghaboussi, 2009). This database is matched with measured deflections using subroutine to obtain the layer moduli in the pavement systems after several iterations. The latest version of this program is Modulus 6.0. This newest version can be run for FWD data including seven sensors easily, and it is able to analyze up to four unknown layer systems EVERCALC Program Evercalc, developed by the Washington State Department of Transportation is also a popular backcalculation program. It uses a WESLEA layered analysis program for forward calculation and a modified Augmented Gauss-Newton algorithm for optimization (Tutumluer et al., 2009). The optimization routine is applied to obtain a set of modulus

51 51 values which provide the best fit between measured and calculated deflections models, basins, when given an initial estimate of elastic modulus and a limiting range of moduli. (Tawfiq, 2003). Also as implied, a set of E values is assumed and the deflection at each sensor is calculated and matched within a pre-specified root mean square (RMS) error range. Each unknown E is varied independently, and a new set of deflections calculated for each variation. For every layer and every sensor, the intercept Aji, and the slope Sji (shown in Figure 2.9) are determined. For numerous deflections and layers, the solution is achieved by developing a set of equations which define the slope and intercept for every deflection and every unknown modulus (Tawfiq, 2003). Log (deflectionj) = Aji + Sji (log Ei) Equation (2.13) Figure 2.9: Relationship Between Deflection and Modulus (Tawfiq, 2003) Evercalc can evaluate up to five layers, ten sensors, and twelve drops per station. After estimating elastic moduli of pavement layers, it can determine the stresses and strains

52 52 at different locations. Also, it runs an inverse solution technique on FWD deflection data to determine a set of layer moduli, (Evercalc User s Guide, 2005). The deflection tolerance, moduli tolerance, and the highest number of iterations can be defined before running the program by the user. When one or more of these conditions are satisfied, the program will terminate. 1. Deflection Tolerance: RMS % 100 Equation (2.14) Where: RMS = root mean square error, dci = calculated pavement surface deflection at sensor i, dmi = measured pavement surface deflection at sensor i, nd = number of deflection sensors used in the backcalculation process. 2. Moduli Tolerance: Expressed by the following equation, (EVERCALC User s Guide, 2005) ε Equation (2.15) Where: Eki and E (k+1) = the i-th layer moduli at the k-th and (k+1)-th iteration, m = number of layers with unknown moduli.

53 53 CHAPTER 3 EVALUATION OF PAVEMENT CONDITION USING FWD AND LWD MEASUREMENTS 3.1 Field Testing This chapter discusses measured deflection data from the FWD and LWD devices. The data consists of (1) FWD filed measured data, (2) LWD field measured data, (3) layer thicknesses of the pavement, and (4) evaluation of materials properties. The Ohio Research Institute for Transportation and the Environment (ORITE) arranged many trips to conduct field tests for collecting deflection data, layer thicknesses, and materials properties around the states of Ohio. The map of counties which were investigated during this thesis work is shown in Figure 3.1. Figure 3.1: Ohio Counties Map (Adapted from ORIL, 2015)

54 54 A total of 68 projects with 99 test sites, grouped into five clusters, in eight different counties were provided for this study. These project sites were first grouped into: 32 test sites in Defiance County (cluster 1), 16 test sites in Harrison and Carroll Counties (cluster 2), 13 test sites in Auglaize and Mercer Counties (cluster 3), 14 test sites in Champaign and Madison Counties (cluster 4), and the remaining 24 test sites were included in Muskingum County known as a cluster 5 (Sargand et al., 2016). Furthermore, the FWD tests were immediately followed by the LWD tests across all test sites. The testing was conducted with thin asphalt layers 3-5 inches ( mm) underlying by cement treated base layer. Full depth reclamation mechanism was used to stabilize 4-6 inches ( mm) base layers. At each test site a minimum of three locations/sections were investigated. A total of three main mass drops (additional one initial LWD seating drop) were performed for both devices at each test location. Also for the FWD device, variable drop mass heights were used to achieve a target load. The target loads for one, two, and three drops were 6000lb (26.68 KN), 9000 (40 KN), and 12000lb (53.37 KN), respectively. However, the same target load lb ( KN) was used for all three main drops and one initial drop during LWD testing as recommended by Mooney et al. (2015). Consequently, three measurements in the same location were taken in order to ensure proper reading and to enhance the accuracy of correlations between FWD and LWD sensor deflections. The measurements of the three test drops of both devices were average and considered herein. As results, the test sites and grouped clusters are presented in Table 3.1.

55 55 Table 3.1: Ohio County Roads by Cluster and Construction Material Used County Name Cluster # Total Test Sites = (99) Defiance 1 32 Harrison Carroll Auglaize Mercer Champaign Madison Muskingum 5 24 Material Type Fiber Cement Cement FDR* Asphalt FDR HMA** Whitetopping Fabric Reinforced Stone Full Depth Grinding Cement FDR Permazine FDR HMA Asphalt FDR Aggregate Overlay HMA Cement FDR HMA Full Depth Grinding Partial Grinding 70/30 asphalt/cement HMA HMA Mechanical FDR Cement FDR HMA Cement FDR Geogrid HMA Motorpave Concrete Steel PCC*** Surge & 411 Brick & 411 Asphalt FDR Lime FDR Brick Fly Ash FDR Tests Performed FWD a, LWD b, PSPA c, & DCP d FWD a, LWD b, PSPA c, & DCP d FWD a, LWD b, PSPA c, & DCP d FWD a, LWD b, PSPA c, & DCP d FWD a, LWD b, PSPA c, & DCP d FDR * - Full Depth Reclamation; HMA ** Hot Mix Asphalt; PCC *** Portland Cement Concrete FWD a Falling Weight Deflectometer; LWD b Lightweight Deflectometer PSPA c Portable Seismic Pavement Analyzer, & DCP d Dynamic Cone Penetration

56 56 To better illustrate a listed materials and performed tests, Some terms in the table above, used in this study refer to the materials types and different procedures in widening/construction of rural roads are listed below for the convenience of the reader (Sargand et al., 2016). 1. Whitetopping: A reclamation approach in which the existing asphalt pavement is overlaid with Portland Cement Concrete (PCC). 2. Surge: Stone which is the product of the primary crushing run. This stone is used as base material for haul roads (a coarse, temporary road built to facilitate the movement of materials and equipment) to protect very soft and wet soils : Also referred to as stabilized crushed aggregate (ODOT item 411, material specification), includes coarse aggregate with a large amount of limestone fines. This aggregate blend is used as an aggregate base and will harden after addition of water and compaction due to the chemical cementation of the large stone combined with line fines. 4. Surge/411: Stabilized crushed aggregate (411) mixed with surge stone that is wetted and compacted. 5. Full Depth Reclamation (FDR): A reconstruction mechanism that pulverizes an existing flexible pavement with the underlying materials to a predetermined depth. Stabilizing agents such as cement, fly ash, lime or Permazine can be added to the pulverized blend. After, this blend can be compacted with the underneath materials in order to create a homogeneous layer as a base for a new pavement structure.

57 /30 asphalt/cement: A mixture of 70% recycled asphalt grindings from milling projects with 30% cement and water. This mixture is used in pavement structure as base material. 7. Permazine: Is an enzyme rich material used for soil stabilization purpose. This material is created by a natural fermentation mechanism and can be mixed with soil and water to produce a cementitious effect that builds a solid base structure. 8. Permazine FDR: Adding Permazine with Full Depth Reclamation (FDR) in order to stabilize the subgrade blend. 9. Motorpave: Usually referred to item 405 Bituminous Cold Mix Pavement. This material is frequently placed in 2 inches lifts and covered with chip seal. 10. Mechanical FDR: A well compacted full depth reclamation without using of any stabilization technique. 11. Lime FDR: Adding lime and water with full depth reclamation to stabilize the subgrade blend. 12. Geogrid: A geosynthetic material that can be used to keep structural integrity in soil structure in order to resist tensile stress in soil. 13. Full/Partial Depth Grindings: Using asphalt products which have been pulverized from another bituminous surface project and recycled for reuse as surface layer or an aggregate base. 14. Fly Ash FDR: Adding of fly ash and water with Full depth reclamation to stabilize the subgrade of the asphalt pavement. 15. Fiber Cement: A concrete pavement reinforced with small fiber.

58 Fabric Reinforced Stone: Using fabric on top of natural subgrade with a compacted overlaid aggregate base layer on top of the fabric. This mechanism is used to enhance the tensile strength of the aggregate base material for protecting the natural subgrade. 17. Aggregate Overlay: Using stone as a surface layer in pavement structure. 18. Asphalt FDR: An asphalt binder added to Full Depth Reclamation (FDR) to stabilize the pulverized subgrade blend. 19. Asphalt Grindings FDR: Adding recycled asphalt grindings with full depth reclamation to the asphalt subgrade blend. 20. Brick/411: Recycled bricks mixed with a 411 material and creates a blend that can be wetted and compacted to bind materials together. This type of material is used as an aggregate base. 21. Cement FDR: Adding cement and water to the full depth reclamation to stabilize the subgrade blend. 22. Concrete/Steel: Recycled concrete and rebar from old buildings, bridges, and pavement structures placed on top of the subgrade. This material can be used as an aggregate base. 3.2 Quantifying Pavement Condition Using FWD Deflections Falling Weight Deflectometer (FWD) testing was used to evaluate the 99 different test sites. For all of these test sites, the nominal 11.8 inches (300mm) plates and three loading drops each, approximately 6000lb (26.7KN), 9000lb (40KN), and 12000lb (53.4KN) were used to measure pavement surface deflections. Every sensor has a subscript

59 59 indicates the distance in inches, from the center of applied load. A typical surface deflections under variable loading conditions for Champaign section of Pisgah road is shown in Figure 3.2. FWD Load Sensor Deflection, mils D4 D5 6 D D D D Sensor Distance, inches 6000 lbf 9000 lbf lbf Figure 3.2: Typical Pavement Surface Deflection Basins Based on Load Levels, Champaign County, and Section Pisgah Road (C236-3) Moreover, a program called FWD-AREA to measure pavement condition was first developed by the Washington State Department of Transportation (2005). The FWD- AREA program computes a deflection basin that has a trapezoidal area developed in the pavement system based on dynamic load, (Jordan, 2013). Previous studies show the modulus of subgrade reaction (MR) is directly correlated with the deflection sensor spaced from center load in about 24 inches (D24). (WSDOT, 2005). The Equation 3.1 below was proposed for estimating MR:

60 M psi Equation (3.1) Where: D24 = 24 inches from the center of the loading plate, mils. MR = Modulus of subgrade reaction, psi. A computer software, Modulus 6.0, was used to backcalculate pavement layer moduli (typically, Surface, Base, Subbase, and Subgrade in this study). Thickness of the layers, determined from prior Dynamic Cone Penetration (DCP) testing and the Poisson ratio selected based on ASTM D5858 (2003) were the main inputs of the Modulus 6.0 software. Meanwhile, coring is a way to physically see and accurately measure different bound layers, helps to determine the bond quality between pavement layers, and identify the subgrade materials in asphalt pavement structure. Figure 3.3 shows DCP operation and sample measurements which was performed in of the tested section. Figure 3.3: Coring and Obtaining Samples, form One of Tested Section

61 61 Figure 3.3 indicates coring procedure which played a substantial role in the pavement investigation during this study. This analysis involves layer s thicknesses and determination of the material properties. Evaluating pavement characteristics and other predictors of pavement service cannot be done or seen visually without site investigation. Therefore this study covered and investigated the aforementioned requirements for local pavement system of the selected counties in Ohio. A complete coring summary along with layer thicknesses is shown in Appendix A FWD Results The results in Figure 3.4 demonstrate a typical shape of the deflection bowl for structural analysis of the pavements. Basically, the upper deflection line define the first dropped load, 6000lb (26.68 KN) in relation to the second and third dropped loads each, 9000lb (40 KN) and 12000lb (53.37 KN), respectively. FWD Sensor Deflection (mils) Radial Distance (in) 6000 lb 9000 lb lb Figure 3.4: FWD Deflection Basins, Various Loads, Cluster # 3, Section of Southland Road (Aug-C3-15), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate

62 62 Based on Figure 3.4, typically a semi bowl with little curvature shows minimum deflection and the resulting stiff layer system in the structural analysis of the pavements system. However, the semi bowl with high curvature around these loads indicates maximum deflection and the resulting weak layer system (see Figure 3.5). The remaining typical deflection basins of every location in cluster-3 for various test sites were provided, and can be seen in Appendix B. FWD Sensor Deflection (mils) Radial Distance (in) 6000 lb 9000 lb lb Figure 3.5: FWD Deflection Basins, Various Loads, Meter Road (CAR-T269-2), Aggregate Overlay Surface Layer, Carroll County, 11.8-in. (300-mm) Plate 3.3 Quantifying Pavement Condition Using LWD Deflections The second nondestructive device examining structural pavements performance during this study was the Light Weight Deflectometer (LWD). The Prima 100 LWD, with additional radial geophones was used to develop a deflection basin for backcalculation of layer moduli. Three consecutive drops of sliding weight (a standard 10kg) and plate

63 diameter of 11.8 inches (300mm) were performed on leveled surface. Figure 3.6 shows process of collecting field deflections using Prima 100 LWD. 63 Figure 3.6: Conducting Tests on Pavement Surface Sections in Defiance As shown in Figure 3.6, two additional geophones/sensors were placed at radial offset distance 12, and 24 inches from the center of the loading steel plate. Three drops (in addition of one initial drop) in each location/section were performed and the deflection basin was taken by geophones. The deflections and loads were measured and stored by a personal digital assistant (PDA). The PDA device has a Bluetooth wireless connection to the Prima 100 LWD apparatus. As previously stated, Evercalc 5.0 software was used to backcalculate pavement layer moduli in this study. The LWD software integrates the velocity transducer signal to determine peak deflection value. According to Shafiee et al. (2013), usually under testing the maximum deflection does not occur at the same instant as the peak load especially for lower stiffness materials as shown in Figure 3.7.

64 Load Cell (psi) Time (millisecond) Load Cell (psi) D0 (mils) D1 (mils) D2 (mils) Figure 3.7: Example of a LWD Output from Field Testing, Auglaize County, Section of Minster Fort Recovery Road, (Aug-C30-16) LWD Results Prima 100 LWD testing with radial geophones followed FWD testing, and three repeat measurements were taken at the same locations/section along the 99 different test sites as the FWD test. The Prima 100 LWD with additional geophones allows the analysis of more layers, and was used in this study to see whether its measurements correlates well with FWD measurements. The correlation between their measurements are presented and discussed in the next chapter. The LWD deflection measurements for cluster-3 (Auglaize + Mercer) are shown in Table 3.2.

65 65 Table 3.2: Prima 100 LWD Sensor Deflection Measurements for Cluster # Recovery - C Blank Pike C Neptune Mendon C161C Harris C175B Dutton C230A Road Name East Shelby - C71 Fairground -FG Kossuth Loop- C216A Secti D 0 D 1 D 2 D 0 D 1 D 2 Road Name Section on Mils Mils Mils Mils Mils Mils Southland C Minster Fort It is significant to mention that the Prima 100 LWD and FWD repeatability are slightly weak in rough or soft surfaces rather than in stiff/hard surfaces. Measured deflections on rough or soft surfaces (typically aggregate overlay and full depth grinding in this study) from multiple sensors across all test sites resulted variations and/or somewhat

66 66 identified to be outliers in the deflection data. Figures were addressed in appendix B in this study. The prima 100 LWD sensor deflection basin for Southland Road is shown in Figure 3.8 as follows: 0 LWD Sensor Deflection, mils lb 2400 lb 2360 lb Radial Distance, (in) Figure 3.8: LWD Deflection Basins, Same Loads, Cluster # 3, Southland Road (Aug-C3-15), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate 3.4 Backcalculation Methodology and Pavement Layer Moduli Table 3.3 shows a basic backcalculation procedure. The output of this methodology are modulus of elasticity, effective structural number, layer coefficients, and subgrade resilient modulus of pavement layer system.

67 Table 3.3: Representation of Backcalculation Procedure (Murillo & Bejarano, 2013) Direct Calculation E, D, ν d, σ, ε Backcalculation E, σ, ε d, D, ν 67 Where: E = Elastic Modulus of materials. Ν = Poisson s Ratio D = Layer thickness D = Defection of the pavement structure Ε = Strain Σ = Stress on each layer of the pavement structure AASHTO Method (Section 5.4.5, FWD) The AASHTO (1993), Design Guide for Pavement Structures establishes a pavement analysis method based on FWD testing results. This method is mostly used to calculate the subgrade resilient module (MR) and effective structural number. The method works based on elastic layer theory using the deflection at a sufficiently large distance from the center load to calculate the MR value. MR is then used as an input parameter thereafter to calculate the effective structural number (pavement structural capacity) and layer coefficients (hereafter referred to as the AASHTO 5.4.5). The MR is calculated using Equation 3.2.

68 68 M. and r 0.7a Equation (3.2 Where: MR = subgrade resilient modulus, psi P = load magnitude, lb (9,000 lb recommended by AASHTO). dr = measured deflection at distance r from the center of the load, inches. r = radial offset (distance from plate center, inches). AASHTO proposes a radial spacing which exceeds 70% of the effective radius (ae) of the stress bulb at the subgrade/pavement interface. The effective radius can be estimated using Equation 3.3. a a D Equation (3.3) Where: ae = effective radius of stress bulb at subgrade/pavement interface, inches. a = FWD load plate radius, inches. D = total pavement depth above subgrade, inches. MR = subgrade resilient modulus, psi. Ep = effective modulus of all pavement layers above the subgrade, psi. AASHTO also presents Equation 3.4 for predicting the modulus of all layers above the subgrade called, the effective modulus of the pavement structure (Ep). This can be determined in terms of calculated MR and total thickness of layers above the subgrade.

69 69 d 1.5 pa Equation (3.4) Where: d0 = measured deflection at the center of load, adjusted to a temperature of 20 C (68 F), inches. p = pressure of load plate (P/πa 2 ), psi. D = total thickness of all layers above subgrade, inches. a = FWD load plate radius, inches. EP = effective modulus of all pavement layers above subgrade, psi. MR = back calculated subgrade resilient modulus, psi. The EP value can be simply determined from equation 3.4 in an excel spreadsheet by using an iterative procedure such as the built in Solver function. Also, this can be estimated by using the bisection method that will produce the measured center deflection (d0). After the Ep has been calculated, Equation 3.3 is used to verify that the reasonable radial distance parameter (r) criteria has been met. Moreover, the effective structural number which represents the structural strength of the overall pavement sustains traffic loadings. The effective structural number (SNeff) for the entire pavement system can be calculated based on the total thickness of the pavement system and its computed effective modulus using Equation 3.5.

70 70 SN D E Equation (3.5) Where: SNeff = effective structural number of in place pavement. D = total pavement depth above subgrade, inches. Ep = effective modulus of all pavement layers above the subgrade, psi. AASHTO procedure was applied based on FWD field-collected data for flexible pavement. The calculated effective structural number, subgrade modulus, total thickness of pavement, effective modulus, and central deflections for each site are shown in appendix C of this study Determining Layer Coefficients from AASHTO Equations There is no standard method of calculating layer coefficients (ai) for flexible pavement based on FWD data. In order to find layer coefficients, knowing effective structural number of pavement system is necessary. Therefore, the effective structural number for every site was determined as discussed in the preceding section. The proposed equation of structural number structural number is a combination of the thicknesses for each layer and layer coefficients of that specific layer as shown in below: SN a D a D a D Equation (3.6)

71 71 Where: a1, a2, a3, = are empirical layer coefficients for the pavements layers (surface, base, and subbase). D1, D2, D3 = thicknesses for the surface, base, and subbase of the pavement layer system. The structural layers coefficients were determined by solving simultaneously equations for structural number of all sites with similar material, while there was only one equation representing each site. Any possible combination of equations which was necessary in computing layer coefficients due to effective structural number was considered. The process consisted of combining equations form several sites in different groups based on similar materials. Solutions containing negative values and values larger than 1.0 were considered non feasible solution in this study. Materials characterized layer coefficient determined from solving simultaneous equation are shown in Table 3.1 below.

72 Table 3.4: Calculated Layer Coefficients Range Based on Material Types, AASHTO Materials Types Calculated Layer Coefficient Range Portland Cement Concrete (PCC) White topping HMA Mechanical FDR Fiber Cement Cement FDR asphalt/cement Concrete Steel Full Depth Grindings Partial Depth Grindings Lime FDR Permazine FDR Fabric Reinforced Stone Fly Ash FDR Brick & Surge & Motorpave Asphalt FDR Geogrid Aggregate Overly The layer coefficients ranged in Table 3.4 was summarized as box plots in Figure 3.9 to show the variability of the data.

73 73 Layer Coefficient Box Plot for Layer Coefficient (FWD AASHTO 5.4.5) Layer Type Bottom 2Q Box 3Q Box Mean Figure 3.9: Box Plot of Layer Coefficients for Each Widening/Construction Treatment, Layer Type Based on AASHTO As shown in Figure 3.9, the bottom of the box represents the first quartile (Q1) and the top represents the third quartile (Q3). The line within the box (a line across the box) represents the median value and lastly the bold dot within the box represents the mean of the response within that group (the mean value). The two lines extending from the box upward and downward, each represents values outside the first and third quartile. Furthermore the horizontal bars at the end of upper and lower extended vertical lines represent the maximum and minimum values respectively. To determine the spread and skew of the data, box plots are useful. The plots can be used to identify outliers for removal from the data analysis. According to Tukey (1977), the space between Q3-Q1 is known as

74 74 inter quartile range (IQR) and this measure is significant in detecting outliers in the data. Any observation falling outside Q3+1.5*IQR or Q1-1.5*IQR could be flagged as potential outlier. In Figure 3.7, the observation falling outside 0.34 or for Mechanical FDR, 0.38 or 0.05 for 70/30 asphalt/cement, 0.25 or 0.1 for Full depth grindings, and 0.83 or 0.05 for Motorpave could be flagged as potential outliers for characterized materials. Meanwhile, the box plot can be further used when comparing various materials. If the boxes do not overlap, the two layer coefficients are difference from each other. If the boxes overlap, but do not contain both medians, the layer coefficients are likely different from each other. When the boxes overlap and contain both medians, then both materials are considered to have the same layer coefficient values AASHTO Method (Section 2.3.5, LWD) The AASHTO (1993), Guide for Design of Pavement Structure describes a procedure for estimating the structural layer coefficients from laboratory data. In this study, AASHTO section hereafter AASHTO was used to determine layer coefficients for granular base and subbase layers using backcalculated modulus values. This section proposed the relationship via equation 3.7 for base materials such as gravel or crushed gravel from its elastic (resilient) modulus. a log E Equation 3.7

75 75 Where: EBS = base layer modulus. a2 = base layer coefficient. For the crushed stone as a subbase layers equation 3.8 was used as described below: a log E Equaiton 3.8 Where: ESB = subbase layer modulus. a3 = subbase layer coefficient. Also, AASHTO provides a graph shown in Figure 3.8, which can be used to determine the structural layer coefficient of asphalt concrete surface course from its elastic (resilient) modulus. Extrapolation procedure was used to determine structural layer coefficients when the surface moduli that exceeded the standard range of proposed chart by AASHTO. The graph is described as follows:

76 76 Figure 3.10: Chart for Estimating Structural Layer Coefficient of Asphalt Concrete (AASHTO, 1993) Moreover, the AASHTO provides the chart show in Figure 3.11 for estimating the structural layer coefficient of cement-treated base materials according to its elastic layer modulus. Extrapolation technique was used to determine structural layer coefficients from the cement-treated base modulus that exceeded the proposed standard range of the chart. The chart is showing below:

77 77 Figure 3.11: Used Chart for Cement-Treated Base Materials, (AASHTO, 1993). The material characterized layer coefficients range determined from the above procedure are shown in Table 3.5.

78 Table 3.5: Calculated Layer Coefficients Range Based on Material Types, AASHTO LWD 78 Materials Types Calculated Layer Coefficient Range Portland Cement Concrete (PCC) White topping HMA Mechanical FDR Fiber Cement Cement FDR asphalt/cement Concrete Steel Full Depth Grindings Partial Depth Grindings Lime FDR Permazine FDR Fabric Reinforced Stone Fly Ash FDR Brick & Surge & Motorpave Asphalt FDR Geogrid Aggregate Overly Also, the results presented in Table 3.5 are graphically presented in the box plot in Figure 3.12

79 79 Layer Coefficient Box Plot for Layer Coefficient (LWD AASHTO 2.3.5) Layer Type Bottom 2Q Box 3Q Box Mean Figure 3.12: Box Plot of Layer Coefficients for Each Widening/Construction Treatment, Layer Type Based on AASHTO As shown in Figure 3.12, few outliers were identified in the above box plot as indicated in Figure Generally, any observation falling outside 0.90 or 0.42 for PCC, 0.25 or 0.05 for Full depth grindings, 0.30 or 0.14 for Concrete Steel, 0.28 or 0.01 for Lime FDR, 0.33 or 0.20 for Brick & 411, and 0.38 or 0.08 for Motorpave could be flagged as potential outliers for characterized materials Rohde s [1994] Method of Determination of Pavement Structural Number and Subgrade Modulus from FWD Testing. Rohde (1994) developed a procedure for obtaining the structural number of a pavement system based on FWD measurements. The structural number equation adopted

80 80 the one modified by the Transport Research Laboratory (TRL) in 1957 which was used in the World Bank Highway Design and Maintenance pavement performance model in the United Kingdom (Janoo, 1994). Modified structural number (SNC) equation is described as: SNC a h SN Equation 3.9 Where: SNC = Modified Structural number, SNsg = 3.51(log CBR) , CBR = in situ California bearing ratio, ai = material and layer coefficient, and hi = layer thickness (inches) Rohde assumed that the surface deflection measured at an offset of 1.5 times the structural pavement thickness (h) is due to the subgrade only. After comparing this deflection with the maximum or peak deflection, he established the Structural Index of the Pavement (SIP). The SIP correlated with the deflection above the subgrade is defined in Equation 3.10 as follows: SIP D D. Equation 3.10

81 81 Where: SIP = Structural Index of pavement, D = Maximum or peak deflection measured under a standard 9000lb (40KN) FWD Load. D1.5Hp = 1.5 times Hp offset measured surface deflection under 9000lb (40KN) of FWD impulse load. Hp = total pavement thickness. Rohde hypothesized the SIP must be fully correlated with the stiffness of the pavement structure and thus the structural number. Rohde investigated and developed the best relationship between structural number and SIP based on regression analysis. A relationship of Equation 3.11 was selected: SN k SIP h Equation 3.11 Where: SN = Structural number, inches SIP = Structural index of pavement (µm), Hp = total pavement thickness (mm), k1, k2, k3 = Coefficient as listed in Table 3.6.

82 Table 3.6: Coefficient for Structural Number versus SIP Relationships, (ROHDE, 1994). 82 Surface Type k1 k2 k3 r 2 * n** Surface Seal Asphalt Concrete * Coefficient of determination **Sample Size Moreover, Rohde (1994) used field-measured FWD deflection data to obtain the subgrade modulus (Esg). He developed a second index called the structural index of subgrade (SIS). This index was defined as: SIS D. D Equation 3.12 Where: Ds = measured deflection spaced 30 inches from the center of the loading plate. The subgrade modulus can be describes as follows: E 10 SIS Hp Equation 3.13 Where: Esg = subgrade modulus, Mpa k4, k5, k6 = coefficients as given in Table 3.7.

83 83 Table 3.7: Coefficient for E versus SIS Relationship, (Rohde, 1994) Total Pavement Thickness k4 k5 k6 r 2 n Hp 380 mm mm < Hp 525 mm mm < Hp The Rohde method of determination of effective structural number, and subgrades modulus was applied to FWD field measured deflections to improve and confirm correlations between FWD and LWD. Results of this procedure are presented in Table 3.8. Table 3.8: Effective Structural Numbers and Subgrade Modulus from Rohde Procedure Road Name Structural Number Subgrade Modulus E sg (ksi) Road Name Structural Number Subgrade Modulus E sg (ksi) Christy-C164 10* 20 Southland-C Blosser-C Minster Fort Recovery-C Mountain Perry-C * 33 Dutton (C230A) 3 11 Arch Hill-C * 22 Neptune Mendon- (C161C) 4 25 Vista View Drive 9* 22 Harris (C175B) 4 20 Air Park 9* 25 East Shelby-C Airport-C * 21 Mansfield-C6-14C 2 10 Mansfield-C Blank PikeC Elliott Road-C Salt Creek Road- C Elliott-C Dietz Ln-C

84 84 Banner School-C70-09 Banner School-C70-11 Table 3.8: Continued 5 10 Apollo-C Canyon-C * 59 Blosser-C Chase-C * 54 Rosedale-C117-01A 2 9 Meter-T Rosedale-C117-01B 3 11 Plum Run-C Rosedale-C Birmingham-C WCC- C Unionvale-C * 54 The Bend-C Bakers Ridge-C Krouse-C * 22 Fountain-C Harding-C Flory- C Kite-1-C Blosser-C Heck Hill-C WCC-l-C Nine Miles-C Hammon-T Nine Miles-C Taylor Blair-C14-S4 2 4 Sullivan-C Taylor Blair-C14-N5 1 5 Lippincott-C MCEO 3 21 Dallas-C Old Troy Pike-C Old Troy Pike-C Charleston Chillicothe-C15B Davis-C Rural Dale-C Pisgah-C Ellis Dam-C Fairground (West) 1 12 Powelson-C Pledge-T Narrows-C Meter-T Kossuth Loop- C216A-03 Friendly Hill-C New Hope-C Fairground (Center) 1 9 Southern-C Fairground (East) 1 18 Norfield-C

85 85 In above table the structural numbers greater than 6 were concrete pavements and/or concrete steel. In addition, materials characterized structural layer coefficients range determined from Rohde method are shown in Table 3.9 Table 3.9: Calculated Layer Coefficients Range Based on Material Types, Rohde [1994] Method Materials Types Calculated Layer Coefficient Range Portland Cement Concrete (PCC) Whitetopping HMA Mechanical FDR Fiber Cement Cement FDR asphalt/cement Concrete Steel Full Depth Grindings Partial Depth Grindings Lime FDR Permazine FDR Fabric Reinforced Stone Fly Ash FDR Brick & Surge & Motorpave Asphalt FDR Geogrid Aggregate Overly To better see the variability of the data and due to the high volume of the collected and analyzed data, Table 3.9 values were graphically plotted. A figure of graphical box plots for the layer coefficients is shown in Figure 3.13.

86 86 Box Plot for Layer Coefficients, Rohde Procedure Layer Coefficient Bottom Layer Type 2Q Box Figure 3.13: Box Plot Showing Layer Coefficients for Each Widening/Construction Treatment as Determined Using Rohde [1994] Procedure In Figure 3.11, a few outliers were identified. Typically any observation falling outside describing ranges could be flagged as potential outlier. For PCC 1.12 or 0.52, White topping 0.80 or 0.40, Mechanical FDR 0.24 or 0.05, Full depth grindings 0.20 or 0.07, Partial depth grindings 0.25 or 0.04, Lime FDR 0.25 or 0.02, Fabric reinforced stone 0.28 or 0.05, Asphalt FDR 0.21 or 0.02, and lastly Geogrid 0.15 or Pavement Layer Moduli As previously discussed, Modulus 6.0 and Evercalc 5.0 were used to backcalculate layer moduli during this study. These programs are the most commonly used backcalculation programs that can evaluate pavement structural capacity up to five different unknown layers, (Tawfiq, 2003). Also, the Dynamic Cone Penetration (DCP) was

87 87 used to identify pavement material types and layer thicknesses. For each tested location/section, a separate layer thickness was assigned based on coring results and DCP testing. According to ASTM-D5858 (2003), a Poisson s ratio, 0.35, along with field temperature obtained from FWD testing were some initial software inputs. Evercalc 5.0 software works based on multi-layer elastic forward calculation subroutines. To begin the backcalculation process, a general file including but not limited to the following input parameters must be generated: Loading plate radius, number of layers, number of sensors, sensor spacing, and Poisson s ratio. Additional input parameters, options pertaining to the treatment of three drops at almost same load levels, and deflection basin for one of the test section, Pisgah road, Pisgah-C-236-3, Champaign County are shown in Figure 3.14 through Figure 3.16 respectively.

88 88 Figure 3.14: Evercalc 5.0 General File Data Entry Screen for Pisgah Road, Champaign County. Figure 3.15: Evercalc 5.0 LWD Deflection File screen for Pisgah Road, Champaign County.

89 89 Figure 3.16: Evercalc 5.0 LWD Deflection Basin for Pisgah Road, Champaign County Similarly, Modulus 6.0 works based on the linear elastic theory. WESLEA, a layered elastic solution platform developed by US Army Corps of Engineers covered in Modulus as a subroutine to perform the forward calculation for building a database of calculated deflection basin (Tutumluer et al., 2009). The general window of Modulus 6.0 is shown in Figure 3.17.

90 90 Figure 3.17: Main Window of Modulus 6.0 (Liu and Scullion, 2001) Figure 3.18: Backcalculation Routine Window, Krouse Road, Defiance County.

91 91 Moreover, a number of backcalculated layer moduli were computed based on material properties. A summary of backcalculated layer moduli from Modulus 6.0, FWD testing and from Evercalc 5.0, LWD testing, are reported in Appendix D. Outliers amongst the characterized material layer moduli for each section were also identified. A similar graphical representation of backcalculated layer moduli, computed from seven FWD, three LWD, sensor deflections with three varying applied loads, for three typical test sections in each location, with the 11.8-in (300mm) plates of both devices can be seen in the box plots presented in Figure 3.19 and Figure 3.20, respectively. Backcalculated Layer Moduli Box Plots, Modulus 6.0 (FWD) Backcalculated Layer Moduli (ksi) ,343 1, Bottom 2Q Box 3Q Box Mean Layer Type Figure 3.19: Box Plot Showing Backcalculated Layer Moduli for Each Widening Treatment as Determined Using Modulus 6.0 Software, FWD Testing.

92 92 In the box plots of Figure 3.19, any observation falling outside describing ranges could be flagged as potential outlier as described: For PCC 3718 or 1190, White topping 2617 or 780, Fiber Cement 1791 or 688, Cement FDR 1459 or 773, Concrete Steel 186 or 8, Lime FDR 1173 or 500, and Motorpave 1298 or 649. Similarly, backcalculated layer moduli box plots based on materials characterization from Evercalc 5.0, LWD testing with three sensor deflections measurements, are shown in Figure Backcalculatd Layer Moduli Box Plot, Evercalc 5.0 (LWD) Bottom 2Q Box 3Q Box Mean Backcalculated Layer Moduli (ksi) ,288 1, Layer Type Figure 3.20: Box Plot Showing Backcalculated Layer Moduli for Each Widening Treatment as Determined Using Evercalc 5.0 Software, LWD Testing.

93 93 Figure 3.20 indicates backcalculated layer moduli from LWD deflection data. Few outliers were identified. Any observation falling outside the ranges could be flagged as potential outlier as described: For PCC 3201 or 1535, White topping 3530 or 64, HMA 1604 or 6, Fiber Cement 1242 or 35, Cement FDR 790 or 6, 70/30 asphalt/cement 230 or 18, Lime FDR 1689 or 946, and lastly for Motorpave 262 or 17.

94 94 CHAPTER 4 RESULTS AND DISCUSSION 4.1 Introduction As described in Chapter 2 of this study, several studies have been done previously to determine the relationship between FWD and LWD. Various pavement structures and material types were examined in these studies. In this chapter the author conducted a comprehensive statistical analysis to correlate FWD and LWD sensor deflections, backcalculated layer moduli, and layer coefficients obtained from their measurements. The best correlation parameters are based on routine regression analysis using a statistical software, the Statistical Package for the Social Sciences (SPSS). Also FWD and LWD correlations was verified and proved by the Rohde method using, effective structural numbers, layer coefficients, and subgrade moduli. 4.2 Regression Analysis As mentioned previously, in order to determine the correlation between the FWD, LWD measurements and prove it by the Rohde methods, a statistical analysis using SPSS and Microsoft excel are used to perform an extensive regression analysis on data described in the previous chapter. The main objective of regression analysis is to obtain the parameter in the least squared error model that can predict the FWD layer coefficients, layer moduli, and effective structural number from the LWD measurements and the Rohde method with their corresponding coefficient of determination, R 2, standard error, and statistical significant level. Linear and nonlinear regression models were utilized in this study. According to Field (2013) a common form of linear regression model is describes as:

95 95 Y b b x ε Equation 4.1 Where: Yi = Dependent variable, b0 = intercept value, b1 = slope of the regression, and εi = residual term (difference between the prediction and actual). The measurements determined from LWD and the Rohde method were used as the independent variable in comparing with their dependent variable, FWD measurements, in all the regression models. However, the Rohde method is used as a dependent variable in the regression model obtained, while comparing it with its independent variable LWD. Moreover, the coefficient of determination, R 2, statistical significance level, and the standard error are considered to be reported for each regression model developed in this study. The coefficient of determination, R 2, is a number that represents the proportion of variation in the dependent variable which is predictable from the independent variable and has a value which ranges from 0 to 1. A perfect correlation exists when the value is equal to one, this means all points lie on the suggested least square line. The significance level is the result for a given null hypothesis test for which a typical P-value of less than or equal to 0.1, 0.05, and 0.01 is considered statistically significant. Lastly, the standard error is define as the standard deviation of the sampling distribution of a statistic or can be the square root of the mean square errors, MSE, (Nazzal, 2003).

96 Comparison FWD and LWD Sensor Deflections The author conducted a comprehensive regression analysis to find the best correlation between FWD and LWD sensor deflection data. To compare collected deflection data, all the deflection data from both the FWD and LWD were normalized to 9000 lb (LWD sensor deflections were extrapolated linearly to 9000 lb). Since the Prima 100 LWD has geophones/sensors up to 24 inches only (600mm) from the center of the loading plate, the deflection data of both devices were compared at 0, 12, and 24 inches about (0, 300, and 600mm) from the center of the loading plate. Each sensor has a subscript (0, 1, and 2.) which represents the deflections at 0, 12, and 24 inches (0, 300, and 600mm) respectively. The Statistical Package for the Social Sciences (SPSS) is used to perform the regression analysis between FWD and LWD sensor deflections individually. Also, FWD and LWD sensor deflection data were filtered to detect outliers using SPSS. Some abnormal deflections amongst normal deflections were identified. Accordingly, a decision was made to exclude/remove outliers due in part to plate vibration and/or soft surface layers from regression modeling prior to data analysis. A typical normalized FWD and LWD sensor deflections of all tested sites, at 0, 12, and 24 inches (D0, D1, and D2) from the center of loading plate and a brief list of the detected outliers are available in appendix E of this study Deflections at the Center of Loading plate, (D 0 ) Deflection data collected from the FWD and LWD measurements at the center of 11.8 inches (300mm) loading plate were compared. For all tested locations/sections, the FWD central deflection data at (0.0 distance) is plotted against LWD. The regression

97 97 analysis yielded a nonlinear model with a power function that gives the best fit based on the best regression coefficient of determination, R 2 = It is worth mentioning that other linear and nonlinear relationships (exponential, polynomial, and linear) that improve the fit model were also checked, but did not fit into the data point. To better illustrate the relationship, a separate plot of regression model was developed and presented in Figure 4.1below: Normalized FWD Sensor Deflections, (mils) FWD versus LWD Center Deflecions, (D 0 ) y = x R² = Extrapolated LWD Sensor Deflections, (mils) Figure 4.1: Comparison Between FWD and LWD Deflections at the Center of Loading Plate, (D0) As previously stated, the regression model shown in Figure 4.1 yielded nonlinear model. Overall, as indicated in the figure above, a strong correlation exists between the Prima 100 LWD and FWD central sensor deflections (D0). A nonlinear regression model is defined in Equation 4.2 as follows.

98 98 D d. Equation 4.2 With R 2 = 0.85, and correlation coefficient R = 0.86, it is important to note that the model was considered statistically significant with a probability level of P = 0.00 < across all test sites. Meanwhile, the nonlinear regression equation at the plate center explored by Horak et al. (2008) produced a moderate correlation (R 2 = 0.61) and the regression equation was reported to be DFWD = (dLWD) Where various layer thickness (75 and 100mm) of sand treated with emulsion constructed on Berea red type sand subbase and subgrade (Horak et al., 2008). The Horak et al. (2008) equation along with the one suggested by the author at the center of loading plate were plotted in Figure 4.2 below: Normalized FWD Sensor Deflection, (mils) y = x R² = Extrapolated LWD Sensor Deflections, (mils) Power (Suggested by the author) Power (Horak et al., 2008) Figure 4.2: DFWD vs. dlwd Correlation, Comparison to, (Horak et al., 2008)

99 Deflections at Radial Offset Distance r = 300mm, (D 1 ) Deflections measured at the second sensor with radial distance of 12-inches about (300mm) from center loading plate of both FWD and LWD were compared. The plot of regression model is presented in Figure 4.3. The regression analysis yielded a nonlinear model with a power function that gives the best fit based on the best regression coefficient of determination, R 2 = Normalized FWD Sensor Deflections, (mils) FWD versus LWD at r = 300mm (D1) y = x R² = Extrapolated LWD Sensor Deflections, (mils) Figure 4.3: Comparison of FWD and LWD Deflections at r = 300mm from the Center of Loading Plate, (D1) As a result, Figure 4.3 indicates that a correlation was made with the correlation coefficient, R = 0.82, between the Prima 100 LWD and FWD sensor deflection at r = 300mm radial distance from center of loading plate. The Prima 100 LWD sensor

100 deflections (D1) correlates better to FWD. A nonlinear regression model with a power function is defined in Equation 4.3 below: 100 D D. Equation 4.3 With R 2 = 0.78, and correlation coefficient R = It is important to note that the model was considered statistically significant with a probability level of P = 0.00 < across all test sites Deflections at Radial Offset Distance r = 600mm, (D 2 ) Lastly, as the LWD has geophones only up to 24-inches (600mm) from the center of the loading plate, therefore, the sensor deflections measured at the third sensor, D2, were also compared with FWD. The regression analysis yielded a linear model that is presented in Table 4.1 below: Table 4.1: Statistical Analysis Model Summary of FWD vs. LWD Sensor Deflections (D2). Model Summary b Std. Error Change Statistics R Adjusted Model R of the R Square F Sig. F Square R Square df1 df2 Estimate Change Change Change a a. Predictors: (Constant), LWD Sensor Deflections (D 2 ) b. Dependent Variable: FWD Sensor Deflections (D 2 ) Overall, as indicated in Table 4.1, a strong correlation with the correlation coefficient, R = 0.85, exists between the Prima 100 LWD and FWD sensor deflections (D2). The coefficient of determination, shown in column three is a measure of how much

101 101 of the variability in the outcome is accounted for by the predictors. For this model, its value is about, R 2 = 0.72, which means that the LWD sensor deflections (D2) accounts for 72% of the variation in the FWD sensor deflections. The adjusted R 2 in column four of the above table gives us an idea of how good our model generalizes while typically, its value should be the same, or very close to the value of R 2. In this example the difference is very small (the difference between the values is = 0.001, or 0.1%). The standard error of 2.45 is reported with the significance level of R 2 which can be tested using F-ratio. So, this model example causes R 2 to change from 0 to 0.72, and this change in the amount of variance explained gives rise to an F-ratio of , which is significant with a probability level of, P = 0.00 < 0.001, means the model is considered statistically significant (Field, 2013). To better illustrate the relationship between FWD and LWD sensor deflection, a separate plot of (D2) regression model was developed and presented in Figure 4.4 below which gives the best fit based on the best regression coefficient of determination value.

102 102 Normalized FWD Sensor Deflections, (mils) FWD versus LWD at r = 600mm, (D 2 ) y = x R² = Extrapolated LWD Sensor Deflections, (mils) Figure 4.4: Comparison of FWD and LWD Deflections at r = 600mm from the Center of Loading Plate, (D2) Figure 4.4 shows a similar and better linear relationship at radial offset distance of 600mm. A linear regression model is defined in Equation 4.4 D D Equation 4.4 With R 2 = 0.72, and correlation coefficient R = 0.85, it is important to note that the model was considered statistically significant with a probability level of P = 0.00 < across all test sites.

103 Area Under Pavement Profile (Deflection Basin Parameter) Upon positive relationships between FWD and LWD sensor deflections in the previous section, the author was interested to examine if the area under pavement profile between AUPP (FWD 4 sensors) and AUPP (LWD 3 sensors) has any relationship. Therefore, for LWD results at radial distances 0, 12, and 24 inches about (0, 300, and 600mm) were considered rather than 0, 12, 24, and 36 inches from the load center. The modified area under pavement profile is now indicated in Figure 4.5 as follows: Figure 4.5: AUPP (LWD 3 Sensors) Modified Deflection Basin Parameter From the results of Figure 4.5 presented above, all LWD sensor deflections were normalized to 9000 pounds. The new AUPP deflection basin shape parameter model is shown in Equation 4.5.

104 104 AUPP 1 2 3D 2D D Equation 4.5 Where: D0 = FWD sensor deflection at the center of the loading plate, mils D1 = FWD sensor deflection 12 inches from the center of the loading plate, mils D2 = FWD sensor deflection 24 inches from the center of the loading plate, mils A comprehensive regression analysis was performed to find the relationship between AUPP (FWD 4 Sensors) and AUPP (LWD 3 Sensors). The findings from regression analyses support the findings demonstrated in the previous section. The results in Figure 4.6 below show evidence of the correlation between AUPP s. Fitting a linear trendline to the data reveals a high correlation across all sites. AUPP FWD versus AUPP LWD AUPP (FWD 4-Sensors) y = x R² = AUPP (LWD 3-Sensors) Figure 4.6: AUPP Comparison of FWD and FWD across All Sites

105 105 As indicated in Figure 4.6, regression analysis yielded a nonlinear model with a power function that gives the best fit based on the best regression coefficient of determination, R 2 = A nonlinear regression model is defined in Equation 4.6 as: AUPP AUPP. Equation 4.6 With R 2 = 0.83, while correlation coefficient R = It is important to mention that the model was considered statistically significant with a probability level of P = 0.00 < Overall, the model presented in Equation 4.6 between FWD (AUPP) and LWD (AUPP) demonstrated high relationships. The results presented in Equation 4.6 was adopted and substituted into Equations 2.3 and 2.4. The tensile strain at the bottom of the asphalt layer (εac), for full-depth asphalt is computed from Equation 4.7 in term of LWD measurements as below: Log ε Log AUPP Equation (4.7 Similarly, for aggregate base pavements, the tensile strain can be predicted using Equation 4.8 as follows: Log ε Log AUPP Equation 4.8

106 106 Where: εac = tensile strain at the bottom of the HMA layer, macrostrain AUPP = Area under Pavement Profile based on LWD 3-sensors deflection, mils. This method of predicting the (εac) is to use the AUPP value from LWD measurements with 3 sensors. The suggested models presented in Equations 4.7 and 4.8 can be used to design overlays and predict remaining life of pavement sections without using backcalculation technique. 4.5 Comparison of Backcalculated Layer Moduli The next component in correlations between FWD and LWD within this study was the backcalculated layer moduli of pavement layer system. For all sites, the FWD backcalculated layer moduli were plotted against backcalculated layer moduli measured with Prima 100 LWD in the box plot of the previous chapter (Figure 3.19 and Figure 3.20, respectively). The SPSS software was used to perform a comprehensive regression analysis on backcalculated layer moduli in order to find the best correlation between backcalculated layer moduli obtained from Modulus 6.0 and Evercalc 5.0. Also, the subgrade modulus obtained from Rohde method was correlated with the backcalculated subgrade modulus. The results of backcalculated layer moduli upon FWD and LWD measurements are now plotted and presented based on material properties in Figure 4.7 below:

107 FWD vs. LWD, Layer Moduli Layer Moduli (ksi) LWD FWD Material Type Figure 4.7: Backcalculated Layer Moduli of Pavement Layers Based on FWD and LWD Measurements Figure 4.7 shows a column chart of the averaged backcalculated layer modulus based on material type. Comparing the FWD modulus with the LWD modulus in the figure above, it is noted, the FWD modulus is typically slightly less than or somewhat equal to the LWD modulus. However for two materials, lime FDR and concrete steel, they are slightly greater than the LWD modulus. To find a better relationship of layer moduli from FWD and LWD deflection data, statistical analysis was conducted. A model summary from the SPSS outputs is shown in Table 4.2.

108 Model Table 4.2: Statistical Analysis, Model Summary of FWD & LWD Procedures. Model Summary b R R Square Adjusted R Square R Square Change Change Statistics F Change df1 df2 Sig. F Change a A discussion of model summary was previously provided with Table 4.1. Table 4.2 herein generated from SPSS is a model summary which describes the overall model and it tells us whether the LWD is successful in predicting FWD. The Table also provides, R = 0.98 and the R 2 = This tells us that the layer modulus of LWD account for 96% of the variation in the layer modulus of the FWD. Also, the model causes R 2 to change from 0 to 0.96, and this change in the amount of variance explained, gives rise to an F-ratio of , which is considered statistically significant across all test sites with a probability level of P = 0.00 < Similar and strong relationship (R 2 = 0.96) provided in Figure 4.8 indicates fitting a linear trendline to all materials. This means the backcalculated layer modulus of LWD, ELWD (ksi), predicts the FWD backcalculated layer modulus, EFWD (ksi). The results are presented below:

109 FWD versus LWD, Layer Moduli y = x R² = PCC White Topping HMA Mechanical FDR Fiber Cement Cement FDR Layer Modulus, FWD (ksi) asphalt/cement Concrete Steel Full Depth Grindings Partial Depth Grindings Lime FDR Permazine FDR Fabric Reinforced Stone Fly Ash FDR Brick & 411 Surge & 411 Motorpave Layer Modulus, LWD (ksi) Asphalt FDR Geogrid Aggregate Overly Linear (Best Fit Model) Figure 4.8: Regression Analysis Fitting Linear Trendline to Data Points The results presented in Figure 4.8 demonstrated the dependent variable, FWD layer modulus, was highly predicted by the independent variable, LWD layer modulus. Results of regression analysis yielded linear model shown in Equation 4.9. E E Equation 4.9

110 110 It is important to note that the suggested model, Equation 4.9 is compatible with the models proposed by Steinert et al. (2005), Nazzal (2003), and Fleming et al. (2000). Their proposed model equations are indicated as follows: E E Equation 4.10 (Steinert et al., 2005), equation with R 2 = 0.94 E 0.97 E Equation 4.11 (Nazzal, 2003), equation with R 2 = 0.94 E 1.03 E Equation 4.12 (Fleming et al., 2000), equation with R 2 = The above three equations along with the one suggested by the author were combined to a single plot shown in Figure 4.9 below:

111 111 Layer Modulus, FWD (ksi) y = x R² = Linear (Suggested by the author ) Linear (Steinert et al., 2005) Linear (Nazzal, 2003) Linear (Fleming et al., 2000) Layer Modulus, LWD (ksi) Figure 4.9: EFWD vs. ELWD, Comparison to Steinert et al. (2005), Nazzal (2003), and Fleming et al. (2000) Comparison of Subgrade Moduli Subgrade modulus of the FWD measurement was compared with the subgrade modulus of the LWD. To better understand the differences, a floating column plots was prepared similarly for each method as indicated in Figure 4.10 through Figure These plots were developed based on the volume of the collected and analyzed data in this study. Consequently a columns chart was produced to compare the analyses results.

112 112 Modulus of the Subgrade (ksi) Defiance Augaize + Mercer (FWD AASHTO 5.4.5) Champaign + Madison Muskingum Harrison + Cluster Carroll Figure 4.10: FWD Measured Modulus of the Subgrade. Values Indicated are Minimum; Mean; and Maximum Respectively. (1 ksi = 6.89 MPa) Figure 4.10 presents the subgrade modulus of the FWD measurements. Values indicated at the bottom of the column is the minimum resilient modulus, the top of the column is the maximum resilient modulus, and the numbers shown in each column bar near the point are the mean values.

113 113 Modulus of the Subgrade (ksi) (LWD AASHTO 2.3.5) Defiance Augaize + Champaign + Muskingum Harrison + Mercer Madison Carroll Cluster Figure 4.11: LWD Measured Modulus of the Subgrade. Values Indicated are Minimum; Mean; and Maximum Respectively. (1 ksi = 6.89 MPa) Modulus of the Subgrade (ksi) (FWD Rohde Method) Defiance Augaize + Champaign + Muskingum Harrison + Mercer Madison Cluster Carroll Figure 4.12: Rohde Method Measured Modulus of the Subgrade. Values Indicated are Minimum; Mean; and Maximum Respectively. (1 ksi = 6.89 MPa)

114 114 The results in Figure 4.10 through Figure 4.12 demonstrate that FWD subgrade modulus in the first and second clusters have a weak subgrade modulus when compared to other clusters. This is consistent with the results from LWD and Rohde methods. Floating column plots reveal from the FWD results an average values of ksi (88.8 MPa), and ksi ( MPa) respectively. Moreover, in the remaining three clusters, Champaign/Madison, Muskingum, and Harrison/Carroll, the average subgrade modulus values were ascending based on clusters variation subsequently in all three comparisons. 4.6 Comparison of Layer Coefficients A regression analysis was also performed to better understand the variation in FWD and LWD corresponding to layer coefficients. AASHTO FWD was compared with AASHTO LWD. The results yielded a linear relationship indicated in Figure AASHTO FWD AASHTO FWD vs. AASHTO LWD Layer Coefficients 0.80 y = 0.735x R² = AASHTO LWD PCC Whitetopping HMA Mechanical FDR Fiber Cement Cement FDR asphalt/cement Concrete Steel Full Depth Grindings Partial Depth Grindings Lime FDR Permazine FDR Fabric Reinforced Stone Fly Ash FDR Brick & 411 Surge & 411 Motorpave Asphalt FDR Geogrid Aggregate Overly Linear (Best Fit Model) Figure 4.13: Regression Analysis Fitting Linear Trendline to All Layer Coefficients Obtained from AASHTO FWD & AASHTO LWD Methods

115 115 The model shown in Figure 4.13 reveal that the AASHTO FWD has a high correlation (R 2 = 0.78) with AASHTO across all sites. The linear regression equation was reported as follows: a a Equation 4.12 Also, layer coefficients obtained from AASHTO LWD was correlated with the layer coefficients of the Rohde method. The regression analysis yielded a linear model. To better explain this, a separate plot was created for all sites corresponding to material properties. The plot is shown in Figure 4.14 below: Rohde Method AASHTO LWD versus Rohde Method Layer Coefficients y = x R² = AASHTO LWD PCC White Topping Mechanical FDR Fiber Cement Cement FDR asphalt/cement Concrete Steel Full Depth Grindings Partial Depth Grindings Lime FDR Permazine FDR Fabric Reinforced Stone Fly Ash FDR Brick & 411 Surge & 411 Motorpave Asphalt FDR Geogrid Aggregate Overly Linear (Best Fit Model) Figure 4.14: Regression Analysis Fitting Linear Trendline to All Layer Coefficients Obtained from AASHTO LWD and Rohde Method

116 116 As presented in Figure 4.14, a proposed linear model has high correlation (R 2 = 0.82). This reveals that the Rohde method tends to improve/confirm correlation of FWD versus LWD. For the convenience of the reader, both models were plotted in a single plot. This is presented in Figure 4.15 below: 0.80 AASHTO FWD y = x R² = Linear (FWD versus LWD) Linear (Rohde Mehod) AASHTO LWD Figure 4.15: FWD vs LWD Layer Coefficient Models, Comparison to Rohde Method The results in Figure 4.15 demonstrate that the FWD layer coefficients are highly correlated with LWD layer coefficients. It is important to understand that both linear trendlines indicate that the relationship is compatible. This was confirmed by the model developed by Rohde method.

117 Comparison of Effective Structural Numbers Effective structural numbers were determined based on AASHTO 1993, section design guide for the pavement structures, and the Rohde method of determination for the pavement structural number from FWD measurements. A separate single column chart was developed to compare the effective structural numbers. The chart for the first cluster (Defiance County), is shown in Figure 4.16 while the remaining column charts are available in appendix F of this study. Effective Structural Number AASHTO equations versus Rohde method AASHTO Equations ROHDE Method Defiance County Roads Figure 4.16: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Defiance County The results in Figure 4.16 demonstrate that the AASHTO effective structural numbers are slightly greater than, or somewhat equal to the Rohde effective structural

118 118 numbers for almost the entire test sites in the first cluster. However, for a few test sites, the Blosser-C72-07, Blosser-C-15, and Flory-C68-08 the result was opposite. In order to understand a better relationship between these two methods, a regression analysis was performed which yielded a linear model shown in Figure AASHTO Equations Effective Structural Number AASHTO Equations vs. Rohde Method y = x R² = Rohde Method Figure 4.17: Regression Model of Effective Structural Numbers Obtained from, the AASHTO Equations and the Rohde Method The model and fitting linear trendline to the data points presented in Figure 4.17 reveal a very high correlation (R 2 = 0.95) across all test sites. On the other hand, the dependent variable, AASHTO equations is highly predicted by the independent variable, the Rohde method. The regression equation for the effective structural numbers was suggested to be:

119 119 SN SN Equation 4.13 Table 4.3 provides a summary of aforesaid regression equations developed by the author to relate FWD and LWD parameters across all test sites. It is important to mention that fitting trendline varied between linear and power based on high regression correlation value of R 2, to see which one generates the best fit to the model.

120 120 Table 4.3: Summary of Regression Analysis of FWD versus LWD Generated from Developed Models Parameter Index Regression Equation Best Fitting Trendline (R 2 ) Deflections at Radial Offset Distance of 0, 300, and 600mm from the Center of the Loading Plate. D o D FWD = (d LWD )^ Power 0.85 D 1 D FWD = (d LWD )^ Power 0.78 D 2 D FWD = (d LWD ) Linear 0.72 Area Under Pavement Profile (AUPP). AUPP (AUPP) FWD 4-Sensors = *[(AUPP) LWD 3-Sensors ] Power 0.83 Backcalculated Layer Moduli E E FWD = (E LWD ) Linear 0.96 Layer Coefficients a a FWD = 0.735(a LWD ) Linear 0.78 Effective Structural Numbers SN (eff) SN(eff) FWD = (SN(eff) LWD ) Linear 0.95

121 121 CHAPTER 5 CONCLUSION AND RECOMMENDATIONS 5.1 Summary The principal goals in this study were to determine the structural adequacy of the low-volume road using nondestructive test (NDT) technology, and to investigate the potential use the Light Weight Deflectometer (LWD). Regression analysis was conducted using the SPSS statistical software between the FWD and Prima 100 LWD to evaluate whether LWD could be employed to measure the stiffness/strength parameters of the existing local pavement materials and embankment. The data was used to correlate: sensor deflections, backcalculated layer moduli, layer coefficients, and the effective structural numbers. Modulus 6.0 and Evercalc 5.0 were chosen to perform backcalculation analyses on pavement layers. The results of the relationship between FWD and LWD was corroborated by the Rohde method for the layer coefficients and subgrade modulus across all test sites. Also, In the course of this study, a modified method of calculating Area under Pavement Profile (AUPP) was devised. 5.2 Conclusion The results presented in this thesis demonstrate a good correlation exists between FWD and LWD. These findings correspond with but not limited to AASHTO 1993 Guide, statistical analysis outputs, and the Rohde method. Statistical analysis demonstrates and suggests that the LWD could be reliably used to evaluate low-volume roads systems. The correlation coefficients (R) and coefficient of determination (R 2 ) values vary from (0.85 to 0.98) and (0.72 to 0.96), respectively for all correlated parameters.

122 122 The results highlight several significant findings. First, in the comparison between FWD and LWD sensor deflections at the center of loading plate, and sensor deflections at r = 300mm, yield a nonlinear model (power function) with quite a high relationship (R 2 =0.85 and R 2 = 0.78) respectively. This appears to be consistent for all sites. On the contrary, the relationship between sensor deflections at radial offset distance, r = 600mm, yield a linear regression model with a moderate relationship (R 2 = 0.72). Also, it was found that FWD and LWD sensor measurements at 0, 300, and 600mm are influenced by the material behavior. This means that on soft soil surfaces (typically aggregate overlay in this study), the FWD and LWD relationships result in a much lower R 2 or deflections identified to be outliers. Also, the results for the Area under Pavement Profile (AUPP) yielded a nonlinear model with moderate correlation (R 2 = 0.83) for the modified AUPP. This modification at radial offset distances 0, 300, and 600mm from the load center using LWD measurements, now appears to be a new valid parameter for overlay design, and it can be used to predict tensile strain on the bottom of an asphalt layer. The subgrade modulus results reveal, soil stiffness/strength consistently increase by cluster variation from one, two, and so forth. The subgrade modulus values in the first cluster, using FWD, LWD, and Rohde method, was found to be ksi (88.8 MPa), ksi (87.36 MPa), and ksi (91.42 MPa) respectively. In the second cluster, Auglaize/Mercer, the LWD obtained subgrade modulus seems to be the weakest subgrade with an average value of ksi (71.77 MPa). However, this cluster had approximately the same subgrade modulus with average values of ksi ( MPa) and ksi

123 123 (96.52 MPa) from the FWD and Rohde methods respectively. Moreover, for the remaining three clusters, Champaign/Madison, Muskingum, and Harrison/Carroll, average modulus of subgrade values were consistently ascending, based on clusters variation across all comparison parameters. In the comparison of effective structural numbers, the AASHTO method results were slightly greater than the Rohde method, or somewhat equal to the Rohde effective structural numbers in all sites. This consistency was compatible with and supported by the Rohde procedure, since he consider the contribution of the subgrade to SN. Finally, in the correlation of the layer coefficient between AASHTO FWD versus AASHTO LWD; the AASHTO FWD was closely predicted by AASHTO LWD. AASHTO LWD could account for 78.0% of the variation in AASHTO FWD. Although the correlation coefficient was high enough, R = 0.88, a moderate fitting linear trendline of the model was established between the AASHTO FWD and AASHTO LWD. Also, in the correlation between Rohde and AASHTO LWD, the Rohde method was predicted very well by the AASHTO LWD. This method accounts for 81.67% of variation in the Rohde procedure and correlation coefficient was found to be R = 0.90; so, a fitting linear trendline to the model was established between the Rohde and AASHTO LWD. Finally, statistical analysis proved that all the correlation models were highly correlated with each other and were considered statistically significant with a probability level of P = 0.00 < The aforementioned observations, as major points with regard to the acceptability of the Prima 100 LWD, as a nondestructive device for low traffic volume follows:

124 The Prima 100 LWD device, with a 300mm diameter load bearing plate, is usable for characterizing pavement layer moduli in the local pavement system. Specifically, because the benchmark test, FWD, layer moduli are highly correlated with the LWD layer moduli, and the FWD is economically prohibitive device for local agencies. The results of this study confirm the hypothesis the LWD is effective and adequate to evaluate in-situ deformation parameters during local pavement investigations, and has the advantage of portability, cost effectiveness, and ease of use. 2. The results presented in this study demonstrate that sensor deflections at the center of loading plate, and the sensor deflections at r = 300mm have high relationship (R2 =0.85 and R2 = 0.78) between FWD and LWD respectively. However, the correlation of sensor deflections at radial offset distance, r = 600mm has a moderate relationship (R2 = 0.72). 3. The author modification of the Area Under Pavement Profile is valid for the Prima 100 LWD measurement in the evaluation of pavement responses at a radial offset distance of 0, 12, and 24 inches (0, 300, and 600mm) from the center of the loading plate. This modification appears to be a new valid parameter for the overlay design, and it can be used to predict tensile strain on the bottom of an asphalt layer. 4. The Prima 100 LWD, is highly affected by inadequate/inaccurate seating of the bearing plate (300mm in this study) on the pavement surfaces. Moreover, radial geophone configuration of 300 and 600mm records critical deflections to the

125 125 backcalculation process, and also produced the most accurate layer moduli backcalculation results. 5. The FWD and the Prima 100 LWD measurements are less repeatable on a rough or soft surfaces compared to a stiff or hard surfaces. Measured deflections on rough or soft surfaces (typically aggregate overlay in this study) from multiple sensors across all test sites resulted in a much lower R2, or somewhat identified to be outliers in the data. 6. Based on observations made during testing, adjustment/measurements of radial sensors, the PDA device Bluetooth connection, and verticality of the guide rod are other factors affecting the Prima 100 LWD results. Also, test operators have to be trained enough to identify and avoid incorrect device reading. 7. As shown in the column chart, Figure 4.7, the Evercalc 5.0 software consistency for the backcalculated modulus from LWD measurements is slightly higher compared to Modulus 6.0 software consistency of the FWD measurements. 5.3 Recommendations This study investigated the feasibility of using Prima 100 LWD to evaluate pavement performances such as: layer moduli, effective structural numbers, and layer coefficients for the low volume roads. Based on the objectives set for this study, the suggested recommendations are drawn as follows: 1. Statistical analysis demonstrated that all regression models are highly correlated with each other and were agreed with those studied previously. Thus, derived

126 126 equations are recommended to use with confidence for structural analyses of the pavement systems in the local roads with the LWD. 2. Follow up studies should investigate the use of the LWD on rough surfaces and soft soil to reevaluate variability of measurements. 3. Finite element analysis is recommended for further correlation study between these nondestructive devices. Such analyses and should consider the rate of stiffness evaluation at various stress and strain levels. 4. It is recommended that laboratory determined resilient moduli should be correlated with the moduli obtained from FWD and LWD. This may develop or modify precise shift factors between field backcalculated and laboratory estimated layer resilient moduli and also revalidate the presented correlation in this study. 5. The author does not prescribe any specific backcalculation software for LWD deflection data analysis. However, particular caution against using any software is needed in the conventional sense in addition of knowing that, incorrect input parameters result in incorrect outcomes. 6. It is highly recommended the use of the modified AUPP should be investigated. The correlations should be made between additional deflection basin parameters and pavement responses. 7. Lastly, further research should include conducting the correlation between FWD and LWD measurements with different plate size and drop height while maintaining the same radial offset distance. This results to better understand the effects of influence depth on the underlying layer systems.

127 127 REFERENCES Ahmed, M. U. (2010). Evaluation of FWD software and deflection basin for airport pavements. Master thesis, Civil Engineering Department, The University of New Mexico, Albuquerque, New Mexico. Alavi, S., LeCates, J. F., & Tavares, M. P. (2008). Falling Weight Deflectometer Usage. A Synthesis of Highway Practice. National Cooperative Highway Research program (NCHRP). Transpiration Research Board, Washington, D.C. Al-Jhayyish, A. K. (2014). Incorporating Chemical Stabilization of the Subgrade in Pavement Design and Construction Practices. Master thesis, Civil Engineering Department, Ohio University, Athens, Ohio. American Association of State Highway and Transportation Officials. (1993). AASHTO Guide for Design of Pavement Structures, Washington, D.C. Appea, A. K. (2003). Validation of FWD testing results at the Virginia smart road: Theoretically and by instrument responses. PhD dissertation, Civil Engineering Department, Virginia Polytechnic Institute and State University. Blacksburg, Virginia. ASTM D , (2015). Standard Guide for Calculating in Situ Equivalent Elastic Moduli of Pavement Materials Using Layered Elastic Theory (volume Road and paving materials; vehicle-pavement systems). ASTM International, West Conshohocken, PA. Commuri, S., Zaman, M., Beainy, P. E. F., Singh, D., Nazari, M., Imran, S., & Barman, M. (2012). Pavement Evaluation Using a Portable Lightweight Deflectometer. (Report No. OTCREOS F). Oklahoma Transportation Center, Oklahoma. Crovetti, J. A., Shahin, M. Y., & Touma, B. E. (1989). Comparison of two falling weight deflectometer devices, Dynatest 8000 and KUAB 2M-FWD. In Nondestructive Testing of Pavements and Backcalculation of Moduli. ASTM International, West Conshohocken, PA, (pp ). Elhakim, A. F., Elbaz, K., & Amer, M. I. (2014). The use of light weight deflectometer for in situ evaluation of sand degree of compaction. HBRC Journal, 10(3), pp Field, A. (2013). Discovering statistics using IBM SPSS statistics. Third edition. Sage Publications Ltd, University of Sussex, England.

128 128 Fleming, P. R., Frost, M. W., & Lambert, J. P. (2009). Lightweight deflectometers for quality assurance in road construction. In IN: Tutumluer, E. and Al-Qadi, IL (eds). Bearing Capacity of Roads, Railways and Airfields: Proceedings of the 8th International Conference (BCR2A 09) (pp ). George, K. P. (2006). Portable FWD (PRIMA 100) for in-situ subgrade evaluation. (No. FHWA/MS-DOT-RD ). University of Mississippi. Gopalakrishnan, K., Kim, S., & Ceylan, H. (2010). Non-destructive evaluation of in-place rehabilitated concrete pavements. Journal of Civil Engineering and Management, 16(4), Gopalakrishnan, K. (2009). Backcalculation of non-linear pavement moduli using finiteelement based neuro-genetic hybrid optimization. Open Civil Engineering Journal, 3, Gopalakrishnan, K., & Thompson, M. R. (2007). Pavement rutting characterization using deflection basin parameters. Indian Journal of Engineering & Materials Sciences, 14, Grau, R. H., & Alexander, D. R. (1994). Nondestructive Testing, Evaluation, and Rehabilitation for Roadway Pavement: Warren County, Mississippi/ Cincinnati, Ohio/ and Berkeley, California. Guide, E. U. (2005). Pavement analysis computer software and case studies. Washington State Department of Transportation, Everseries User s Guide, Olympia, Washington. Guthrie, W., Cooley, D., & Eggett, D. (2007). Effects of reclaimed asphalt pavement on mechanical properties of base materials. Transportation Research Record: Journal of the Transportation Research Board. Volume. 2005, (pp ). Hoaglin, D. C., & Iglewicz, B. (1987). Fine-tuning some resistant rules for outlier labeling. Journal of the American Statistical Association, 82(400), Horak, E., & Emery, S. (2006). Falling weight deflectometer bowl parameters as analysis tool for pavement structural evaluations. In Research into Practice: 22nd ARRB Conference. Horak, E., Maina, J. W., Guiamba, D., & Hartman, A. (2008). Correlation study with the light weight deflectometer in South Africa. Hossain, M. S., & Apeagyei, A. K. (2010). Evaluation of the Lightweight deflectometer for in-situ determination of pavement layer moduli. (No. FHWA/VTRC 10-R6).

129 Huang, Y. H. (2004). Pavement analysis and design. Second edition. Pearson prentice Hall. University of Kentucky. 129 Icenogle, P., & Kabir, M. S. (2013). Evaluation of Non-Destructive Technologies for Construction Quality Control of HMA and PCC Pavements in Louisiana. (No. FHWA/LA. 12/493). Report. Baton Rouge, LA: Louisiana Department of Transportation. Janoo, V. C. (1994). Layer Coefficients for NHDOT Pavement Materials. (No. Crrel-SP ). Cold Regions Research and Engineering Lab Hanover NH. Jordan, B. B. (2013). Asphalt Perpetual Pavement Design: Utilizing Existing Pavement Systems in Ohio. Master thesis, Ohio University. Keller, G., & Sherar, J. (2003). Low-Volume Roads Engineering: Best Management Practices. Transportation Research Record: Journal of the Transportation Research Board, (1819), Kim, Y. R., & Park, H. (2002). Use of FWD multi-load data for pavement strength estimation. (Report No. FHWA/NC/ for the North Carolina Department of Transportation). Raleigh, NC. Mehta, Y., & Roque, R. (2003). Evaluation of FWD data for determination of layer moduli of pavements. Journal of Materials in Civil Engineering, 15(1), Mooney, M. A., Surdahl, R. W., Grasmick, J. G., Senseney, C. T., & Voth, M. (2015). Comparison of Multiple Sensor Deflection Data from Lightweight and Falling Weight Deflectometer Tests on Layered Soil, Geotechnical Testing Journal, Vol. 38, No. 6, 2015, (pp ). Murillo Feo, C. A., & Urrego, L. E. B. (2013). Correlation between deflections measurements on flexible pavements obtained under static and dynamic load techniques. In Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris (pp ). Nazzal, M. D. (2007). Field evaluation of in-situ test technology for Q (C)/Q (A) during construction of pavement layers and embankments. Master thesis, The Department of Civil and Environmental Engineering, Louisiana State University and Agricultural and Mechanical College, Baton Rouge, Louisiana. Pologruto, M. (2004). Study of In Situ Pavement Material Properties Determined from Falling-Weight Deflectometer Testing. In Transportation Research Board 85th Annual Meeting. (No ).

130 130 Qin, J. (2010). Predicting Flexible Pavement Structural Response Using Falling Weight Deflectometer Deflections. Master thesis, Civil Engineering Department, Ohio University, Athens, Ohio. Rada, G., Nazarian, S., Visintine, B., Siddharthan, R. V., & Thyagarajan, S. (2015). Pavement structural evaluation at the network level. Final Rep., FHWA, Washington, DC. Rohde, G. T. (1994). Determining Pavement Structural Number from FWD Testing, TRR Transportation Research Board, Washington DC. Romanoschi, S., Hossain, M., Gisi, A., & Heitzman, M. (2004). Accelerated pavement testing evaluation of the structural contribution of full-depth reclamation material stabilized with foamed asphalt. Transportation Research Record: Journal of the Transportation Research Board, (1896), Ryden, N., & Mooney, M. A. (2009). Analysis of surface waves from the light weight deflectometer. Soil Dynamics and Earthquake Engineering, 29(7), Samb, F., Fall, M., Berthaud, Y., & Bâ, M. (2013). Resilient modulus of compacted lateritic soils from senegal at OPM conditions. Geomaterials, 3(04), 165. Sargand, S., Green, R., Burhani, A., Alghamdi, H., & Jordan, B. (2016). Investigation of In-Situ Strength of Various Construction / Widening Methods Utilized on Local Roads, (134991). Senseney, C., & Mooney, M. (2010). Characterization of two-layer soil system using a lightweight deflectometer with radial sensors. Transportation Research Record: Journal of the Transportation Research Board, (2186), Shafiee, M. H., Nassiri, S., & Khan, M. R. H. (2013). Evaluation of New Technologies for Quality Control/Quality Assurance (QC/QA) of Subgrade and Unbound Pavement Layer Moduli. Steinert, B. C., Humphrey, D. N., & Kestler, M. A. (2005). Portable falling weight deflectometer study. (Report No. NETCR52), Department of Civil and Environmental Engineering, University of Maine, Orono, Maine. Tang, S., Cao, Y., & Labuz, J. F. (2012). Structural Evaluation of Asphalt Pavements with Full-Depth Reclaimed Base. (Report No. MN/RC ), Department of Civil Engineering, University of Minnesota, Minneapolis, MN.

131 131 Tawfiq, K. (2003). Utilizing the falling weight deflectometer in evaluating soil support values of pavement layers. (Accession No ), National Technical Information Service, Alexandria, VA. Thompson, M. (1999). Hot-mix asphalt overlay design concepts for rubblized Portland cement concrete pavements. Transportation Research Record: Journal of the Transportation Research Board, (1684), Tukey, J. W. (1977). Exploratory data analysis.-reading, Mass.: Addison-Wesley. Tutumluer, E., Pekcan, O., & Ghaboussi, J. (2009). Nondestructive pavement evaluation using finite element analysis based soft computing models. (NEXTRANS Project No. 010IY01), Department of Civil and Environmental Engineering, University of Illinois, Urbana-Champaign, Chicago. Von Quintus, H. L. (2009). NDT technology for quality assurance of HMA pavement construction National Cooperative Highway Research program (NCHRP). Transpiration Research Board, (Volume 626). Washington, D.C.

132 132 APPENDIX A: PAVEMENT LAYER THICKNESSES AND MATERIAL PROPERTIES BY COUNTY. Table A1: Layer Thicknesses and Material Properties, Defiance County Layer 1 Layer 2 Layer 3 Road Name Core ID Thickness (in) Material Thickness (in) Material Thickness (in) Material Rosedale C117 (A) Rosedale C117 (B) Harding C195 Rosedale C117 Willams Center Cecil C123 Hammon T187 Blosser C72 Blosser C72 Flory C HMA HMA HMA HMA HMA Chip Seal Chip Seal Chip Seal HMA Fiber Cement FDR Cement FDR Asphalt FDR Cement FDR Cement FDR Cement FDR Cement Fiber Cement Fiber Cement * * * * * * * * * Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade

133 Banner School- C70 Elliot C53 Banner School Rd C70 The Bend C134 Krouse- C Mansfield C6 Mansfield C6 Blosser C72 Fountain Street C39 Table A1: Continued Fiber HMA Cement Fiber HMA Cement HMA Fiber Cement Fiber HMA Cement Fiber HMA Cement Fiber HMA Cement HMA Chip Seal Chip Seal Full Depth Grindings FDR Cement Fiber Cement * * * * 133 Aggregate Base Aggregate Base Natural Subgrade Aggregate Base Aggregate Base Aggregate Base Natural Subgrade Natural Subgrade Natural Subgrade

134 134 Williams Center Cecil C123 ( BX1, BX2, BX3) Elliot C53 Christy C164 Table A1: Continued C BX BX3 1.5 Fabric Chip 8.0 Reinforced C2 1.5 Seal 5.0 Stone F C HMA Fabric Reinforced Stone * White HMA 3.0 topping *Placed Layer on the Natural Subgrade, No Thickness Was Measured. * Natural Subgrade Natural Subgrade Aggregate Base

135 135 Road Name Birmingham- (C10) Unionvale- (C12) Bakers Ridge (C51) Plum Run (C8) Table A2: Layer Thicknesses and Material Properties, Harrison County Core ID Thickness (in) Layer 1 Layer 2 Layer 3 Material Thickness (in) HMA HMA HMA HMA Material FDR Asphalt FDR Cement FDR Cement FDR Permazine *Placed layer on natural subgrade, No thickness was measured. Thickness (in) * * * * Material Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade

136 136 Road Name Pledge (T370) Meter (T269-02) Meter- (T269-03) Canyon (C54) Apollo (C12) Chase (C66) Table A3: Layer Thicknesses and Material Properties, Carroll County Core ID Thickness (in) Layer 1 Layer 2 Layer 3 Material Aggregate Overlay Aggregate Overlay Thickness (in) * * HMA HMA HMA HMA 12.0 Material Natural Subgrade Natural Subgrade Cement FDR Cement FDR Cement FDR Cement FDR *Placed Layer on the Natural Subgrade, No Thickness Was Measured. Thickness (in) * * * * * * Material Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade

137 137 Road Name Kossuth Loop (C216A) East Shelby (C71) Blank Pike (C160) Southland (C3) Minster Fort Recovery(C30 ) Fairgrounds Table A4: Layer Thicknesses and Material Properties, Auglaize County Cor e ID Thicknes s (in) Layer 1 Layer 2 Layer 3 Material Thicknes s (in) Full Depth * Grinding s HMA Material Natural Subgrade Full Depth Grinding s Partial HMA Depth Grinding s Partial Depth HMA 5.0 Grinding s Full Depth HMA 3.0 Grinding s Full Depth Grinding s * Natural Subgrade *Placed Layer on the Natural Subgrade, No Thickness Was Measured. Thicknes s (in) * * * * * * Material Natural Subgrad e Natural Subgrad e Natural Subgrad e Natural Subgrad e Natural Subgrad e Natural Subgrad e

138 138 Table A5: Layer Thicknesses and Material Properties, Mercer County Road Name Core ID Layer 1 Layer 2 Layer 3 Thickness (in) Material Thickness (in) Material Material Dutton (C230A) HMA /30 asphalt/cement Natural Subgrade Neptune Mendon Rd. (C161C) HMA /30 asphalt/cement Natural Subgrade Harris (C175B) HMA /30 asphalt/cement Natural Subgrade

139 139 Road Name Pisgah (C236) Heck Hill (C26) Nine Mile (C37) Kite (C22) Sullivan Rd (C45) Lippincott (C115) Old Troy Pike (C193) W. Dallas Rd. (C184) Table A6: Layer Thicknesses and Material Properties, Champaign County Core ID Thickness (in) Layer 1 Layer 2 Layer 3 Material Thickness (in) Material Thickness (in) Mechanical HMA FDR Mechanical HMA FDR Chip FDR * Seal Cement Mechanical HMA FDR Chip Mechanical Seal FDR Chip Seal Chip Seal HMA Mechanical FDR Mechanical FDR Cement FDR Nine Mile Chip 13.0 Mechanical (C37) Seal 13.0 FDR Old Troy Chip Mechanical Pike Seal FDR (C193) *Placed Layer on the Natural Subgrade, No Thickness Was Measured. * * * * * Material Aggregate Base Aggregate Base Natural Subgrade Aggregate Base Aggregate Base Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade Natural Subgrade

140 140 Road Name MCEO Charleston- Chillicothe (C15B) Davis Rd (C95) Taylor Blair-C14- S4 Table A7: Layer Thicknesses and Material Properties, Madison County Core ID Thickness (in) Layer 1 Layer 2 Layer 3 Material Thickness (in) HMA Chip Seal Chip Seal Material Thickness (in) Geogrid * FDR Cement FDR Cement Chip HMA 10.0 Seal Taylor Chip FDR Blair-C Seal Cement N *Placed Layer on the Natural Subgrade, No Thickness Was Measured * * Material Natural Subgrade Natural Subgrade Natural Subgrade Aggregate Base Aggregate Base

141 141 Table A8: Layer Thicknesses and Material Properties, Muskingum County Road Name Co re ID Thickn ess (in) Layer 1 Layer 2 Layer 3 Layer 4 Materia l Thickn ess (in) Materia l Thickn ess (in) Materi al Thickn ess (in) Materi al Airport (797) Arch Hill Dietz Lane (449) Elis Dam(C 49) Friendl y Hills (418) Mt Perry (C30) Narrow s (C76) Aggreg Concret HMA ate e Steel Base Motorp ave Concret e Steel * Natural Subgra de Natural Brick & HMA 6.0 * Subgra de Natural Chip FDR * Subgra Seal Fly Ash de Aggreg Chip FDR ate Seal Lime Base HMA Concret e Steel * Chip 4.5 Motorp Seal 5.0 ave Natural Subgra de Aggreg ate Base * Natural Subgra de * - * - * * Natural Subgra de * - * Natural Subgra de

142 New Hope (C20) Norfiel d (C64) Powels on (C49) Rural Dale (C31) Salt Creek (C44) Vista View AirPark HMA 16.0 Table A8: Continued Surge & 411 * Natural Subgra de Motorp Chip HMA ave Seal Chip Motorp Seal ave Chip Seal FDR Asphalt Brick & HMA PCC PCC Concret e Steel Concret e Steel * * * FDR Lime Natural Subgra de * * Bricks * Natural Subgra de Natural Subgra de Souther Natural Motorp Surge n * Subgra ave & 411 (C107) de *Placed Layer on the Natural Subgrade, No Thickness Was Measured 142 Aggreg ate Base Natural Subgra de * - Natural Subgra de * - * - * -

143 143 APPENDIX B: TYPICAL FWD AND LWD DEFLECTION BASINS FWD Sensor Deflection (mils) lbf 9000 lbf lbf 28 Radial Distance (in) Figure B1: Deflection Basins for Three Loads, Cluster # 2, Section of Minster Recovery Road (Aug-C30-16), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate. LWD sensor Deflection, mils Radial Distance, (in) 2701 lb 2589 lb 2521 lb Figure B2: LWD Deflection Basins Same Loads, Meter Road (CAR-T269-2), Aggregate Overlay Surface Layer, Carroll County, 11.8-in. (300-mm) Plate.

144 FWD Sensor Deflection (mils) Radial Distance (in) lbf 9000 lbf lbf Figure B3: FWD Deflection Basins for Three Loads, Cluster # 2, Section of East Shelby Road (Aug-C71-8), HMA Surface layer, Auglaize County, 11.8-in. (300-mm) Plate FWD Sensor Deflection (mils) lbf 9000 lbf lbf 30 Radial Distance (in) Figure B4: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Blank Pike Road (Aug-C160-12), HMA Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate.

145 145 FWD sensor Deflection (mils) lbf 9000 lbf lbf Radial Distance (in) Figure B5: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Kossuth Loop (Aug-C216A-3), Full depth Grindings layer, Auglaize County, and 11.8-in. (300- mm) Plate. FWD Sensor Deflection (mils) Radial Distance (in) 6000 lbf 9000 lbf lbf Figure B6: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Fairground (Aug-FG-18), Full Depth Grindings Layer, Auglaize County, 11.8-in. (300-mm) Plate.

146 146 FWD Sensor Deflection (mils) lbf 9000 lbf lbf Radial Distance (in) Figure B7: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Neptune Mendon Road (MER-C161C-7), HMA Surface Layer, Mercer County, 11.8-in. (300-mm) Plate. FWD Sensor Deflection (mils) Radial Distance (in) 6000 lbf 9000 lbf lbf Figure B8: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Harris Road (MER-C175B-8), HMA Surface Layer, Mercer County, 11.8-in. (300-mm) Plate.

147 FWD Sensor Deflection (mils) Radial Distance (in) lbf 9000 lbf lbf Figure B9: FWD Deflection Basins for Three Loads, Cluster # 2, Section of Dutton Road (MER-C230A-3), HMA Surface Layer, Mercer County, 11.8-in. (300-mm) Plate. LWD Sensor Deflection, mils Radial Distance, (in) 2336 lb 2395 lb 2360 lb Figure B10: LWD Deflection Basins Same Loads, Cluster # 2, Kossuth Loop (Aug- C216A-3), Full Depth Grindings Surface Layer, Auglaize County, 11.8-in. (300-mm) Plate

148 148 APPENDIX C: AASHTO PROCEDURE OUTPUTS USING FWD SENSOR DEFLECTIONS Table C1: AASHTO Equations Outputs Calculated from FWD Sensor Deflection Using 11.8-in. (300mm) Plate. Road Name Subgrade M R (ksi) Effective E P (ksi) Central Deflection d 0 (mil) Total thickness D (in) Effective Structural Number (S eff ) Road Name Subgrade M R (ksi) Effective E P (ksi) Central Deflection d 0 (mil) Total thickness D (in) Effective Structural Number (SN eff ) Christy- C East Shelb- C Blosser- C72-07 Mt Perry- C30-11 Arch Hill- C82-03 Vista View Air Park Mansfield- C6-14C Blank Pike- C Salt Creek- C44-19 Dietz Ln- C Apollo-C Airport- C Canyon- C

149 149 Table C1: Continued Mansfiel d-c Chase-C Elliott- C Meter-T Elliott- C53-18 Banner School- C70-09 Banner School- C70-11 Blosser- C72-15 Rosedale -C117-01A Rosedale -C117-01B Plum Run- C8-06 Birmingham -C10-02 Unionvale- C12-03 Bakers Ridge-C51-04 Fountain- C39-16 Flory- C Rosedale -C Blosser- C WCC- C WCC-C

150 150 Table C1: Continued The Bend- C Krouse- C Taylor Blair-C14- S4 Taylor Blair-C14- N5 Harding- C Hammon- T Kite-1- C Charleston - C15B Heck Hill- C62-07 Nine Miles- C37-12 Nine Miles- C37-20 Sullivan- C MCEO Davis-C95-03 Rural Dale- C31-18 Ellis Dam- C49-08 Lippinco tt-c Dallas- C Powelson- C49-16 Southland- C

151 151 Table C1: Continued Old Troy Pike- C Old Troy Pike- C Pisgah- C Minster Fort -C30-16 Dutton (C230A) Neptune M- C161C Harris (C175B)

152 152 APPENDIX D: SUMMARY OF BACKCALCULATED LAYER MODULI FROM FWD AND LWD TESTING Table D1: Summary of Averaged Backcalculated Layer Moduli Computed from FWD Sensor Deflections Using 11.8-in. (300mm) Plate, Modulus 6.0 Software. Backcalculated Layer Road Name Moduli (ksi) Road Name Backcalculated Layer Moduli (ksi) E 1 E 2 E 3 E 1 E 2 E 3 E 4 Christy-C Elliott-C Blosser-C a 3 Elliott-C * Mt Perry-C Banner School-C * Arch Hill -C Blosser-C * Vista View Drive Rosedale-C117-01A * Air Park Rosedale-C117-01B * WCC-C Airport-C Lippincott-C a 13 Mansfield-C6-14 & 14C Dallas-C The Bend-C Old Troy Pike-C Krouse-C Old Troy Pike-C a 1 Harding-C * Southland-C Kite-1-C Minster Fort Recovery -C Pisgah-C Nine Miles-C a 4 Heck Hill-C Dutton-C230A Neptune Mendon-C161C * Nine Miles-C a 3 East Shelby-C * Sullivan-C a 14 Blank Pike-C *

153 153 Table D1: Continued Harris-C175B Salt Creek-C Fountain Street-C a 6 Dietz Ln-C * Flory- C a 3 MCEO Office Drive * Blosser-C a 6 Apollo-C * WCC a 12 Canyon-C Hammon-T a 7 Taylor Blair-C14-N5 399 a 10 3 Taylor Blair-C14-S4 626 a 10 Chase-C * Charleston Chillicothe-C15B a 5 Meter-T * Davis-C a 12 Plum Run-C * Rural Dale-C a 3 Unionvale-C * Ellis Dam-C a 4 Powelson-C a Fairground (Center) 26 5 ** Bakers Ridge -C51-04) * Fairground (East) ** Narrows-C a 43 4 Fairground (West) 23 7 ** Friendly Hill-C a 42 8 Pledge-T ** Norfield-C a 6 Meter-T ** Southern-C * New Hope-C20 20 a 11 Birmingham-C * Kossuth Loop-C216A ** Rosedale-C * * Three Layer System; ** Two Layer System; a Top Layer Was Combined With the Bottom Layer

154 154 as following: Similarly, an averaged of backcalculated Layer moduli from LWD sensor deflection using Evercalc 5.0 are shown in table D2 Table D2: Summary of Averaged Backcalculated Layer Moduli Computed from LWD Sensor Deflections Using 11.8-in. (300mm) Plate, Evercalc 5.0 Software. Backcalculated Layer Backcalculated Layer Moduli (ksi) Road Name Moduli (ksi) Road Name E 1 E 2 E 3 E 1 E 2 E 3 E 4 Christy-C Elliott-C Blosser-C a 6 Elliott-C * Mt Perry-C Banner School-C * Arch Hill -C Blosser-C * Vista View Drive Rosedale-C117-01A * Air Park Rosedale-C117-01B * WCC-C Airport-C Lippincott-C a 13 Mansfield-C6-14 & 14C * Dallas-C The Bend-C Old Troy Pike-C a 6 Krouse-C Old Troy Pike-C a 7 Harding-C * Southland-C Kite-1-C Minster Fort Recovery -C Pisgah-C Nine Miles-C a 7 Heck Hill-C

155 155 Table D2: Continued Dutton-C230A Neptune Mendon-C161C * Nine Miles-C a 7 East Shelby-C * Sullivan-C a 12 Blank Pike-C * Harris-C175B Salt Creek-C * Fountain Street-C a 6 Dietz Ln-C * Flory- C a 5 MCEO Office Drive * Blosser-C Apollo-C * WCC a 10 Canyon-C * Hammon-T a 8 Taylor Blair-C14-N5 329 a 15 5 Taylor Blair-C14-S4 572 a 8 Chase-C * Charleston Chillicothe-C15B a 6 Meter-T * Davis-C a 14 Plum Run-C * Rural Dale-C a 6 Unionvale-C * Ellis Dam-C a 6 Powelson-C a Fairground (Center) ** Bakers Ridge -C51-04) * Fairground (East) 27 6 ** Narrows-C a Fairground (West) 41 9 ** Friendly Hill-C a Pledge-T ** Norfield-C a 14 Meter-T ** Southern-C * New Hope-C Birmingham-C * Kossuth Loop-C216A ** Rosedale-C * * Three Layer System; ** Two Layer System; a Top Layer Was Combined With the Bottom Layer

156 156 APPENDIX E: FWD AND LWD SENSOR DEFLECTIONS Table E1: Normalized/Extrapolated to 9000 Pounds Sensor Deflections (D0, D1, and D2) at Radial Offset Distance 0, 12, 24 inches from the Center of the Load. Normalized FWD Sensor Deflections Extrapolated LWD Sensor Deflections D 0 D 1 D 2 D 0 D 1 D 2 D 0 D 1 D 2 D 0 D 1 D 2 (mils) (mils) (mils) (mils) (mils) (mils) (mils) (mils) (mils) (mils) (mils) (mils)

157 157 Table E1: Continued

158 158 Table E1: Continued

159 159 Table E1: Continued

160 Table E2: Deleted Outliers/ Abnormal Sensor Deflections Obtained from FWD and LWD Testing FWD LWD D 0 (mils) D 1 (mils) D 2 (mils) D 0 (mils) D 1 (mils) D 2 (mils) * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *No outlier was identified. 160

161 APPENDIX F: EFFECTIVE STRUCTURAL NUMBERS OF AASHTO EQUATIONS AND THE ROHDE METHOD 161 Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Auglaize County Roads Figure F1: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Auglaize County. Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Mercer County Roads Figure F2: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Mercer County.

162 162 Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Madison County Roads Figure F3: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Madison County. Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Champaign County Roads Figure F4: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Champaign County.

163 163 Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Muskingum County Roads Figure F5: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Muskingum County. Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Carroll County Roads Figure F6: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Carroll County.

164 164 Effective Structural Number AASHTO Equations vs. Rohde Method AASHTO Equations ROHDE Method Harrison County Roads Figure F7: Effective Structural Numbers Based on County Roads, AASHTO Equations versus Rohde Method of Harrison County.

165 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Thesis and Dissertation Services!