INVESTIGATION OF GRADED AGGREGATE BASE (GAB) COURSES

Transcription

1 INVESTIGATION OF GRADED AGGREGATE BASE (GAB) COURSES R.L. Baus T. Li submitted to The South Carolina Department of Transportation and The Federal Highway Administration February 26 Department of Civil and Environmental Engineering 3 Main Street Columbia, SC 2928 (83) cee@engr.sc.edu This research was sponsored by the South Carolina Department of Transportation and the Federal Highway Administration. The opinions, findings and conclusions expressed in this report are those of the authors and not necessarily those of the SCDOT or FHWA. This report does not comprise a standard, specification or regulation. (FHWA/SCDOT Report No. FHWA-SC-6-3)

2 1. Report No. FHWA-SC Government Accession No. 2. Recipient s Catalog No. 4. Title and Subtitle Investigation of Graded Aggregate Base (GAB) Courses 5. Report Date February Performing Organization Code 7. Author(s) R.L. Baus and T. Li 9. Performing Organization Name and Address Department of Civil and Environmental Engineering University of South Carolina Columbia, South Carolina Sponsoring Agency Name and Address South Carolina Department of Transportation P.O. Box 191 Columbia, South Carolina Performing Organization Report No 1. Work Unit No. (TRAIS) 11. Contract or Grant No. 14. Sponsoring Agency Code 15. Supplementary Notes Prepared in cooperation with the Federal Highway Administration 16. Abstract This report summarizes a study undertaken to investigate the feasibility of relaxing current South Carolina Department of Transportation (SCDOT) graded aggregate base (GAB) gradation specifications and layer thickness restrictions. The study included a review of historical and current SCDOT specifications and practices, a literature review and survey of state highway agency practice, and laboratory and field data collection and analysis. Seven granular base materials used by the SCDOT were included in laboratory plate load and Soil Stiffness Gauge (SSG) tests. In addition, two field test sections were constructed and tested using a Falling Weight Deflectometer (FWD) and SSG. Routine laboratory tests were also performed on the granular materials to determine basic physical properties and compliance with SCDOT specifications. Based on tests results, it is proposed that the maximum percent passing the No. 4 sieve for Macadam be relaxed from the current specification limit of 5 % to 6% (the current SCDOT limit for passing the No. 4 sieve for Marine Limestone). It is also proposed that the SCDOT allow GAB layer thickness greater than 8 in. on a trial basis. Differences in backcalculated layer coefficients for base layers constructed in the laboratory and at field sites were observed in this study. Laboratory test results are in good agreement with results reported by other researchers. It is recommended that the SCDOT consider the feasibility of re-evaluating layer coefficients used for GAB materials. Also included in the study was a preliminary investigation of SSG applicability for assessing compacted GAB materials. Study results suggest that the SSG offers an alternative tool for pavement material quality assurance and construction control. It is suggested that the SCDOT study the SSG further and consider SSG implementation for material characterization in future mechanistic-empirical pavement design approaches. 17. Key Words Flexible Pavement Structure, Base Layer, Granular Material, Gradation, Base Thickness, Resilient Modulus, Plate Loading Test, FWD Test, Soil Stiffness Gauge 19. Security Classif. (of this report) Unclassified 2. Security Classif. (of this page) Unclassified Form DOT F 17.7 (8 72) Reproduction of completed page authorized 18. Distribution Statement No restrictions. This document is available to the public through the National Technical Information Service, Springfield, VA No. of Pages 22. Price i

3 ACKNOWLEDGEMENTS This project was funded by the South Carolina Department of Transportation and the Federal Highway Administration. Their support is greatly appreciated. The authors would like to acknowledge the assistance provided by personnel at the Research and Materials Laboratory of the South Carolina Department of Transportation. Several individuals at the SCDOT provided their time and insights to the project. They include Dr. Andy Johnson, Melissa Campbell, and Mike Lockman. The authors would also like to thank Vulcan Materials Company and Martin Marietta Aggregates for their donations of GAB materials. ii

4 TABLE OF CONTENTS ACKNOWLEDGEMENTS... ii TABLE OF CONTENTS... iii CHAPTER 1 - INTRODUCTION...1 Background Information...1 Project Objectives...2 Scope of Study...2 Research Approach...2 Justification...2 Project Tasks...4 CHAPTER 2 LITERATURE REVIEW...5 SCDOT Practice...5 Gradation and Compaction Requirements (Historical and Current)...5 GAB Layer Coefficient...7 State Agency Survey...8 Summary of Survey Responses...9 Related Technical Information...13 Characterization of Unbound Granular Materials...13 Resilient Modulus...13 Triaxial Test...14 CBR Test...14 Falling Weight Deflectometer (FWD)...15 Other Testing Techniques...16 Factors that Influence Resilient Modulus of Unbound Granular Materials...16 Permanent Deformation Resistance...18 CHAPTER 3 LABORATORY PLATE LOAD TESTING PROGRAM...22 Introduction...22 Laboratory Testing Program...22 Test Pit and Plate Load Apparatus...22 Materials for Laboratory Testing...24 Subgrade...24 Base Materials...25 Placement and Test Procedures...28 Experimental Results...3 Static Plate Loading and CBR Tests on Subgrade...3 GAB Cyclic Plate Load Tests...32 GAB Static Load Tests...37 iii

5 CHAPTER 4 LABORATORY PLATE LOAD TEST RESULTS...42 Cyclic Plate Load Test Results and Analysis...42 Summary...48 Static Plate Load Test Results and Analysis...48 Introduction...48 Finite Element Analysis...49 Linear Programming to Determine Deflection at Optimum Water Content...52 Backcalculation of Resilient Modulus...57 Summary...61 CHAPTER 5 LABORATORY SOIL STIFFNESS GAUGE TESTING PROGRAM AND RESULTS...62 Introduction...62 Experimental Program...62 SSG Laboratory Testing Results...64 Comparison between SSG and Plate Modulus...68 Summary...69 CHAPTER 6 - FIELD SOIL STIFFNESS GAUGE AND FWD TESTING PROGRAM AND RESULTS...72 Introduction...72 US 61 Field Test Section without HMA...73 SC 72 Field Test Section without HMA...76 US 61 and SC 72 Field Test Section with HMA in Place...8 Summary...85 CHAPTER 7 SUMMARY AND CONCLUSIONS...86 Summary...86 Conclusions...86 Influence of Gradation...86 Influence of Base Layer Thickness...87 Recommendations and Future Studies...9 REFERENCES...91 iv

6 CHAPTER 1 INTRODUCTION Background Information For flexible highway pavement construction, the South Carolina Department of Transportation (SCDOT) uses three types of unbound graded aggregate base (GAB) materials (Macadam, Marine Limestone, and Recycled Portland Cement Concrete). Each type of GAB is described in the SCDOT s Standard Specifications for Highway Construction as follows: Macadam base course materials are composed of crushed stone (excluding marine limestone) or slag filled and bound with screenings. The fine aggregate component of Macadam GAB is produced by the crushing operations. Marine Limestone base course materials are produced from crushed limestone from a single source or deposit. The fine aggregate component of Marine Limestone GAB is limestone particles produced by the crushing or mining operations. Recycled Portland Cement Concrete base course materials consist of crushed, graded, recycled portland cement concrete mixed together with sand, sand-gravel, soil or other materials. The coarse aggregate component of Recycled Portland Cement Concrete GAB consists of sound, durable particles of recycled portland cement concrete excluding crushed concrete block or pipe. The fine aggregate component is produced by the concrete crushing operations, or may be sand, soil, or other acceptable fine-grained material. Materials retained and passing the No. 4 sieve are the coarse and fine aggregate components of GAB, respectively. Cost and availability govern the contractor s selection of the GAB type to be used on a construction project. Relative usage of the three types of GAB materials for flexible highway construction is as follows: Macadam GAB is commonly used, Marine Limestone GAB is used for some Coastal Plain projects, and Crushed Recycled Portland Cement Concrete GAB is used infrequently. Macadam and Recycled Portland Cement Concrete share the same SCDOT gradation specifications. Marine Limestone gradation specifications are similar, but allow somewhat fine gradations (specifically, higher percentages passing the ½-inch, No. 4, No. 3, and No. 2 sieves are permitted). It is not known how current SCDOT gradation specifications were established. Current SCDOT flexible pavement design policy limits GAB thickness to a maximum of 8 inches. It is believed that this policy resulted from an investigation of the relative strength of flexible pavement components conducted for the SCDOT at Clemson University (Busching et al., 1971). Current SCDOT Standard Specifications for Highway Construction state that field compaction shall be done with equipment capable of obtaining the required density to the full depth. Compaction shall be done at near optimum moisture until the entire base course is compacted to not less than 1% of maximum laboratory density as determined by AASHTO T 18 (Method D). If the total compacted thickness of the graded aggregate base course is more than 8 inches (a condition not allowed by current SCDOT design 1

7 practice), the Standard Specifications state that the base course should be compacted in two or more layers of approximately equal thickness. Project Objectives This report summarizes a study conducted for the SCDOT to 1) investigate the feasibility of relaxing current GAB gradation specifications and 2) investigate the feasibility of allowing GAB layer thicknesses greater than 8 inches in flexible highway pavement structures. The study included only a limited number of the GAB materials used by the SCDOT and a limited number of laboratory and field tests. Scope of Study Time and resource limitations as well as the timeliness of new highway construction projects suitable for the inclusion of GAB test sections necessitated a study of limited scope. Therefore, the general objectives to investigate the technical and construction issues associated with relaxation of GAB gradation specifications and layer thickness restrictions were limited as described below. The investigation related to relaxing current SCDOT gradation specifications was limited to relaxing the passing No. 4 sieve specification for Macadam. More specifically, relaxing the maximum percent passing the No. 4 sieve from the current limit of 5% to 6% (the current SCDOT limit for passing the No. 4 sieve for Marine Limestone) was investigated. This investigation included full-scale laboratory tests on three commonly used Macadam GABs (tested with the percent passing the No. 4 sieve meeting current SCDOT specifications and with the percent passing the No. 4 sieve increased to near 6%). No Recycled Portland Cement Concrete GAB material testing was conducted to investigate the feasibility of relaxing the passing the No. 4 sieve specification from the current value of 5% to the Marine Limestone value of 6%. The investigation related to relaxing the current SCDOT GAB thickness maximum of 8 inches was limited to the following. Four commonly used GAB materials (two granite GABs, one marble-schist GAB, and one Marine Limestone GAB) were subjected to fullscale laboratory tests with compacted layer thicknesses of 6, 9, and 12 inches. The two granite GABs and one marble-schist GAB are Macadam and were tested in both the meeting and exceeding the maximum percent passing the No. 4 sieve specification condition (as mentioned above). All laboratory GAB layers were compacted in 3 inch lifts on a supporting subgrade material of compacted sand. Two Macadam GAB materials were tested at field test sections with compacted layer thicknesses between 6 and 12 inches. All field test section GAB layers were compacted as a single lift. Research Approach Justification The performance of unbound GAB pavement layers depends on the properties of the aggregates used. NCHRP Report 453 (Saeed et al., 21) states the poor performance of 2

8 unbound GAB layers in flexible pavements may be manifested by fatigue cracking, rutting, and other pavement distresses. Two important aggregate properties cited as contributors to pavement performance are shear strength and stiffness. Other important properties cited are durability (as might be determined by the Magnesium Sulfate Soundness test), and toughness (as might be determined by the Los Angeles abrasion or Micro-Deval test). Frost susceptibility is also cited for cold weather applications. The SCDOT uses AASHTO flexible pavement design methods and Structural Number (SN) to quantify pavement structure. The GAB layer s contribution to SN is the product of GAB layer thickness D 2 (in inches) and layer coefficient a 2 (in 1/inches). The value of layer coefficient quantifies the material quality (influenced by the material s mineralogy, gradation, and other factors that affect mechanical properties) and is a measure of the ability of a unit thickness of the material to function as a structural component of the pavement. Layer coefficient may also be influenced by layer thickness, layer location in the pavement structure, traffic level, and failure criterion (Appendix GG, AASHTO 1986). An accepted way to quantify the quality of GAB as a component in a flexible pavement structure is to compute layer coefficient a 2 as a function of modulus. GAB materials are nonlinear and therefore the value of modulus within a GAB layer will depend on the stress magnitude within the layer (which is influenced by the magnitude of the load, depth and thickness of the GAB layer, and other factors). At the AASHO Road Test, the average value for layer coefficient for untreated granular base course materials was.14. A well-known relationship between a 2 -value for untreated granular base course materials and resilient modulus (M R ) was developed by Rada and Witczak (1981). The relationship is: a 2 =.249 x log 1 M R.977 Using this equation, a GAB resilient modulus value of 3, psi gives the average AASHO Road Test a 2 -value of.14. Higher modulus values give higher values of a 2. Clearly, factors that affect GAB stiffness (including in situ stress level, gradation, layer depth and thickness, etc) affect a 2 (and, as stated above, the layer s ability to function as a structural component of the pavement). The SCDOT currently uses an a 2 -value of.18 for all graded aggregate base course materials (corresponding to M R = 44,3 psi by the Rada and Witczak equation above). The two classic failure mechanisms for flexible pavements are fatigue cracking and rutting. Fatigue cracking is influenced by tensile strains in the hot mix asphalt layer. Strain magnitude is influenced by load magnitude, layer thicknesses, and the stiffness of the underlying layers. Rutting is indicated by permanent deformations that appear in the wheelpaths. One important AASHO Road Test finding was that rutting was due primarily to decrease in thickness of the pavement layers. Results summarized in Huang (24) indicate that about 6% of rutting was due to permanent deformation of base and subbase layers (with the remaining 4% due to surface layer deformation (about 3%) and subgrade deformation (about 1%)). This suggests that a GAB material s ability to resist 3

9 permanent deformations under repeated wheel load repetitions may be important when assessing GAB performance. In this study, the research approach used was to measure GAB stiffness and resistance to permanent deformation to assess the affects of relaxing current SCDOT gradation specifications for Macadam GAB and allowing layer thicknesses greater than 8 inches for Macadam and Marine Limestone GAB. In situ GAB stiffness was measured in full-scale laboratory pit tests (static plate tests and GeoGauge tests) and in field test sections (Falling weight deflectometer and GeoGauge tests). Resistance to permanent deformation was measured by full-scale laboratory pit tests (1, cycle plate tests). Project Tasks The basic research approach for this project included the following tasks: Task 1. Literature review including a state agency survey. Task 2. Full-scale laboratory plate tests on GAB materials. Task 3. Installation and testing of GAB test sections at SCDOT highway construction projects. Task 4. Analysis of test results. 4

10 CHAPTER 2 LITERATURE REVIEW SCDOT Practice Gradation and Compaction Requirements (Historical and Current) A review of SCDOT gradation and compaction requirements for untreated granular base materials revealed the following. The earliest gradation specifications that could be found dated back to Specifications at that time stated that base courses were to be placed in two 3-inch layers and were to consist of hard, durable stone with stone dust and screens spread evenly on the surface. No compacted density requirement was specified. Acceptable density was determined by the opinion of the Engineer. Gradation specifications are listed in Table 2.1 below. Table 2.1 Macadam Base Course Gradation Specifications (1939) Crushed Stone Screen and Dust Min. Max. Passing 1 ½ Screen 1%.. Passing No. 4 Sieve 1% Passing No. 4 Sieve % 4.% SCDOT Macadam base specifications dated 1955 specified that the base course be placed in one uniform layer. The grading requirements for composite mixture of coarse and fine aggregate are listed in Table 2.2. Maximum base course layer thickness and density were not specified. Table 2.2 Macadam Base Course Gradation Specifications (1955) Sieve Designation Percentage by Weight Passing Min. Max. 2 ½ ½ ½ 4 65 No No No No Maximum single lift for base courses were specified in the SCDOT specifications dated For a required base layer thickness of 8 inches or less, the base may be constructed in one layer. Where the required thickness is more than 8 inches, the base was to be constructed in two or more layers of approximate equal thickness and the maximum compacted thickness of any one layer should not exceed 6 inches. A few small changes were made to the grading requirements of Macadam base course material as listed in Table

11 Table 2.3 Macadam Base Course Gradation Specifications (1964) Sieve Designation Percentage by Weight Passing Min. Max. 2 ½ ½ ½ 4 67 No No No No A compacted density requirement was specified in the SCDOT specifications dated The in-place density was required to be not less than 1 percent of maximum laboratory density as determined by AASHTO T 18 (Method D). Maximum single lift was increased from 6 inches to 8 inches. Gradations specifications were altered with the most notable change being substantial increases in the maximum percent passing the 1 and ½ sieves (from 88 to 1%, and 67 to 75%, respectively). The revised gradation specifications are given in Table 2.4 below. Table 2.4 Macadam Base Course Gradation Specifications (1986) Sieve Designation Percentage by Weight Passing Min. Max ½ ½ No No No Liquid Limit: 25 Maximum Plasticity Index: 6 Maximum It should also be mentioned that on page 45 of Busching et al. (1971) there is a reference to an additional SCDOT base course gradation specification that was apparently in effect in This specification (referenced as South Carolina Department of Highways Specification 45B3) is different from the 1964 and 1986 gradation specifications cited above. SCDOT Research and Materials Laboratory personnel were unable to provide any historical record of this specification. Current Macadam base course gradation specifications (2) remain the same as in SCDOT specifications dated As mentioned in Chapter 1, the same gradation specifications also apply for Recycled Portland Cement Concrete GAB. A gradation specification for Marine Limestone GAB was added and is shown in Table

12 Table 2.5 Marine Limestone Base Course Gradation Specifications (2) Sieve Designation Percentage by Weight Passing 2" 1 1 1/2" " 7 1 1/2" 5 85 No No No. 2 2 Liquid Limit 25 Max. Plasticity Index 6 Max. Current (2) compaction specifications (summarized in Chapter 1 of this report) require GAB compaction at near optimum moisture until the entire base course is compacted to not less than 1% of maximum laboratory density as determined by AASHTO T 18 (Method D). If the total compacted thickness of the graded aggregate base course is more than 8 inches (a condition not allowed by current SCDOT design practice), the base course should be compacted in two or more layers of approximately equal thickness. GAB Layer Coefficient The SCDOT uses a layer coefficient (a 2 -value) of.18 to represent the quality of all unbound GAB materials. This value was established as a result of the investigation conducted by Busching et al. (1971). This investigation is also believed to have led to the SCDOT s current design policy of limiting untreated granular base layer thickness to 8 inches. The Busching study involved the construction and testing of flexible pavement test sections. The test sections were constructed within the confines of two concrete test pits. Each test section was approximately 8 ft x 12 ft in plan area. The GAB materials used were South Carolina GAB materials (unbound crushed Granite-gneiss and Fossiliferous Limestone) and a Dolomitic Limestone from the AASHO Road Test. GAB layer thicknesses of 5 and 1 inches were tested (supported by different subbase materials placed either 5 or 15 inches thick). Two subgrade conditions were investigated (modulus of subgrade reaction, k = 5 pci and 275 pci). All test sections had a 3-inch asphaltic concrete surface layer. Surface displacements were produced by a dual tire hydraulic loading system. The loads required to produce.1-inch and.2-inch surface deflections were used for data analysis. Stiffness modulus values were computed using the increments of load per inch of base per inch of deflection (using loads corresponding to deflection increments from to.1 inch and.1 to.2 inch). Base layer coefficients were then computed for assumed base layer thicknesses of 4, 6, 8 and 1 inches. A comparison of selected recommended layer coefficients for base thicknesses of 4, 6, 8, and 1 inches constructed with AASHO Dolomitic Limestone and South Carolina Type 2 Macadam (crushed Granite-gneiss) is given in Table

13 Table 2.6 Recommended Layer Coefficients for GAB Materials (Busching et al., 1971) Pavement Components Base Layer Coefficient for Base Layer Thickness of: AASHO Dolomitic Limestone South Carolina Type 2 Macadam Crushed Granite-gneiss (On weak support) 1 (On firm support) ,2 see Table 28 (page 81) of Busching et al. (1971) for definitions of weak and firm support The findings presented in Table 2.6 show a 2 -values decreased for the 1-inch layer thickness. Not shown in Table 2.6 are computed a 2 -values for unbound Fossiliferous Limestone base (.21 for all base layer thicknesses). It is presumed that the SCDOT s current a 2 -value of.18 for all GAB materials (unbound Macadam, Marine Limestone, and Recycled Portland Cement Concrete) is based on the approximate average of the Crushed Granite-gneiss, Fossiliferous Limestone, and perhaps the AASHO Dolomitic Limestone a 2 -values presented in the Busching study. The results in Table 2.6 are based on applied dual tire load vs measured surface deflection data. Modulus values for base course materials were not determined. State Agency Survey To obtain information about relevant recent research activities and state highway agency (SHA) practice, a survey was conducted. In early 22, a short survey questionnaire (Table 2.7) was sent to all US SHAs. The survey included questions on several topics, including base layer thickness and single-lift thickness requirements, gradation and compaction specifications, and relevant research activities on unbound granular base materials. Twenty-five SHAs responded to the survey. Responding states are shown in Fig Data from the survey and information obtained from the literature, SHA websites, etc. are combined and summarized in this chapter. Table 2.7 Survey Questions 1. For flexible pavement construction, does your agency permit unbound aggregate base course thickness greater than 8 inches? 2. If your agency does permit unbound aggregate base course thickness greater than 8 inches, a) What is the maximum total thickness permitted? b) What is the maximum single-lift thickness permitted? 3. We would greatly appreciate the following (either in paper or electronic form, or a link on your agency s web page to online specifications/information): a) Your agency s unbound aggregate base compaction and gradation specifications, and b) Any research reports or summaries of internal investigations related to unbound aggregate base course thickness for flexible pavement construction. 8

14 WA OR ID MT ND SD MN WI MI MI NY ME VTNH MA CTRI CA NV AZ UT WY NM CO NE KS OK IA MO AR IL IN KY TN OH PA NJ MDDE WV VA NC SC TX LA MS AL GA FL AK HI HI HI HI States Responding (25) States Not Responding (25) HI Fig. 2.1 US SHAs Responding to Survey Summary of Survey Reponses Question 1. For flexible pavement construction, does your agency permit unbound aggregate base course thickness greater than 8 inches? Of the 25 SHAs responding to the survey, 15 SHAs reported unbound aggregate base course thickness greater than 8 inches is permitted (see Fig. 2.2). Kansas indicated unbound aggregate base course is rarely used. Illinois uses unbound GAB for local roads only. Mississippi uses only bituminous base course. 9

15 N/A States Not Responding (25) Greater than 8 in. Less than or Equal to 8 in. Fig. 2.2 SHAs Responses to Question 1 Question 2. a) What is the maximum total thickness permitted? Twenty SHAs responded explicitly to this question. Colorado and Montana reported 24 unbound GAB (UGAB) layers were implemented in projects. Nebraska uses 4 to 5 layer thickness, and Alaska and Wisconsin limited layer thickness to 6. Kansas reported no layer thickness limit for dense graded base with 8% to 2% fine content, and unbound drainable base was limited to 6. Ohio reported using 6 UGAB or 6 UGAB with 4 of permeable base. Fifty percent of the SHAs reported limits greater than 12, 22% of SHAs had a limit greater than 8 and less than 12, 11% of SHAs impose a limit of 8, and 17% of SHAs limited UGAB layer thickness to less than 8. Graphical summaries are provided in Fig. 2.3 and

16 > 8" and <=12", 22% =8", 11% <8", 17% > 12", 5% Fig. 2.3 UGAB Layer Thickness Reported by SHAs States Not Responding (25) Less than or Equal to 8 in. Greater than 8 in. Less than or Equal to 8 in. Greater than 12 in. N/A Fig 2.4. SHAs Responses to Question 2. a) 11

17 Question 2. b) What is the maximum single-lift thickness permitted? Eighteen SHAs gave explicit answers to this question. Illinois reported limiting single-lift thickness to 4, Louisiana has a limit of 12, and New York has a limit of 15. The remaining 15 SHAs reported a single-lift thickness limit of 6 to 8. Question 3. a) Your agency s unbound aggregate base compaction and gradation specifications. Responding SHAs provided limited information on compaction and gradation and specifications. Table 2.8 summarizes compaction specification information from survey results and a review of selected agency web sites. As mentioned, the SCDOT GAB compact specification is 1% of AASHTO T 18 (Method D). Kentucky, Ohio, and Kansas reported using strip test to determine density requirements for granular base materials. Table 2.8 SHA Compaction Standards State Name Compaction Specification Maine 95% AASHTO T 18 C or D, and corrected by Adjustment Chart Illinois 1% AASHTO T 99 and corrected by AASHTO T 224 Washington 95% WSDOT Test Method Alaska 98% AASHTO T 18 D Indiana 1% AASHTO T 99 New 95% AASHTO T 99 Hampshire Utah 97% AASHTO T 18 D +/-2% optimum water content Wisconsin AASHTO T 99 C, (replacement of the fraction) Grading requirements for percent passing the No. 4 and No. 2 sieves for 8 SHAs are compared in Table 2.9. Note that all SHAs have a passing No.4 sieve limit above 5%. The maximum reported limit for material passing the No. 2 sieve is 12%. Table 2.9 Grading Requirements for No. 4 and No. 2 Sieves State Name Percent Passing No. 4 Sieve No. 2 Sieve Low Limit High Limit Low Limit High Limit Delaware 2 5 N/A N/A Washington 25 N/A N/A N/A Florida Tennessee N/A N/A Alaska Nebraska N/A 93 N/A 3 New Hampshire South Carolina

18 Question 3. b) Research reports or summaries of internal investigations related to unbound aggregate base course thickness for flexible pavement construction. Little information was obtained from the survey responses about research activities directly related to unbound aggregate base course thickness. Reported investigations included the following: Kansas reported investigating aggregate base course drainability; Iowa reported a study to establish layer coefficients for some local materials; Georgia reported construction of a test section that includes 12-inch UGAB lifts. In 1996, a study (Kimley-Horn and Associates, Inc., 1996) was conducted to evaluate the feasibility of compacting unbound GAB lifts thicker than permitted by the North Carolina DOT. Study test results demonstrated that 1 and 12 in. lifts of GAB were successfully placed and compacted, and the required gradation, density, and water content were maintained throughout the entire depth of the compact lift. A similar investigation was conducted by ICAR researchers (John, et al., 1998). Five full-scale test sections were constructed using a variety of material types and single lift thicknesses ranging from 12 to 21. Density and shear wave velocity were measured using nuclear density gauge and spectral analysis of surface waves techniques. Their findings indicated that density in excess of 1 percent of maximum as determined by AASHTO T 18 can be achieved for thicknesses of up to 21 using standard compact equipment. Related Technical Information Characterization of Unbound Granular Materials Resilient Modulus Resilient modulus, M R, is the fundamental material property used to characterize the quality of subgrade and UGAB materials. Resilient modulus is the elastic modulus based on the recoverable (resilient) strain under repeated loads. M R is expressed as: σ d M R = ε r where σ d = the applied deviator stress; σ d = σ 1 σ 3 ; ε r = resilient strain. M R was introduced in the 1986 AASHTO flexible pavement design procedures and is a key input property for the mechanistic-empirical pavement design procedures proposed in NCHRP Project 1-37A (22). For older AASHTO flexible pavement design procedures, resilient modulus of granular base material can be correlated to layer coefficient a 2. Appendix L of AASHTO Guide for Design of Pavement Structures (1993) lists four approaches for determining design resilient modulus. The four approaches are 1) laboratory testing (repeated load triaxial testing), 2) estimation by correlation to other 13

19 test results or physical properties, 3) nondestructive testing (NDT) of in situ materials, and 4) determination from original design and construction data. Triaxial Test Laboratory repeated load triaxial tests have been widely used to determine resilient modulus and permanent deformation characteristics of unbound granular materials. In this test, cylindrical specimens are subjected to a series of load pulses applied with a distinct rest period, simulating the stresses caused by multiple wheels moving over the pavement. A constant all-around confining pressure applied on the specimen simulates the in situ lateral stresses. The total recoverable axial deformation response of the specimen caused by the stress pulses is used to calculate resilient modulus. There have been several triaxial test methods presented by AASHTO, ASTM, and the SHRP LTPP program for measuring M R of soils and unbound granular materials. NCHRP Project 1-28, Laboratory Determination of Resilient Modulus for Flexible Pavement Design, completed in 1997, produced yet another set of testing procedures for M R determination of unbound materials. In order to harmonize these testing methods, a recommended method was developed in NCHRP Project 1-28A Harmonized Test Methods for Laboratory Determination of Resilient Modulus for Flexible Pavement Design. This test protocol is reported to reduce testing variability and the time required to complete testing. Despite improvements made over the years, there are uncertainties as well as limitations associated with triaxial testing. A study by Ke et al. (2) showed that reproducing the in situ internal structure of granular materials with current laboratory specimen preparation techniques is not possible because of sample disturbance and differences in aggregate orientation, moisture content, and level of compaction. Karasahin et al. (1993) demonstrated stress conditions in a triaxial cell are different from those in pavement structures due to inherent equipment flaws. The minor principle stress in a triaxial cell is kept constant during testing while both major and minor principle stresses are cycled under wheel loadings. In addition, inherent instrumentation flaws create uncertainty in the measurement of sample deformation. Darter et al. (1995) reported that because of the complexity of repeated load triaxial tests, most SHAs do not routinely measure resilient modulus using triaxial testing but rather estimate resilient modulus from experience or by correlation equations developed from physical properties or CBR test results. CBR Test The California Bearing Ratio (CBR) test was introduced in 1929 by Jim Porter working as Soils Engineer for the state of California (Brown S. F., 1997). This test is used to quantify the quality of compacted soil, soil-aggregate combinations, and aggregate base materials using a numerical value of CBR. In this well-known test (AASHTO Test Method T 193), a CBR value is computed from piston force and piston penetration measurements. There are a number of empirical equations to correlate CBR value to resilient modulus of unbound base or subgrade materials. Heukelom and Klomp (1962) 14

20 proposed a well-known correlation using dynamic compaction and in situ resilient modulus and CBR values. The correlation can be expressed as: M R (psi) = 15 CBR The coefficient 15 can vary from 75 to 3. Available data indicate that the Heukelom and Klomp equation provides acceptable M R predictions for fine-grained soils and fine sands with CBR values less than about 2 (Huang, 24). Materials with CBR values higher than 25 often have their M values overestimated. Another correlation for subgrade soil was proposed by Powell et al. (1984). This correlation was primarily developed from data relating modulus measured by wave propagation to in situ CBR results. The correlation is: M R (psi) = 255 CBR.64 The CBR test measures penetration resistance and thus provides an indirect measurement of undrained shear strength. Pavement materials generally function at stress levels within the elastic range. The CBR test provides no information about material resilience. Brown et al. (1987) states that resilient modulus is not a simple function of CBR. But, nevertheless, partially due to its simplicity and well-acceptance among pavement engineers, the CBR test is widely used to evaluate the strength of paving materials, and to correlate to resilient modulus. Falling Weight Deflectometer (FWD) The FWD has become a popular device for nondestructive measurement of loaddisplacement behavior of constructed pavement structures or pavement layers. The device may be van-integrated, mounted on a trailer, or hand-portable. The FWD test involves applying a dynamic load on the pavement surface through a circular metal plate. By varying the drop height and weight, a range of peak impact forces can be produced to simulate actual traffic loads. Sensors are used to measure load-deflection history at the center of the plate and deflection history at several radial distances from the plate. The measured deflected shape of the pavement surface under peak impact load is called the deflection basin. Pavement surface temperature can also be measured using an infrared temperature sensor. Measured load-deflection information is often used in back-analysis procedures for the purpose of computing in situ moduli of the various pavement layers. There are two basic types of backcalculation models. One involves numerous iterations of a linear or nonlinear elastic analysis program. The other involves matching the measured deflection basin to a number of previously calculated deflection basins. A static pavement response model is usually used in the backcalculation procedure without considering the inertial effects. This simplification was investigated by Tam and Brown (1989), whose work indicated the inertial effects were generally insignificant and static modeling can provide reasonable solutions. Different testing and analysis procedures may produce different results. Research conducted by Rauhut and Jordahl (1992) showed that the coefficient of variation R 15

21 (COV) of backcalculated modulus values ranged from 13% to 67% for four Strategic Highway Research Program sections. Collop et al. (21) performed statistical analysis of in situ pavement moduli backcalculated from FWD tests. The study indicated that although backcalculated mean modulus values of the asphalt layer were determined to an accuracy of ± 25% and a confidence level of 95% with a relatively small sample size, for base layers a rather larger number of samples was required to achieve the same accuracy and confidence level. Johnson and Baus (1993) investigated a number of basin-matching backcalculation programs. They indicated those programs tend to underestimate the modulus of the unbound base course and overestimate the modulus of the asphalt concrete-bound top layer and the subgrade. Other Testing Techniques Wave propagation techniques have also been used for the determination of in situ modulus of subgrade and pavement materials. Dynamic Cone Penetrometer (DCP) testing can be used to determine in situ layer thickness and penetration resistance. Penetration resistance can be correlated to in situ modulus. Plate load tests have been used on flexible pavement components for design or evaluation purposes. The loading plate is also used in the Falling Weight Deflectometer (FWD) testing to simulate dynamic wheel load on highway pavements. Advantages of plate tests include 1) the magnitude of load and the state of stress caused by the load reasonably approximate those in highway pavement structures, and 2) tested materials are either in situ or can be constructed in a way that approximates the in situ internal structure of pavement materials. Konrad and Lachance (21) performed DCP and plate load tests to evaluate base and subbase materials. The results indicated a good correlation can be found between modulus inferred from the plate load tests and the penetration tests. In this project, cyclic and static plate load tests were conducted to access permanent deformation resistance and resilient modulus of the unbound granular base materials. A relatively new device called the Soil Stiffness Gauge (SSG, a.k.a. GeoGauge) measures material stiffness (modulus of subgrade reaction or stress-strain modulus) directly using steady-state vibrations. In this project, engineering properties of unbound granular base materials were evaluated using the SSG. Factors that Influence Resilient Modulus of Unbound Granular Materials The resilient modulus of granular materials is not a constant stiffness property but depends upon various factors including stress state, water content, dry density, and gradation. (Seed, 1967; Thompson, 1969; Hicks and Monismith, 1971; etc.). Lekarp et al. (2) illustrated that the effect of stress level on the resilient behavior is the most significant factor. Stress level of GAB materials in base courses primarily depends on base layer location within the pavement structure, layer thickness, and wheel load magnitude. Granular materials are known to exhibit nonlinear behaviors under traffic load. Uzan (1992) introduced a universal model, applicable to all types of unbound paving materials ranging from very plastic clays to clean granular bases. This model was 16

22 used in the NCHRP Project 1-28A on the development of a harmonized M R test protocol. This model, also known as K1 K 3 model, can be represented as: θ τ = + MR K Pa P a P a where ( ) K ( oct 1) K M R = resilient modulus; P a = atmospheric pressure to normalize stresses and modulus; K 1, K 2, K 3 = regression constants, dependent on material type and physical properties and are obtained from regression analysis; θ = stress invariant, or the sum of the three principle stresses; τ oct = octahedral shear stress, which can be expressed as: 1 τoct = (( σ σ ) + ( σ σ ) + ( σ σ ) ) / Octahedral shear stress is equal to deviator stress when the stress condition of granular materials in a triaxial test or under real traffic loads has an axis of symmetry. A study by Lekarp et al. (2) demonstrated that M R of granular materials increases with increasing confining stress and sum of principal stresses. The influence of octahedral stress is minimal. For granular materials, it is known that M R increases with a decrease in moisture content and an increase in density. Hicks and Monismith (1971) used triaxial tests to evaluate the influence of water content, dry density, and confining stress. Their findings indicated a steady decrease of M R with increasing degree of saturation up to optimum water content and decreasing dry density. Numerous other investigations confirm these findings. A study by Hicks and Monismith (1971) showed the influence of gradation was not well defined for two types of aggregate materials. For one material, the resilient modulus increases as the percentage passing No. 2 sieve increased, while for another material, the opposite trend was observed. Shaw (198) studied the effect of aggregate grading using triaxial test results. A comparison was made between 4-mm maximum size broadly graded granular material and a 3-mm single-sized stone from the same source. The broadly graded material was found to be stiffer than the singlesize stone. Thom (1988) conducted a series of repeated load tests on 1-mm maximum sized crushed dolomitic limestone and found high stiffnesses for uniformly graded materials, but broadly graded materials showed higher shear strengths. Kamal et al. (1993) compared the mechanical behavior of six gradings of unbound granular materials. Testing results showed the effect of the percentage passing the No. 4 sieve on resilient modulus and permanent strain was not defined. Santha (1994) studied the effects of the physical properties of 15 granular materials on the resilient modulus. It was found that M R decreased slightly with increasing percent passing the No.4 sieve. Rahim and George (24) investigated the relevance of soil index properties in resilient modulus for Mississippi soils. A result of this study was a proposed 17

23 correlation equation that indicates M R decreases with increasing percent passing the No. 2 sieve. A review of the literature suggests that the percent passing the No. 4 and No. 2 sieves may be important influencing factors for well-graded granular materials, but their influence on resilient modulus is not well defined and likely material-specific. Permanent Deformation Resistance Permanent deformation characteristics of unbound granular materials are important for flexible pavement performance. AASHO Road Test findings indicated a major part of rutting occurred inside the base and subbase layers. Another finding was that rutting in the wheel path was primarily caused by lateral movements of pavement materials instead of material densification. A study by Thompson and Smith (199), confirmed by other researchers, showed that permanent deformation under repeated load application may provide a more definite evaluation of pavement materials than resilient modulus in some cases. Repeated load tests, typically as an extension of the resilient modulus tests, have often been used to determine permanent deformation of pavement materials. The tests usually involve applying loading applications of up to 1, repetitions and recording permanent deformation at a number of designated cycles. Compared to resilient behavior characterization, fewer models have been developed to describe permanent deformation of pavement materials. Gidel and Horny (21) suggest the following reasons: 1. Permanent deformation tests are expensive and time consuming. Permanent deformation is strongly dependent on stress history. Only one stress level is generally applied on each specimen per test; a large number of tests (at least ten each involving a large number of cycles) is therefore necessary to investigate how stress levels affect permanent deformation. 2. It is difficult to predict field rut depth from laboratory test results. In flexible pavements the material has a very complex loading history (initial phase of pavement construction, highly varied traffic loading, variations in climatic conditions) which is extremely difficult to simulate under laboratory testing conditions. Barksdale (1972) conducted repeated load triaxial tests with an average of 1, load applications on different granular materials. The load ramped up to the peak value and then back to the trough value in a period of.1 second with a 1.9 second period of no load separating the load applications. It was suggested that a reasonable range of load repetition is from 1, to 1,,. From this work a qualitative rutting index was defined to evaluate pavement performance. A relationship between deformation and load applications was suggested as: p ε 1 ( N) = a + blog( N) where 18

24 p N = number of load cycles; ε 1 ( N) = permanent axial strain at Nth number of cycles; a and b are regression parameters. The Barksdale study showed that permanent deformations were highly dependent on the applied load and increased when confining pressure decreased and deviator stress increased. The effect of density on the permanent deformation was also investigated. An increase of permanent axial strain of about 185% was observed when the material was compacted at 95 % instead of 1 % of maximum compaction density. (AASHTO T 99). Diyaljee and Raymond (1982) developed an equation using regression methods to predict the permanent deformation under long term repeated loading using static stress-strain data and a minimum number of cycles of repetitive load test data. Based on their results on Conteau Dolomite railroad ballast data from other researchers, an equation for cohesionless materials was proposed: ε p 1 ( where N ) = Be ex N m B = value of strain at X = for the first cycle; X = the ratio of the repeated deviator stress to the failure deviator stress under static loading; n, m = regression parameter. An example expression for subgrade sand with 35 kpa of confining pressure would be p ε 1 =.4e N 4.7 x.12 A study by Sweere (199) on granular materials showed that a log-log approach is appropriate for a large number of cycles: log( ε p ( N)) = a+ blog( N) 1 Sweere s model is essentially the same as the power model proposed by Monismith in 1975 (Monismith et al., 1975) though the later was based on a silty clay with a LL = 35 and PI = 15. The Monismith model can be expressed as: ε = an p 1 b One major finding of the Monismith study was that the exponent b depends only on soil type. The tested soils had a b parameter between.154 to.332 and an a parameter between.467 and The effect of factors such as applied stress history and moisture content are included into parameter a. An asymptotic model proposed by Hornych and Paute (1993) for unbound granular materials is: 19

25 ε N = A[1 ( ) ] + ε (1) 1 p B p 1 1 This model assumes that permanent strain approaches a finite limit as N tends towards infinity. Other models relate permanent deformation to applied stresses. Representatives are those by Hyde (1974) and Lekarp et al. (1998). A model proposed by Hyde (1974) can be expressed as: ε p 1 q = a σ 3 Lekarp et al. (1998) used the repeated load triaxial equipment to test five different granular materials. A model that relates permanent axial strain to stress path length and stress level was proposed as: p ε 1( Nref ) q = a( ) L/ P p where b max p ε 1( Nref ) is the accumulated permanent axial strain at a given number of cycles; L is the length of the stress path in kpa; a and b are regression parameters; P is a reference stress and is equal to 1 kpa. Lekarp showed that the accumulation rate of permanent strain would eventually reach zero, if the stress ratio was low. However, at high stress ratios the accumulation of permanent strain was more progressive, indicating that a threshold stress ratio must exist above which accumulation of permanent strain will cause failure. This threshold stress ratio is called the shakedown limit. Beside the number of load repetitions and stress level, the influence of gradation on permanent deformation may be important. Kamal et al. (1993) conducted laboratory and full-scale tests on unbound granular materials with 8 different gradations. The percent passing the No. 4 sieve varied from approximately 13% to 6%. The results indicated the rut depth was more than twice as great for the open-graded materials as for the well-graded materials. A study by Barksdale et al. (1997) shows permanent deformation of crushed granite gneiss with 16% fines content was more than twice that for 1% fines content (see Fig. 2.5). Similar results were found by Belt and Ryynanen (1997). The Belt and Ryynanen study also showed that open-graded or well-graded unbound granular materials do not necessarily behave as well in real pavement structures as would be expected based on repeated load triaxial test results. This is said to be because the rotation of principal stress directions during real traffic 2

26 loading conditions, which significantly influences the permanent deformation behavior of unbound granular materials. Fig. 2.5 Influence of Number of Load Repetitions and Material Quality on Permanent Deformation (Barksdale et al. (1997)) 21

27 CHAPTER 3 LABORATORY PLATE LOAD TESTING PROGRAM Introduction Seven GAB materials were tested in the laboratory with compacted layer thickness of 6, 9, and 12 inches. Table 3.1 lists the material types, sources, and abbreviations used in this report. Crushed Granite (CGr), Crushed Marble-schist (CMs), and Crushed Limestone (CL) are commonly used in South Carolina to construct base courses. Modified crushed granite and marble-schist are those with additional material passing the No. 4 sieve. Table 3.1 Materials Tested in the Laboratory Material Type Source Abbreviation Crushed Granite Vulcan Materials Company, Columbia Quarry CGr A Modified Crushed Granite Vulcan Materials Company, Columbia Quarry MCGr A Crushed Marine Limestone Martin Marietta Aggregates, Berkeley Quarry CL Crushed Granite Martin Marietta Aggregates, Jefferson Quarry CGr B Modified Crushed Granite Martin Marietta Aggregates, Jefferson Quarry MCGr B Crushed Marble-schist Modified Marble-schist Vulcan Materials Company, Blacksburg Quarry CMs Vulcan Materials Company, Blacksburg Quarry MCMs Throughout the testing program, test conditions were designed to simulate in-service conditions of stress levels, moisture, and density. Sample preparation followed, as closely as possible, standard methods and procedures of sample splitting, handling, compaction, and moisture control. Laboratory testing apparatus and equipment were carefully maintained and calibrated in accordance with manufacturers manuals. The FWD is maintained by SCDOT technicians. Laboratory Testing Program Test Pit and Plate Load Apparatus To provide control over material gradations, water content, and construction, full-scale laboratory pavement models were constructed in a concrete test pit housed in the Department of Civil and Environmental Engineering laboratory facility at the University of South Carolina. 22

28 A schematic plan view of the laboratory test pit is shown in Fig Test apparatus and cross-section of the test pit are illustrated in Fig The test pit was 13 feet by 1 feet in plan. Plate tests were performed with a MTS Axial-Torsion Test System capable of applying an axial load of 5, pounds at a frequency range of to 1 Hz. Plate deflections were measured by three linear variable differential transformers (LVDTs).The LVDTs were mounted on two reference beams and were placed equally apart on a 17.8 in. diameter metal plate. The metal plate had the same diameter as the large plate used for FWD testing. Load was measured by a load cell aligned collinearly with a hydraulic actuator. Deflection and load data were collected to a computer through a high speed digital/analog data acquisition system. An initial attempt had been made to simulate flexible plate conditions by gluing a rubber pad from a Dynatest FWD to the bottom of the metal plate. Initial experiments using the rubber pad showed that deformations of the rubber pad itself dominated total deflection. Therefore the pad was removed and a rigid plate was used. To insure proper contact between the rigid metal plate and the base layer, a thin hydrostone membrane was applied to the contacting area. Fig. 3.1 Schematic Plan View of the Laboratory Test Pit 23

29 Fig. 3.2 Cross Section of the Laboratory Test Pit and Plate Test Apparatus Materials for Laboratory Testing Subgrade The subgrade material was an in-place sand for general geotechnical testing purposes. The thickness of the sand subgrade layer is approximately 1 feet. The sand is underlain by a permeable gravel deposit. The gradation curve for the sandy subgrade material is shown in Fig According to the ASTM soil classification it is medium sand. 24

30 1 9 8 Subgrade Percent Passing Particle Size mm-log Fig. 3.3 Subgrade Sand Gradation Curve Base Materials All of the base (GAB) materials tested in the laboratory were fabricated by the quarries identified in Table 3.1. The materials were shipped then stored in containers at the testing facility. Sieve analysis and moisture-density tests were performed on each GAB material. Gradation results are shown in Table 3.2 and Fig Table 3.2 Gradations of the Macadam and Limestone GAB Materials Sieve Percentage by Weight Passing Designation CGr A MCGr A CL CGr B MCGr B CMs MCMs 2" /2" " /2" No No No

31 1 Percent Passing CGr A MCGr A CL CGr B MCGr B CMs MCMs * Density (pcf) ** Optimum water content (%) Particle Size mm-log Fig. 3.4 Gradation Curves for GAB Materials Moisture-density tests were performed in accordance with AASHTO T 18 (Method D). Results are shown in Fig. 3.5 and Table 3.3. Fig. 3.5 shows that dry densities are not particularly sensitive to changes in compaction water content. Table 3.3 shows that the difference in maximum dry densities for modified and unmodified materials is not remarkable, especially for crushed granite. 26

32 CGr A MCGr A 15 CL CGr B Dry Density (pcf) MCGr B CMs MCMs Water Content (%) Fig. 3.5 Dry Density vs. Water Content for GAB Materials (AASHTO T 18 D) Table 3.3 Optimum Moisture Content and Maximum Density Material Optimum Water Content (%) Maximum Dry Density (pcf) AASHTO T 18 (Method D) CGr A MCGr A CL CGr B MCGr B CMs MCMs

33 Placement and Test Procedures An electric Wacker compactor was used to compact the in-place subgrade. In situ CBR and static plate loading tests were performed to determine its elasticity and strength characteristics. Prior to placement of the granular material, the subgrade surface was leveled. GAB materials were transported from the quarry to the testing facility by SCDOT maintenance personnel. The base materials were mixed with water as necessary to achieve optimum moisture then covered with plastic sheeting for short-term storage prior to placement. Just prior to placement, conventional oven drying moisture content tests were used to confirm water contents within 1% of optimum. A GAB material to be tested was spread and shaped in the test pit. Compaction commenced immediately and continued without interruption until the desired level of density was achieved. Compaction was performed using the electric Wacker in 3. in. lifts. Effort had been made to ensure a leveled base surface for each lift. The thickness of each compacted layer was carefully controlled to be within 1/4 in. tolerance. Where the base course was deficient by more than ¼ in., such areas were scarified and re-compacted with base material added. For each material, base layer thicknesses of 6. in., 9 in., and 12 in. were constructed and tested. After compaction, a hydrostone mixture was applied to the contacting area between the plate and the base layer. The plate was lowered on to the hydrostone and a 1 psi pressure was applied to form a thin hydrostone membrane between the GAB and the loading plate. Curing of the membrane took approximately 3 hours. Meanwhile, the base was covered with plastic sheeting to allow even distribution of moisture (curing) through the base depth. Just prior to testing, the MTS and data acquisition systems were turned on, warmed-up, and confirmed to be fully operational. The LVDTs were inspected for wear and verticality with the metal loading plate. Elliott et al. (1998) summarized repeated loading test configurations for various permanent deformation studies (see Table 3.4). Table 3.4 shows load frequency variations from 2 to 12 repetitions per minute. Previous research (Seed and Fead, 1959; Barksdale, 1972) indicated the effect of load pulse shape has little effect on the resilient modulus measurements. Barksdale et al. (1997) suggested the haversine load pulse as the likely best approximation of traffic loading of base materials. For this study, a haversine wave form was used for cyclic loading tests and a triangle wave form was used for static loading tests. Load durations of 1 second for cyclic loading tests and 1 seconds for static loading tests were used. 28

34 Table 3.4 Summary of Test Configurations for Various Permanent Deformation Studies (after Elliott et al. (1998)) Load Frequency (repetitions per minute) Load Duration (seconds) Seed and Fead (1959) Larew and Leonards (1962) Barksdale (1972) Monismith et al. (1975) Poulsen et al. (1979) Lentz (1979) Raad and Zeid (199) Behzadi and Yandell (1996) Elliott et al. (1998) Rest Period (seconds) No. of Applications (in thousands) 1 6 to or

35 After the curing, 1, cycles of haversine load with frequency of 1 Hz was applied to the granular base layer. The cyclic loading test took approximately 28 hours. During the cyclic loading test, the base course remained covered with the plastic sheeting to maintain the moisture of the GAB material. Subsequently, the plastic sheeting was removed and 3 repetitions of static loading tests were performed approximately every 2 days as the water content decreased due to evaporation. Load and deflection data were recorded for the third load application. Maximum plate pressure on the GAB material was 5 psi. (Preliminary plate tests directly on the sandy subgrade used a maximum plate pressure of 2 psi). These maximum pressures are assumed to simulate typical stresses within typical flexible pavement structures. A complete cycle of the tests on one GAB material required about 2 months, including preliminary laboratory moisture-density and sieving analysis. Prior to each cyclic loading and static loading test, SSG tests were performed on the GAB material. SSG testing is discussed later in Chapter 5. At the completion of testing for each GAB material, nuclear gauge tests were performed by SCDOT personnel. Comparisons of measured density achieved in the test pit and maximum laboratory density (as determined by AASHTO T 18 (Method D)) are given in Table 3.5. All base materials achieved above 1% RC except the Crushed Limestone (CL), which had an RC of 97%. This might be due to Wacker malfunction when compaction of the 12 in. crush limestone base layer was performed. Table 3.5 data confirms that compacted density very near maximum laboratory density was achieved in the testing pit using the Wacker compactor and 3. inch lifts. Table 3.5 Comparison of GAB Test Pit Density and Maximum Laboratory Density (AASHTO T 18 (Method D)) Base Material CGr A MCGr A CL CGr B MCGr B CMs MCMs AASHTO T 18D (pcf) Nuclear Gauge (pcf) Relative Compaction 1% 11% 97% 12% 14% 11% 1% Experimental Results Static Plate Loading and CBR Tests on Subgrade Preliminary plate tests conducted on the sandy subgrade showed linear pressure-deflection behavior for plate pressures up to 2 psi (see sample data shown in Fig. 3.6a). In situ CBR tests were performed on the compacted sand anticipating that CBR values could be correlated with resilient modulus. CBR test results are shown in Fig. 3.6b. Nine CBR tests were conducted with one test at the center of the load plate and eight other tests positioned in a uniform circular pattern 25. in. from the center test. The result from one test was disregarded due to testing error and thus the results from eight tests are shown in Fig. 3.6b. 3

36 Plate Pressure (psi) Subgrade Vertical Deflection (in.) Fig. 3.6a Plate Load Test Results for the Sandy Subgrade #1_sand*23.xls 35 3 Resistance (psi) Penetration (in.) Fig. 3.6b CBR Test Results for the Sandy Subgrade 31

37 GAB Cyclic Plate Load Tests Plastic behavior of the granular materials in the plate loading test was observed by applying a repeated plate load to the surface of the compacted GAB material (maximum plate pressure = 5 psi) and measuring the accumulation of nonrecoverable deformation versus the number of load cycles. To assure proper seating of the loading plate, the first five cycles of loading were regarded pre-test seating cycles. Deflection and plate pressure data were recorded at a frequency of 1 Hz throughout the test until 1, load repetitions were applied. Fig. 3.7 shows typical results obtained from the cyclic plate load tests. Resilient deflections (i.e., recoverable deformations) were calculated by subtracting permanent deformation from total deflection measured at 5 psi. Typical results are shown in Fig Except for the CMs GAB, no significant change in resilient deflection was observed as the number of cycles increased. This is especially the case after 1 cycles. As resilient deflections are inversely proportional to GAB densities, the results appear to be in good agreement with AASHO Road Test data, which showed that permanent deformation is caused primarily by lateral movements of the materials instead of material densification. Due to fast data acquisition rate during a cyclic test, only a few hundred cycles could be recorded at a time. This required the data collection be performed manually at regular intervals during the 28 hours testing period. The test usually began in the late afternoon after GAB compaction and apparatus setup, which meant overnight data collection. During the testing program, two unavoidable schedule changes caused unsuccessful data collection (specifically the 6 in. MCGr A and the 12 in. CGr B base tests). 32

38 1-2 Cyclic Loading Test Results for 6 in. CGr A Base Resilient deflection.1 Vertical Deflection (in.).8.6 Permanent deformation Total deflection Number of Cycles 1-2 Cyclic Loading Test Results for 9 in. CGr A Base.12 Vertical Deflection (in.) Resilient deflection Permanent deformation Total deflection Number of Cycles 33

39 1-2 Cyclic Loading Test Results for 12 in. CGr A Base Resilient deflection Vertical Deflection (in.) Permanent deformation Total deflection Number of Cycles Fig. 3.7 Three Typical Cyclic Plate Load Test Results (CGr A).35 Resilient Deflection (in.) in. CGr A 9 in. CGr A 12 in. CGr A Log(Number of Cycles) 34

40 .35 Resilient Deflection (in.) in. MCGr A 12 in. MCGr A Log(Number of Cycles).35.3 Resilient Deflection (in.) in. CGr B 9 in. CGr B Log(Number of Cycles) 35

41 .35 Resilient Deflection (in.) in. MCGr B 9 in. MCGr B 12 in. MCGr B Log(Number of Cycles) Resilient Deflection (in.) in. CMs 9 in. CMs.1 12 in. CMs Log(Number of Cycles).3 Resilient Deflection (in.) in. MCMs 9 in. MCMs 12 in. MCMs Log(Number of Cycles) Fig. 3.8 Resilient Deflection vs Load Repetitions 36

42 GAB Static Load Tests At the completion of the cyclic loading test, three cycles of static loading tests were performed. The first two load cycles were considered plate seating. Vertical deflection and plate pressure data were recorded for the third static load repetition. Static deflection curves for the seven GAB materials are shown in Fig Due to a power surge that adversely affected data collection, the deflection data were abandoned for one test (6 in. CGr B). Increasing layer thickness helped decrease plate deflections. The deflection curves indicate that the granular base materials exhibited stress-hardening properties. The influence of water content is pronounced. Fig. 3.1 shows the sensitivity of plate deflection to water content. For CGr and MCGr materials, a 1% drop in water content results in approximate.1 in. decrease in plate deflection at plate pressure of 5 psi..5 Plate Pressure (psi) Decreasing water content.5 Plate Pressure (psi) Decreasing water content Vertical Deflection (in " CGr A W=3.8% 6" CGr A W=3.4% 6" CGr A W=2.7% Vertical Deflection (in " MCGr A W=4.5% 6" MCGr A W=2.9% 6" MCGr A W=1.8%.5 Plate Pressure (psi) Decreasing water content.5 Plate Pressure (psi) Decreasing water content Vertical Deflection (in Vertical Deflection (in " CGr A W=4.4% 9" CGr A W=1.8% 9" CGr A W=1.2% " MCGr A W=3.4% 9" MCGr A W=2.% 9" MCGr A W=.9% 37

43 .5 Plate Pressure (psi) Decreasing water content.5 Plate Pressure (psi) Decreasing water content Vertical Deflection (in Vertical Deflection (in " CGr A W=4.% 12" CGr A W=2.%.25 12" MCGr A W=4.% 12" MCGr A W=2.8%.3 12" CGr A W=1.2%.3 12" MCGr A W=2.% Plate Pressure (psi) Plate Pressure (psi) Decreasing water content.5 Decreasing water content Vertical Deflection (in " CL W=11.5% 6" CL W=9.3% 6" CL W=8.1% 6" CL W=6.% Vertical Deflection (in " CL W=1.7% 9" CL W=9.9% 9" CL W=8.% Plate Pressure (psi) Plate Pressure (psi) Decreasing water content.5 Decreasing water content Vertical Deflection (in " CL W=8.9% 12" CL W=5.4% 12" CL W=4.% Vertical Deflection (in " MCGr B W=6.2% 6" MCGr B W=4.6% 6" MCGr B W=2.7% 38

44 Plate Pressure (psi) Plate Pressure (psi) Decreasing water content.5 Decreasing water content Vertical Deflection (in " CGr B W=2.% 9" CGr B W=.8% 9" CGr B W=.7% Vertical Deflection (in " MCGr B W=6.% 9" MCGr B W=3.% 9" MCGr B W=1.5% Plate Pressure (psi) Plate Pressure (psi) Decreasing water content.5 Decreasing water content Vertical Deflection (in Vertical Deflection (in " CGr B W=3.5% 12" CGr B W=3.%.25 12" MCGr B W=4.5% 12" MCGr B W=3.2%.3 12" CGr B W=1.3%.3 12" MCGr B W=1.2% Plate Pressure (psi) Plate Pressure (psi) Decreasing water content.5 Decreasing water content Vertical Deflection (in Vertical Deflection (in " CMs W=4.1% 6" CMs W=3.4% 6" CMs W=2.3% " MCMs W=4.% 6" MCMs W=1.8% 6" MCMs W=1.5% 39

45 Vertical Deflection (in Plate Pressure (psi) Decreasing water content Vertical Deflection (in Plate Pressure (psi) Decreasing water content " CMs W=3.% 9" CMs W=1.6% 9" CMs W=1.5% " MCMs W=4.% 9" MCMs W=1.7% 9" MCMs W=1.6% Vertical Deflection (in Plate Pressure (psi) Decreasing water content Vertical Deflection (in Plate Pressure (psi) Decreasing water content.25 12" CMs W=3.4% 12" CMs W=1.5%.2 12" MCMs W=3.8% 12" MCMs W=2.9%.3 12" CMs W=1.1%.25 12" MCMs W=2.% Fig. 3.9 Static Loading Test Results 4

46 Water Content.% 1.% 2.% 3.% 4.% 5.% 6.%.15.2 Vertical Deflection (in.) MCGr A TH*=6in. MCGr A TH=9 in. MCGr A TH=12 in. CGr A TH=6 in. CGr A TH=9 in. CGr A TH=12 in. * TH: Layer Fig. 3.1 Static Deflection vs Water Content for MCGr A and CGr A (Plate Pressure = 5 psi) 41

47 CHAPTER 4 LABORATORY PLATE LOAD TEST RESULTS Cyclic Plate Load Test Results and Analysis From cyclic plate load tests, permanent deformations were obtained by subtracting the resilient deflections from total deformations. Permanent deformations were plotted against the number of cycles. Log-log plots were found to describe approximately linear relations between permanent deformation and number of cycles. The permanent deformation at the1 th cycle was introduced into the log-log model to help data interpretations using the following model: ε1 log( ) = 2b + blog( N) p ε (1) where p 1 1 p ε (1) is permanent strain at the 1 th cycle; strain ratio; b is regression parameter. ε p ε1 p 1 (1) is defined as the permanent Permanent strain is calculated by dividing the measured permanent deformation by the base thickness. Log-log plots of the permanent strain ratio against the number of cycles are shown in Fig Water content values and regression lines are included as well. It should be noted that although efforts were made to control base layer water content during each cyclic loading test, some variations were not avoidable. Fig. 4.1 data suggest that there is no significant difference between parameter b (slope of the trend line) for 6 in., 9 in., and 12 in. base layer thickness for any of the GAB materials tested. The results are in good agreement with work by Monismith et al. (1975) mentioned in Chapter 2 (see p. 19). Monismith et al. showed b values are dependent only on material type. Least-square regression fitting was used for the permanent deformation data for 6 in., 9 in., and 12 in. base layers collectively. Generally high values of 2 coefficient of determination ( R ) are as shown in Fig

48 1 6 in. CGr A W= 5.2% 9 in. CGr A W=4.7% 12 in. CGr A W=4.8% Permanent Strain Ratio 1 Least-square regression R-square=.74 1 b Number of Cycles 1 9 in. MCGr A W=4.7% 12 in. MCGr A W=5.7% Permanent Strain Ratio 1 Least-square regression R-square= Number of Cycles 43

49 1 Permanent Strain Ratio 1 6 in. CL W=11.8% 9 in. CL W=11.5% 12 in. CL W=11.6% Least-square regression R-square= Number of Cycles 1 6 in. CGr B W=3.9% 9 in. CGr B W=3.2% Permanent Strain Ratio 1 Least-square regression R-square= Number of Cycles 44

50 1 Permanent Strain Ratio 1 6 in. MCGr B W=6.3% 9 in. MCGr B W=6.2% 12 in. MCGr B W=5.5% Least-square regression R-square= Number of Cycles 1 6 in. CMs W=4.8% 9 in. CMs W=3.7% Permanent Strain Ratio 1 12 in. CMs W=4.2% Least-square regression R-square= Number of Cycles 45

51 1 Permanent Strain Ratio 1 6 in. MCMs W=4.8% 9 in. MCMs W=4.8% 12 in. MCMs W=4.6% Least-square regression R-square= Number of Cycles Fig. 4.1 Log-Log Plot of Permanent Deformation and Number of Cycles Regression parameter a in Monismith s power model (Monismith et al., 1975) was not investigated in this study based on the following reasons: Parameter a is strongly dependent on water content and stress history. In this study, permanent deformations were measured over a range of water contents for the 7 granular base materials. In addition, the applied stress histories for 6 in., 9 in., and 12 in. base layers are different. Insufficient water content and stress history data were generated by the testing program to allow for meaningful investigation of parameter a. The primary purpose of the cyclic plate loading test was to compare permanent deformation resistance of the GAB materials. The influences of water content and stress history were beyond the scope of the project. According to the literature, measurement of parameter a is prone to be erratic. It is determined mostly by the first few load repetitions hence it is heavily influenced by seating and material conditioning procedures. The use of parameter a to predict pavement rutting introduces additional uncertainties. GAB materials may be compacted in single or multiple lifts and experience unknown load applications due to heavy construction equipment before placement of HMA layer. The regression parameter b, the slope of the least-square regression line, represents the rate of permanent deformation accumulation with load applications. In this study, b parameter is used for comparison of rutting resistance (see Fig. 4.2). The value of b is assumed to be an indictor of the GAB material s inherent resistance to rutting (independent of water content and stress history). A smaller b value implies a greater resistance to rutting. 46

52 .45 Regression Parameter b CGr A MCGr A CL CGr B MCGr B CMs MCMs Material Type Fig. 4.2 Log-Log Regression Parameter b Fig. 4.2 shows b values vary from.13 to.35. From Fig. 4.2 it can be observed that the permanent deformation resistance of CL (limestone) is approximately average of the other 6 GAB (non-limestone) materials. Also, it appears that the percent passing No. 4 sieve alone may not be a predictor of rutting resistance. The influence of fines content on permanent deformation is not investigated (note that unmodified and unmodified GAB materials used in this study had essentially the same fines content only the percent passing the No. 4 sieve was increased to modify the GAB material). For two GAB materials (CGrA and CMs) the b parameter increased with modification (suggesting a decrease in rutting resistance). For one material (CGrB) the b parameter decreased with modification (suggesting an increase in rutting resistance). Uniformity and gradation (curvature) coefficients (Cu and Cc) values were computed for each GAB material. Results are presented in Table 4.1 along with parameter b values. Typically, granular materials with a Cu value greater than 6 and a Cc value between 1 and 3 are considered well-graded. Comparison of parameter b and Cc (all Cu values exceed 6) does not reveal any meaningful relation between rutting resistance and Cc. 47

53 Table 4.1 GAB Materials Cu and Cc Values Material Type b D6 (mm) D1 (mm) D3 (mm) Cu Cc CGr A MCGr A CL CGr B MCGr B CMs MCMs Summary The following are concluded from the cyclic plate loading tests: The accumulation of permanent deformation of the 7 GAB materials can be described using a log-log approach. The cyclic plate loading test result confirmed that the b value of the log-log model is an indicator of material type, and is independent of water content and stress history. For the 7 GAB materials, percent passing No. 4 sieve alone is not a determinable parameter for material rutting resistance. Generally, approximately a 1.5 times difference in the b parameter value for unmodified and modified GAB materials is observed. A relationship between rutting resistance and percent passing the No. 4 sieve could not be established using the limited results from this study. A relationship between rutting resistance and gradation curve coefficients could not be established using the limited results from this study. Static Plate Load Test Results and Analysis Introduction Static plate load tests provided surface deflection data for thoroughly conditioned GAB layers (static plate load testing was performed after cyclic testing). The static tests were performed for different GAB layer thicknesses (6, 9, and 12 in.) and different GAB water contents (at and below OMC). Plate deflections were used to backcalculate resilient modulus of the GAB material. For granular materials, resilient modulus is known to be sensitive to stress state and water content. The stress state within a GAB layer varies with pavement geometry, the thickness of GAB layer, and the magnitude and configuration of the wheel load; water content is a function of climate and other factors. Backcalculation of in situ resilient moduli of pavement layer materials is often done using Falling Weight Deflectometer (FWD) data. The peak impact load applied through the FWD load plate and corresponding peak surface deflections are used for 48

54 modulus backcalculation. Peak loads near 9 kips (one half of the standard 18-kip single axle design load) are typically used for analysis. The resulting backcalculated pavement layer moduli provide the basis for pavement analysis and elastic layer design computations. Backcalculated moduli values represent in situ material conditions at the time of FWD testing. Cohesive and granular subgrade moduli and granular base materials moduli are dependent on water content. Little information concerning the water content of unbound base layers is available. In one study by Ping et al. (25), in situ water content data for base materials in Florida were collected over a one year period. It was found that average in situ water content was generally close to compaction OMC from laboratory Proctor tests. Finite Element Analysis Finite Element Analysis (FEA) was performed to determine plate pressure on base layer responses similar to base layer responses within typical flexible highway flexible pavements under a 9-kip single wheel load. A backcalculation program was developed to compute GAB resilient modulus. A Linear Programming approach was used to extrapolate the plate deflections at optimum water content. In this study, FEA was performed using ANSYS, a commercial finite element program package, to simulate flexible pavements under 9-kip single wheel load and base layer under plate load conditions. Pavement components were characterized as linear elastic materials with Poisson s ratio of.35. Five conditions were analyzed. Conditions 1 to 4 consisted of a 4 in. HMA layer, 8 in. GAB layer, and subgrade. A 9- kip single wheel load was simulated by a contact pressure of 8 psi applied to circular area of 6 in. radius. Resilient moduli of the HMA layer and the base layer were varied, representing typical changes that may occur in the field due to environmental factors. Condition 5 consists of 8 in. GAB layer on subgrade under static plate loading (simulating the laboratory static plate loading condition). Fig. 4.3a and Fig. 4.3b present schematic representations of the finite element models for Condition 1 through 5. Table 4.2 summarizes assumed modulus values for Conditions 1 though 5. 49

55 Fig. 4.3a FEA Model for Conditions 1 through 4 Fig. 4.3b FEA Model for Condition 5 5

56 Table 4.2 Resilient Modulus for FEA Models Typical Flexible Resilient Modulus Pavement (4" HMA layer) Resilient Modulus (8" GAB layer) Resilient Modulus (Subgrade) Condition 1 14 ksi 5 ksi 15 ksi Condition 2 1 ksi 25 ksi 15 ksi Condition 3 5 ksi 5 ksi 15 ksi Condition 4 5 ksi 25 ksi 15 ksi Plate Load Test Average Plate Pressure Resilient Modulus (8" GAB layer) Resilient Modulus (Subgrade) Condition 5 3 psi 5 ksi 15 ksi Axial symmetric models were used with dimensions of 12 in. x 12 in. An 8-node rectangular-element type was used, with each node having two degrees of freedom and 9 o element angles. The total number of elements was 14,4, each having dimensions of 1 in. x 1 in. A comparison of solutions was made between the ANSYS results and the results obtained using KENLAYER (Huang, 24), a computer program based on Burmister s layered theory. Fig. 4.4a shows predicted vertical stresses in the middle of the 8 in. base layer for Condition 3. It can be seen that ANSYS and KENLAYER yield essentially the same mid-layer stresses. Fig. 4.4b compares vertical deflections at two depths for Condition 1 determined using ANSYS and KENLAYER. The two solutions check quite closely, with a maximum discrepancy not over 5%. Radial Distance (in.) Vertical Stress (psi) In the middle of the 8 in. base layer (Ansys) In the middle of the 8 in. base layer (Kenlayer) Condition Fig. 4.4a Comparison of ANSYS and KENLAYER Linear Solutions for Vertical Stress in the GAB Layer (Condition 3) 51

57 Vertical Deflection (in.) Radial Distance (in.) At the surface of the subgrade (Ansys) At the surface of the subgrade (Kenlayer) In the middle of the 8 in. base layer (Ansys) In the middle of the 8 in. base layer (Kenlayer) Condition 1 Fig. 4.4b Comparisons ANSYS and KENLAYER Linear Solutions for Vertical Deflection (Condition 1) Deflection and stress solutions at the axis of symmetry were obtained using ANSYS at two depths - at the middle of the 8 in. GAB layer and at the surface of the subgrade. The former is usually considered as a representative location to compute the modulus of the GAB layer and the later is critical for evaluation of permanent deformation for flexible pavements. Results are shown in Figs. 4.5a and 4.5b. These figures confirm that a plate pressure of 3 psi approximates the GAB stress and vertical deflection states for a typical highway flexible pavement under a 9-kip single wheel load. Linear Programming to Determine Deflection at Optimum Water Content Linear relations were observed between deflections at the plate pressure of 3 psi and water content. (see Fig. 4.6). To determine the deflections at optimum water content determined with the laboratory Proctor test, a Linear Programming (LP) technique was used. The problem is, for each GAB material, to find parallel straight lines that best fit the deflection data for 6 in., 9 in., and 12 in. base layer thicknesses, respectively. Mathematically, this requires the regressions produce the smallest average root-mean-square (RMS) error. The average RMS error is defined as follows: RMS = N i= 1 w ( m w N w m c ) 2 52

58 3 25 In the middle of the 8" GAB layer At the Surface of Subgrade Vertical Stress (psi) Condition 1 Condition 2 Condition 3 Condition 4 Condition 5 Fig. 4.5a Comparison of Vertical Stress for Condition 1 to Condition In the middle of the 8" GAB layer At the Surface of Subgrade Vertical Deflection (in.) Condition 1 Condition 2 Condition 3 Condition 4 Condition 5 Fig. 4.5b Comparison of Vertical Deflection for Condition 1 to Condition 5 53

59 and Average RMS = where m i=1 RMS m N = number of water content points per base layer; w m = measured deflection; w c = calculated deflection; m = number of curves for a material. Solutions were obtained using Microsoft Excel Solver. The calculated parallel 2 regression lines are shown in Fig 4.6. The LP algorithm produced convincing R values, varying from.82 to.96, as presented in Fig The figures clearly demonstrate the linear relation between water content and measured deflection for each GAB material tested..23 Deflection (in.) R-square= in. CGr A 9 in. CGr A in. CGr A 6 in. LP Regression.11 9 in. LP Regression 12 in. LP Regression.9.% 1.% 2.% 3.% 4.% 5.% Water Content (%) 54

60 .23 Deflection (in.) R-square=.84 6 in. MCGr A.13 9 in. MCGr A 12 in. MCGr A.11 6 in. LP Regression 9 in. LP Regression 12 in. LP Regression.9.% 1.% 2.% 3.% 4.% 5.% Water Content (%).25.2 R-square=.82 Deflection (in.) in. CL 9 in. CL.5 12 in. CL 6 in. LP Regression 9 in. LP Regression 12 in. LP Regression 2.% 4.% 6.% 8.% 1.% 12.% Water Content (%) in. CGr B 12 in. CGr B 9 in. LP Regression 12 in. LP Regression Deflection (in.) R-square=.96.9.% 1.% 2.% 3.% 4.% 5.% Water Content (%) 55

61 Deflection (in.) in. MCGr B 9 in. MCGr B 12 in. MCGr B 6 in. LP Regression 9 in. LP Regression 12 in. LP Regression.11 R-square=.9.9.5% 1.5% 2.5% 3.5% 4.5% 5.5% 6.5% Water Content (%) Deflection (in.) in. CMs 9 in. CMs 12 in CMs 6 in. LP Regression 9 in. LP Regression 12 in. LP Regression R-square= % 1.5% 2.5% 3.5% 4.5% 5.5% Water Content (%) Deflection (in.) in. MCMs 9 in. MCMs 12 in. MCMs 6 in. LP Regression 9 in. LP Regression 12 in. LP Regression.9 R-square= % 1.% 1.5% 2.% 2.5% 3.% 3.5% 4.% 4.5% 5.% Water Content (%) Fig. 4.6 Water Content versus Plate Deflection (Plate Pressure = 3 psi) 56

62 Backcalculation of Resilient Modulus Deflection values extrapolated at the optimum water content were obtained from Fig A backcalculation program was developed to compute the resilient modulus of the GAB layer from measured plate deflection. It should be noted this algorithm produces a mathematically rigorous solution since there is only one variable, i.e., the resilient moduli of the GAB. The modulus of subgrade was calculated directly from plate test results on subgrade. Fig. 3.6a (Chapter 3) shows the subgrade material exhibits essentially linear elastic properties under the static plate load for load pressures up to 2 psi. The deflection of the rigid plate can be expressed as follows (Huang 24): 2 π (1 ν ) qa w = 2E where w = surface plate deflection; ν = Poisson s ratio; q = average plate pressure; a = plate radius; E = half-space modulus of elasticity. The deflection under the center of a flexible plate can be expressed as: 2 2(1 ν ) qa w = E A comparison of the two equations indicates that the deflection under a rigid plate is 79% of that under the center of a flexible plate. Although based on a homogenous elastic half-space, Yoder and Witczak (1975) demonstrated that the same factor of.79 can be applied for layered elastic systems. Using the center of a flexible plate deflection equation, the modulus of subgrade was calculated as 15,581 psi. The resilient modulus of the subgrade can also be estimated from the CBR test data. Using Heukelom and Klomp and Powell et al. equations presented in Chapter 2, the average resilient modulus of the subgrade was found to be 16,48 psi and 11,822 psi, respectively. The former is very close to the resilient modulus of the subgrade calculated from plate test results. In this study, the subgrade modulus value of 15,581 psi calculated from the plate test was used. An algorithm, based on layered linear elastic theory, was developed to back calculate the modulus of the GAB base materials at the optimum water content. A computer program was developed in VC++ language. It consists of two components, namely, a forward calculation routine and a control module. The forward routine is based on ELSYM5 (Kopperman, 1986), and is called iteratively by the control 57

63 module. The input parameters are subgrade modulus, base layer thickness, Poisson s ratio of the base material, and plate pressures and corresponding deflections. The output of the program is the calculated resilient modulus of the base layer. Note the deflection values were divided by the factor of.79 in the backcalculation program to account for the rigidity of the steel loading plate. A flow chart presenting the implementation of the algorithm is shown in Fig The control module repeatedly calls the forward calculation routine until the assumed GAB modulus produces a computed deflection value equal to the measured deflection. Computation time of this algorithm is rather short; requiring only a few seconds on a personal computer with a 2.4 GHz processor. Fig. 4.7 Implementation of the Linear Backcalculation Algorithm Calculated GAB resilient modulus values varied from 13 ksi to 75 ksi, as shown in Fig The resilient modulus of CL is remarkably greater than those for other GAB materials. It was found that increased percentage passing No. 4 sieve increased, though not significantly, resilient modulus of the GAB materials. Note the resilient modulus was linearly backcalculated from the static plate load test data, hence it includes the influence of base layer thickness. In general, resilient modulus remained practically the same or dropped slightly when the base layer thickness increased. An exception applies for the MCGr A base material, which became stiffer when the base layer thickness was increased from 6 in. to 9 in., then to 12 in. Also, there was a notable reduction in CL modulus for the 12 inch layer thickness. 58

64 in. Base Layer 9 in. Base Layer 12 in. Base Layer Resilient Modulus (ksi) CGr A MCGr A CL CGr B MCGr B CMs MCMs GAB type Static deflection_regression_7gab.xls Fig. 4.8 Backcalculated Resilient Moduli of GAB Materials for Different Layer Thicknesses (at Optimum Water Content) An accepted way to quantify the quality of GAB materials as a component in flexible pavement structure is to compute layer coefficient a 2. There are many correlations between layer coefficient, a 2, and resilient modulus, M R for unbound granular base materials. A well-known expression developed by Rada and Witczak (1981) is: a 2 =.249 log( M R).977 Using this equation, the backcalculated GAB resilient modulus values give layer coefficients ranging from.5 to.24, with an average value of.11 (see Figure 4.9). The average a 2 for CL is.2 and the average a 2 for all other materials is.9. The results are in good agreement with the work by Busching et al. (1971), who recommended an a 2 value of.21 for CL, and.1 for Crushed Granite-gneiss. Deflection, backcalculated resilient modulus, and layer coefficient results are summarized in Table

65 .27 Layer coefficient of the base layer in. Base Layer 9 in. Base Layer 12 in. Base Layer CGr A MCGr A CL CGr B MCGr B CMs MCMs GAB type Fig. 4.1 Layer Coefficients, a 2, for GAB Layers Table 4.3 Deflection, Resilient Modulus, and Layer Coefficient Values for GAB Layer Materials 1 Material Deflection (mil.) for Base Layer Thickness of Resilient Modulus (ksi) for Base Layer Thickness of Layer coefficient (1/ in.) for Base Layer Thickness of 6 in. 9 in. 12 in. 6 in. 9 in. 12 in. 6 in. 9 in. 12 in. CGr A MCGr A CL CGr B N/A N/A N/A.6.7 MCGr B CMs MCMs Evaluated at the optimum water content; stress state approximates a typical flexible pavement under a 9-kips single wheel load. 6

66 Summary The following may be concluded from the static plate loading tests: FEA indicated a plate on GAB pressure of 3 psi approximates base layer stress states for a typical highway flexible pavement structure under a 9-kip single wheel load. Water content has pronounced influence on granular base materials and a linear relation is suggested between the plate deflection and the water content. Increasing percentage passing No. 4 sieve slightly increased the resilient modulus of the GAB materials. Resilient modulus (and layer coefficient) remains practically the same or drops slightly when the base layer thickness increased from 6 in. to 12 in. for most of the GAB materials tested. The average a 2 for CL is.2 and the average a 2 for all other materials is.9. These results are in good agreement with the work by Busching et al. (1971), who recommended an a 2 value of.21 for CL, and.1 for Crushed Granite-gneiss. 61

67 CHAPTER 5 LABORATORY SOIL STIFFNESS GAUGE TESTING PROGRAM AND RESULTS Introduction In this chapter, GAB modulus data obtained using a soil stiffness gauge (SSG) are compared to modulus values backcalculated from plate load test data. The SSG is a relatively new device that provides a means for the direct measurement stiffness or modulus of in-place materials. Its use has been proposed as an alternative to in situ density tests for construction quality control of pavement materials. Over the past several decades, nuclear methods have been widely used to determine in situ material densities for quality control purpose. With the movement of pavement design from empirical to mechanistic-empirical, it is increasingly desirable to directly measure stiffness or modulus of as-placed pavement materials. The SSG was originated from a study co-sponsored by the Federal Highway Administration (FHWA) and the U.S. Department of Defense s Advanced Research Programs Administration (ARPA). ARPA authorized FHWA researchers to supervise the redesign of a military device that used acoustic and seismic detectors to locate buried land mines. The FHWA s partners in this cooperative research and development project were Humboldt Manufacturing Co. of Northridge, IL, Bolt, Beranek & Newman (BBN) of Cambridge, MA, and CNA Consulting Engineers of Minneapolis, MN (Fiedler,1998). A result of this cooperative endeavor is the SSG, introduced to civil engineering field in approximately late A Humboldt SSG is shown in Fig Using surface measurements, a SSG measures as-placed stiffness or modulus of base, subbase, and subgrade materials. Experimental Program The SSG measures the impedance at the material surface. During SSG measurements, wave energy is generated and propagated through the material and bounced back to the exciting source of the SSG ring-shaped foot. In a laboratory setup where the SSG testing is conducted on a material within a container (such as steel compaction mold) the boundary conditions may become a concern. A study by Phillip and Pu (23) indicated for soils with Poisson s ratios up to.4, the SSG stiffness measured in a cylindrical steel mold of 5.9-in. (15-mm) diameter was approximately twice the free-field value. In a related study, Lenke et al. (1991) used geotechnical centrifuge modeling technique to evaluate dynamic soil-structure interaction induced by rigid circular footings. Their results demonstrated the importance and influence of container shapes and boundary materials to approximate true radiation damping of a vertically excited circular footing. Cubical containers with compliant energy-absorbing boundary materials were suggested to reasonably minimize reflected wave energy. Sawangsuriya et al. (22) investigated the boundary effects of different sizes of cubical containers using finite element analysis. Their study indicated the boundary effects became negligible for cubical container width 62

68 greater than approximately 24 inches. Fig. 5.1 Soil Stiffness Gauge (Humboldt Mfg. Co.) The Humboldt SSG used in this study has a modulus measurement range of 4 to 9 ksi.. The device weighs approximately 22 lbs., is cylindrical with a diameter of 11 in., a height of 1.5 in. The SSG rests on the soil surface, sitting on a ring shaped foot. A shaker is attached above the foot and excites the footing vertically, producing small changes in forces and displacements at 25 steady-state frequencies between 1 and 196 Hz. Two sensors attached to the shaker are used to record the force and displacement time history of the foot. The soil stiffness is measured at every frequency and the averaged value is displayed. Additional information is given in Fiedler (1998) and Wu (1998). In the GAB testing pit, SSG tests were performed on the compacted base course prior to every cyclic loading and static loading test. For each base layer, SSG tests were performed at about 1 different locations approximately equally spaced over the base layer surface. To evaluate the repeatability of the SSG measurement, after each test, the SSG was lifted from the surface, then reseated for a second test. The SSG manufacturer (GeoGauge User Guide, Humboldt Mfg. Co., 2) states good measurements require good contact between the SSG ring-shaped foot and the soil surface (at least 6% of the foot s surface). To assure proper seating, a thin layer of moist GAB fines (material passing a No. 3 sieve) was periodically applied between the SSG foot and compacted GAB surface. In addition, to achieve good seating, the SSG was placed on the material 63