Quantifying the Influence of Geosynthetics on Pavement Performance (2017)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

13 CHAPTER 4. EXPERIMENTS, MODELING, AND FINDINGS Introduction This chapter on the experiments, modeling, and findings of the research project reviews the tests that were used commercially and identifies the geosynthetic properties that were most relevant to pavement performance prediction. Triaxial laboratory tests at multiple stress states were run to demonstrate the effect of geosynthetics on the measured cross-anisotropic properties of the base courses in which they were embedded. LST tests were conducted with different base course thicknesses and asphalt layer thicknesses, and a single concrete layer thickness. Because pavement edges and joints were of prime importance to concrete pavement performance, the loads were applied at the edge of the concrete pavement. The slippage that was observed between the geosynthetics and base courses revealed the part of the data from a pullout test that was most useful in characterizing the contribution of geosynthetics to reinforcing a base course. Finite element models of each pavement layer and the interfaces between the geosynthetic and base course were used to determine the properties of each layer that were needed to replicate the LST measurements. In general, the comparisons of the predicted and measured results were excellent. In order to provide the Pavement ME Design software with the capability to include the effects of embedded geosynthetics, ANN models predicting the critical strains in asphalt pavements were prepared. The models could reproduce the results of thousands of runs with the analytical software, representing a wide variety of pavement structures and layer material properties. Because of the insensitivity of critical stresses in concrete pavements to either type or location of geosynthetics, no ANN models were developed for rigid pavements. Geosynthetic Application and Reinforcement Mechanisms Geogrids and geotextiles have been the most commonly used geosynthetic products in unbound base layers (i.e., within the layer or as a subgrade/base interface layer) as a means of enhancing the performance of flexible and rigid pavements. Beneficial effects of the geosynthetic layer were identified in the responses of pavements under traffic loading through two major mechanisms (see Figure 4.1):  Lateral confinement, which was produced by the interface frictional interaction and interlocking between base course aggregates and the geosynthetic layer. Significant tensile stress was generated in the geosynthetic layer when a spreading motion was created by traffic loading, which in turn reduced the vertical stress and shear stress dramatically due to the increased base course stiffness (2, 43).  Vertical membrane effect. The inward shear stress caused by membrane deformation reduced the outward shear stress generated by repetitive wheel loading. As a result, the vertical stress was reduced and distributed widely around the geosynthetic layer (6).

In by geotex separatio modulus Availabl Perform T physical layer and performa geosynth propertie        Test stan    F addition to tiles was an n reduced th of the base e Test Meth ance he geosynth properties, m aggregates/ nce-related etic properti s included: Tex-621-J than the m ASTM D ASTM D5 ASTM D6 ASTM D7 ASTM D7 ASTM D6 dards for ev ASTM D5 ASTM D4 the 15 per ASTM D4 igure 4.1. M the above m other impor e base cour course and t ods for Ev etic properti echanical p soils. Many geosynthetic es are listed , to measure aximum ag 1777, to mea 818, to dete 637, to mea 737, to mea 748, to mea 706, to dete aluating geo 199, to mea 751, to mea cent passing 491, to mea echanisms ajor reinfor tant functio se contamin hen increase aluating Ge es that were roperties, an test method properties. below. Spe the apertur gregate size sure the dim rmine the r sure the rib sure the jun sure the fle rmine the g textile prop sure the she sure the app particle siz sure the per 14 of Geosyn cement mec n that prolon ation (2), wh d the pavem osynthetic related to p d interface s were cond The test sta cifically, tes e size, whic in the base ensions of esistance to tensile stiff ction streng xural rigidit eogrid-aggr erties includ et thickness arent openi e. meability, w thetics in P hanisms, th ged the pav ich signific ent service Properties R avement pe properties b ucted to eva ndards and t t standards h must be at course grad the geogrid installation ness. th and junct y. egate/soil sh ed: . ng size, whi hich affect HMA/PCC avement e layer sepa ement servi antly increa life. elated to P rformance in etween the luate the pa he correspo for evaluatin least 50 pe ation (5). ribs. damage. ion efficien ear interfac ch should b s the drainag ration provi ce life. Laye sed the resil avement cluded the geosynthetic vement nding g geogrid rcent greater cy. ial propertie e smaller th e function. ded r ient s. an

15  ASTM D5493, to measure the permittivity, which influences the filtration function.  ASTM D4595, to measure the tensile stiffness.  ASTM D6241, to determine the California bearing ratio (CBR) puncture strength.  ASTM D6706, to determine the geotextile-aggregate/soil interfacial properties. Selection of Test Methods for Determining Geosynthetic Properties A comprehensive review of available test methods for determining the performance- related geosynthetic properties is provided in Appendix A. The criteria of test method selection included the following:  The test method should have the characteristics of simple operation, short test time, and low cost.  The test method should be repeatable and reliable.  The test method should be applicable to different types of geosynthetics.  The determined geosynthetic properties should be directly related to pavement performance.  The determined geosynthetic properties should be capable of being input into the finite element program. Based on the selection criteria above, the direct tension test and the pullout test were determined to be the best tests to measure the tensile sheet stiffness of geosynthetics and determine the geosynthetic-aggregate/soil interfacial properties, respectively, both of which significantly affected the performance of geosynthetic-reinforced pavements. Direct Tension Test to Determine Geosynthetic Sheet Stiffness The geosynthetic sheet stiffness was closely related to the performance of geosynthetic reinforcement on unbound aggregates (49). Specifically, an increase in the tensile stiffness of the geosynthetic increased the resilient modulus of the geosynthetic-reinforced unbound aggregates. In addition, it reduced the permanent deformation of the material. Two test standards, ASTM D6637 and ASTM D4595, as mentioned above, were employed to measure the force- strain relationships of geogrids and geotextiles, respectively, through the direct tension test. The sheet stiffnesses of geogrids and geotextiles were determined using Equations 4.1 and 4.2, respectively. Note that the sheet stiffness at a small strain stage (i.e., tensile strain was less than 1 percent) was typically considered a key value in base reinforcement applications. Ea TM s     (Geogrid) (4.1) TM Et     (Geotextile) (4.2) where M is the sheet stiffness of the geosynthetic; E is the tensile modulus of the geosynthetic; a is the cross-section area of the geogrid rib; s is the spacing of the geogrid ribs; t is the thickness of the geotextile; T is the applied incremental tensile force; and  is the corresponding incremental tensile strain.

Pullout T T using the embedde pullout fo variable d The pullo nonlinear geosynth the interf Equation G where G G is the maximum model co the interf δ est to Deter he interactio pullout test d in the base rce was rec ifferential t ut force ver stage, and etic interact acial shear m 4.3. m ( ) 1 x     x is the in shear modu shear stres efficients th acial shear m mine Geosy n between a (49). Figure course was orded and th ransformers sus geosynt critical stage ion. The pul odulus betw  ax 1 cosh( s A terfacial she lus of the ba s applied on at can be de odulus mo Figure 4 nthetic-Agg ggregates a 4.2 is a sch pulled out e displacem (LVDTs). T hetic displac . Each stage lout test dat een the ge ) sinh G x B  ar modulus se aggregate the geosyn termined fro del are prese .2. Schema 16 regate/Soil nd the geosy ematic plot of the aggre ent of the g ypical pullo ement curv had differe a in the thre osynthetic a ( )x C  as a function s; s is the thetic-aggre m the pullo nted in App tic Plot of t Interfacial nthetic laye of the pullo gate layer by eosynthetic ut test data e had three s nt mechanis e stages wer nd the aggre of the geos thickness o gate interfac ut test data. endix B. he Pullout Properties r was comm ut test. The a tensile pu was measur are shown i tages: linea ms of aggre e interpreted gates, as sh ynthetic em f the shear z e; and  , A The details Test only quanti geosynthetic llout force. ed using lin n Figure 4.3 r stage, gate- to determin own in (4 bedded loca one; max is , B , and C of developin fied The ear . e .3) tion; the are g

17 Figure 4.3. Pullout Force versus Geosynthetic Displacement in a Pullout Test Laboratory Methodology for Quantifying Influence of Geosynthetics The application of geosynthetics had the potential ability to reduce the thickness of the base courses, improve performance, and extend the service life of the pavement structure. Accurate and efficient laboratory characterizations of geosynthetic-reinforced materials were important for including geosynthetic products in pavement design. To develop a laboratory methodology compatible with the current Pavement ME Design software, it was necessary to quantify the characteristics of geosynthetic reinforcement in terms of the resilient properties and permanent deformation properties of the geosynthetic-reinforced UGMs. Influence of Geosynthetics on Cross-Anisotropic Properties of UGMs UGMs were found to exhibit cross-anisotropic resilient behavior (i.e., the properties in the vertical plane were different from the properties in the horizontal plane, while the properties in the horizontal plane were the same in all directions) (21, 50). The cross-anisotropic nature of the base course was demonstrated to be a major factor that influences pavement performance (51). Therefore, quantifying the influence of geosynthetics on the resilient properties of UGMs required the evaluation of the effect of geosynthetics on the cross-anisotropic properties of the base course. One crushed granite material was used in the evaluations in this project. Three types of geogrids and one type of geotextile were selected to reinforce the UGMs. Appendix C elaborates on the material information, including the aggregate gradation, moisture-dry density relationship, and geosynthetic properties. The aggregate specimens were fabricated as 6-inch-diameter and 6-inch-high cylinders at the optimum moisture content using a modified compaction effort (ASTM D1557-12). The influence of the geosynthetic layer depended on its location within the base course. To evaluate this effect, the geosynthetic layer was placed in the middle of the specimen, one-quarter below the middle of the specimen, and at the bottom of the specimen, respectively (see Figure 4.4). 0 1000 2000 3000 4000 5000 6000 0 0.5 1 1.5 2 2.5 Pu llo ut F or ce (l b/ ft ) Relative Displacement (inch) Linear Stage Nonlinear Stage Critical Stage

F T geosynth Universa radial def anisotrop      (c) igure 4.4. S riaxial tests etic-reinforc l Testing M ormations o ic propertie Resilient Resilient Shear mo Poisson’s Poisson’s (a) Control Geosynthet quarter bel 6 inc 6 inc chematic Pl were condu ed aggregat achine (see F f the specim s of each spe modulus of modulus of dulus of the ratio of the ratio of the specimen ic reinforced ow the midd h h ot of Aggre cted on both e specimens igure 4.5). en. The test cimen: the aggregat the aggregat aggregate m aggregate m aggregate m one- le 6 inch 1.5 inch 18 gate Specim the unreinf using the ra During each data were u e matrix in t e matrix in t atrix in the atrix in the atrix in the (b) Geo (d) G ens with/w orced aggre pid triaxial test, LVDT sed to calcu he radial di he axial dir axial plane, axial plane, radial plane synthetic rei eosynthetic bot 6 i 6 i ithout Geo gate specim test (RaTT) s measured late the foll rection, rE . ection, zE . rzG . rz . , rr . nforced in t reinforced tom nch nch synthetic ens and cell with th the axial an owing five he middle at the 3 inch e d

T equations rewritten   where  strain in t A the loadin static stre employed the dynam resilient he loading p of the cros in the incre 1 1 rz r r rz r z E E E E     1 1 rz r r rz r z E E E E     r is the stres he radial dir ccording to g protocol, ss states ass in the triax ic stress co strain was ac Figure 4.5 rotocol used s-anisotropic mental form rr r r z rz r r E E                 rr r r z rz r r E E          s in the radi ection; and the cross-an including th ociated with ial test, as s nsisted of 1 hieved after . Configur in the triax aggregate in Equation r z        r z          al direction; z is the str isotropic co e compress correspond hown in Tab .5 seconds o 25 repetitio 19 ation of Ra ial test was specimens, a 4.5 for the z is the st ain in the ax nstitutive re ion, shear, a ing dynami le 4.1. In ea f loading an ns in the dy pid Triaxia developed b s shown in small strain ress in the a ial direction lation, three nd extension c stresses in ch stress sta d 1.5 secon namic loadi l Test ased on the Equation 4. protocol. xial directio . stress mode modes (20 the three str te, every lo ds of unload ng. Uni Test Mac Rap Test Unb Agg Spe constitutive 4, which is (4 (4 n; r is the s were used ). A total of ess modes w ading cycle ing. A stabl versal ing hine id Triaxial Cell ound regate cimen .4) .5) in 10 ere of e

20 Table 4.1. Triaxial Test Protocol for Determining Cross-Anisotropic Properties Stress State Static Stress (psi) Dynamic Stress (psi) Compression Shear Extension z r cz cr sz sr ez er 1 5.8 3.6 0.7 0 1.5 −0.7 −0.7 0.7 2 7.3 3.6 1.5 0 1.5 −0.7 −1.5 0.7 3 10.2 5.8 1.5 0 1.5 −0.7 −1.5 1.5 4 18.9 8.7 2.9 0 2.9 −1.5 −1.5 1.5 5 21.8 10.2 2.9 0 2.9 −1.5 −1.5 1.5 6 24.7 14.5 2.9 0 2.9 −1.5 −2.9 2.9 7 31.9 17.4 4.4 0 4.4 −2.2 −2.9 2.9 8 36.3 20.3 4.4 0 4.4 −2.2 −2.9 2.9 9 36.3 17.4 4.4 0 4.4 −2.2 −2.9 2.9 10 36.3 15.2 4.4 0 4.4 −2.2 −2.9 2.9 The measured axial and radial strains in every loading mode were analyzed using the system identification method to back-calculate the five cross-anisotropic properties— rE , zE , rzG , rz , and rr —based on the constitutive model presented in Equations 4.4 and 4.5. The results of the calculated cross-anisotropic properties are presented in Appendix C. The cross- anisotropic properties of the control specimens were compared to those of the geosynthetic- reinforced specimens by calculating the normalized material property ratio of the control specimen to the reinforced specimen. Tables 4.2, 4.3, and 4.4 show examples of the comparison results of the aggregate specimens with a geosynthetic layer at the different locations. In these tables, the parameter AR represents the anisotropic ratio, which is the ratio of horizontal modulus to vertical modulus.

21 Table 4.2. Influence of Geosynthetic on Material Properties—Geosynthetic Location: Mid-Height Stress State Geosynthetic Type r geosynthetic r control E E   (%) z geosynthetic z control E E   (%) rz geosynthetic rz control G G   (%) geosynthetic control AR AR (%) 1 Geogrid 123 120 127 103Geotextile 153 92 110 166 2 Geogrid 117 131 129 89Geotextile 157 109 107 144 3 Geogrid 126 120 113 105Geotextile 144 98 99 147 4 Geogrid 118 121 108 98Geotextile 131 100 110 131 5 Geogrid 124 116 116 107Geotextile 132 104 103 127 6 Geogrid 122 115 113 106Geotextile 127 99 104 128 7 Geogrid 112 111 114 101Geotextile 124 103 104 120 8 Geogrid 111 112 117 99Geotextile 124 95 99 131 9 Geogrid 121 109 122 111Geotextile 117 98 103 119 10 Geogrid 110 110 126 100Geotextile 122 102 103 120

22 Table 4.3. Influence of Geosynthetic on Material Properties—Geosynthetic Location: One-Quarter below the Middle Stress State Geosynthetic Type r geosynthetic r control E E   (%) z geosynthetic z control E E   (%) rz geosynthetic rz control G G   (%) geosynthetic control AR AR (%) 1 Geogrid 121 110 118 110 Geotextile 132 85 109 155 2 Geogrid 109 122 120 89 Geotextile 125 97 101 129 3 Geogrid 112 124 115 90 Geotextile 118 102 109 115 4 Geogrid 119 120 114 99 Geotextile 122 95 117 128 5 Geogrid 108 119 121 91 Geotextile 124 99 104 125 6 Geogrid 111 115 106 96 Geotextile 115 91 94 126 7 Geogrid 113 124 127 91 Geotextile 106 103 108 103 8 Geogrid 119 108 117 110 Geotextile 109 95 104 114 9 Geogrid 115 110 114 104 Geotextile 107 89 108 120 10 Geogrid 109 108 111 101 Geotextile 110 87 101 126

23 Table 4.4. Influence of Geosynthetic on Material Properties—Geosynthetic Location: Bottom Stress State Geosynthetic Type r geosynthetic r control E E   (%) z geosynthetic z control E E   (%) rz geosynthetic rz control G G   (%) geosynthetic control AR AR (%) 1 Geogrid 110 109 105 101Geotextile 104 97 116 107 2 Geogrid 107 112 98 96Geotextile 104 93 112 112 3 Geogrid 109 110 89 99Geotextile 108 95 115 114 4 Geogrid 103 105 102 98Geotextile 94 102 106 93 5 Geogrid 103 98 105 105Geotextile 107 96 102 111 6 Geogrid 99 103 100 96Geotextile 116 103 103 113 7 Geogrid 95 97 109 98Geotextile 104 95 97 109 8 Geogrid 102 96 107 107Geotextile 106 93 97 114 9 Geogrid 105 95 109 110Geotextile 102 92 97 110 10 Geogrid 95 98 102 97Geotextile 97 93 98 104 As Tables 4.2 and 4.3 illustrate, the modulus ratios, r geosynthetic r control E E   , z geosynthetic z control E E   , and rz geosynthetic rz control G G   , of the geogrid-reinforced specimens were larger than 100 percent in every stress state, which showed that the geogrid increased rE , zE , and rzG of the aggregate matrix specimen. However, a review of the existing studies conducted on the effects of geosynthetic reinforcement on UGMs revealed that the geogrid had only a slight influence on the vertical modulus zE when the specimens were fabricated as 6-inch-diameter and 12-inch-high cylinders (14, 15). In this study, researchers found that the application of geogrids increased the vertical modulus of UGMs by approximately 10–20 percent when the specimens were fabricated as 6-inch-diameter and 6-inch-high cylinders. It was inferred that the benefits of geosynthetic reinforcement were significantly influenced by the dimension of the UGM specimen. This phenomenon also explained why geogrid-related increases in the resilient modulus of base courses were found in in-service reinforced pavement sections and full-scale pavement sections by back-calculating the modulus of the reinforced base course using falling weight deflectometer (FWD) data (53).

24 In contrast to the geogrid, the application of a geotextile slightly reduced the vertical modulus of the UGM specimen but significantly raised the horizontal modulus. As a result, the geotextile increased the anisotropic ratio of the specimen by 20~60 percent, which indicated that the geotextile made the specimen more isotropic (54). Compared to Tables 4.2 and 4.3, the test results in Table 4.4 demonstrate that placing the geosynthetic at the bottom of the UGM specimen did not show any benefits related to an increase in the cross-anisotropic properties. Influence of Geosynthetics on Permanent Deformation Properties of UGMs Rutting or accumulated permanent deformation has been the primary distress for unbound aggregate bases in flexible pavements. It may also be a major factor in the faulting of jointed concrete pavements. Many studies on in-service or large-scale pavement sections found that the application of geogrids significantly reduced the rutting distress of the flexible pavements (8, 55, 56). In the laboratory, the permanent deformation behavior of UGMs with and without geosynthetics was characterized by the repeated load triaxial tests (see Figure 4.5). It was known that the permanent deformation behavior of UGMs was mainly affected by the stress level (57). The stress level also significantly influenced the effects of the geosynthetic on the reduction of the permanent strain of UGMs (25). In this study, the reduction of the permanent strain (RPS) was defined as:  % 100%permanent strain without geosynthetic permanent strain with geosyntheticRPS permanent strain without geosynthetic   (4.6) As shown in Figure 4.6, the RPS of the geogrid-reinforced UGM was only 13.5 percent when the deviatoric shear stress d was 10 psi. This indicated that the effect of geogrid reinforcement was not significant when the deviatoric shear stress was small. When the deviatoric shear stress reached 19 to 28 psi, the RPS increased to a value between 28.4 and 36.5 percent, which indicated that the reduction of permanent deformation was greater at high deviatoric shear stress levels. Figure 4.6. Effect of Stress Level on Reduction of Permanent Strain 0 10 20 30 40 σd=10 psi σd=19 psi σd=25 psi σd=28 psi R ed uc tio n of P er m an en t S tra in (% )

25 In order to characterize the stress-dependent permanent deformation behavior of UGMs with and without geosynthetics, a new permanent deformation model was proposed, as shown in Equations 4.7 to 4.9. The proposed model was able to determine the accumulated permanent deformation at any specific stress state and number of load repetitions.    0 2 1m nNp e J I K          (4.7)   2sin 3 3 sin    (4.8)   6cos 3 3 sin cK    (4.9) where 2J is the second invariant of the deviatoric stress tensor; 1I is the first invariant of the stress tensor; 0 ,  ,  , m , and n are model coefficients; and c and  are cohesive shear strength and friction angle, respectively. In this model, the two terms, 2J and 1I K  , were incorporated into the Tseng-Lytton model (58), which was used to reflect the influence of a stress state on the permanent deformation of the UGM. Figure 4.7 illustrates the concept of the permanent deformation model. The Drucker- Prager plastic yield criterion (59), which was widely applied to rock, concrete, and other pressure-dependent materials, was the basis of this model. As shown in Figure 4.7, the black dot represents the current stress state in the  1 2I J plane; the parameter 2J represents the softening effects of the deviatoric shear stress on the UGM, and a higher 2J yields a larger permanent deformation. Thus, the power coefficient m in Equation 4.7 was always a positive number. In addition, the term  1I K  indicated the hardening/strengthening effect of the hydrostatic stress on the UGM, which was highly affected by the material cohesion and internal friction angle. A higher  1I K  value resulted in a smaller plastic deformation; thus, the power coefficient n in Equation 4.7 was always a negative number.

26 Figure 4.7. Illustration of the Stress-Related Terms in the Proposed Model Table 4.5 shows the seven stress levels designed to determine the coefficients of the proposed rutting model. Stress States 1, 2, 3, and 4 employed the same 1I but different 2J , whereas Stress States 1, 5, 6, and 7 applied the same 2J with various 1I . This test protocol allowed for quantifying the influence of 1I and 2J on the permanent deformation behavior of UGMs with and without geosynthetics, individually. Note that Stress State 4 represented a hydrostatic state, which could also be used to verify that the plastic behavior of UGMs was marginal under the hydrostatic condition. Table 4.6 presents the other two stress states used to validate the determined coefficients in the permanent deformation model. Table 4.5. Proposed Permanent Deformation Test Protocol—Proposed Stress Levels for Calibration of Model Coefficients Stress State Confining Pressure, σ3 (psi) Deviatoric Stress, σd (psi) Bulk Stress, I1 (psi) Second Invariant of Shear Stress Tensor, J2 (psi2) 1 4.0 28.0 40.0 261.3 2 7.0 19.0 40.0 120.3 3 10.0 10.0 40.0 33.3 4 13.3 0 40.0 0 5 7.0 28.0 49.0 261.3 6 10.0 28.0 58.0 261.3 7 13.0 28.0 67.0 261.3 Failure Envelope Hardening Capacity (Available Strength) Softening Force Stress State ඥܬଶ I1 ඥܬଶ ൌ αܫଵ ൅ ܭ

Table Stress S 8 9 T deformat comparis different Figure 4. errors (R general, a stress lev the meas Figur T predictio model co States 8 a the mode model wa 0 0 0 0 0 1 A cc um ul at ed P la st ic S tr ai n (% ) 4.6. Propos tate C P he coefficie ion curves u ons of labor stress levels 8. The recor MSEs) were smaller RM el was relati ured perman e 4.8. Comp est data from n accuracy o efficients. F nd 9 to the l had small s able to qu .0 .2 .4 .6 .8 .0 1 ed Permane V onfining ressure, σ3 (psi) 5.0 15.0 nts of the m sing the solv atory-measu for the teste ded perman calculated SE indicate vely small, ent deforma arison of L Stress Stat f the rutting igure 4.9 co permanent d RMSE value antify the ef 10 Number of L nt Deforma alidation o Devia Stress, 25 28 odel were de er function red and mo d specimen ent strain sta to evaluate t d a better g which indic tion curves ab-Measur Deform es 8 and 9 s model. The mpares the eformation s, which sh fect of the s 100 oad Cycles 27 tion Test P f Model Co toric σd (psi) .0 .0 termined by in the softw del-predicte s. Stress sta rted from th he goodness oodness of f ated that the . ed and Pro ation Curv hown in Tab se stress sta measured pe curves predi owed that th tress level o 1000 in Logarithm rotocol—P efficients Bulk Stress (psi) 40.0 73.0 fitting the are Excel. F d accumulat te is abbrevi e 15th load of model f it (60). The model accu posed Mod es le 4.6 were tes were no rmanent de cted by the e model had n the perma 10000 Scale roposed Str , I1 Seco Shea measured pe igure 4.8 pr ed permane ated as “S” cycle. The r it at various determined rately captu el-Predicte used to vali t used in det formation cu model. The a high accu nent deform 100000 ess Levels nd Invarian r Stress Ten J2 (psi2) 208.3 261.3 rmanent esents the nt strains at in the legen oot-mean-sq stress states RMSE at ea red the tren d Permanen date the ermining th rves for Str predictions racy. The ation behav Granite Granite Granite Granite Granite Granite Granite Predict S Predict S Predict S Predict S Predict S Predict S Predict S for t of sor, d of uare . In ch d of t e ess of ior S1 S2 S3 S4 S5 S6 S7 1 2 3 4 5 6 7

28 of the UGM with and without geosynthetics. Table 4.7 lists the model coefficients determined for the tested UGM with and without geosynthetics. Table 4.7. Determination of Model Coefficients for the UGM with and without Geosynthetics Material Type Permanent Deformation Model Coefficients ε0 ρ β m n Unreinforced 0.149 72.4 0.247 1.70 −2.16 TX-Geogrid Reinforced 0.079 48.3 0.174 1.68 −2.10 BX-Geogrid Reinforced 0.082 31.2 0.182 1.64 −2.01 Geotextile Reinforced 0.112 60.4 0.261 1.76 −2.18 Note: TX = triangular geogrid; BX = rectangular geogrid. Figure 4.9. Validation of Prediction Accuracy of Proposed Permanent Deformation Model This study also evaluated the effects of the type of geosynthetic and location of geosynthetic on the permanent deformation behavior of UGMs. Figure 4.10 shows an example of the influence of the type of geosynthetic on the permanent deformation of the UGM. In Figure 4.10, TX denotes a type of geogrid with triangular apertures, while BX denotes another type of geogrid with rectangular apertures. Figure 4.10 demonstrates that the geogrid with the triangular apertures reduced the permanent deformation of the UGM more than the geogrid with the rectangular apertures. The aperture openings for the triangular (23 mm) and rectangular (25 mm) geogrids were similar. This finding was consistent for all of the tested stress states (see Appendix C). Since the maximum size of aggregates in the UGMs was 19 mm, the apertures of the TX and BX geogrids were around 1.21 and 1.32 times the maximum aggregate size, respectively. Figure 4.11 presents an example of the effect of the location of the geosynthetic on 0 0.2 0.4 0.6 0.8 1 1 10 100 1000 10000 A cc um ul at ed P la st ic S tr ai n (% ) Number of Load Cycles in Logarithm Scale Granite S8 Granite S9 Proposed Model S8 Proposed Model S9 Proposed Model: RMSE=0.031 (S8) RMSE=0.011 (S9)

29 the permanent deformation behavior of the UGM. It was found that the geogrid placed in the middle of the UGM had a greater effect on the RPS than the geogrid placed at one-quarter below the middle of the UGM. The geogrid placed at the bottom did not show any improvement on the resistance to permanent deformation. Some studies reported that the pavement structure with the geogrid placed at the interface between the base layer and subgrade had less permanent deformation than the unreinforced pavement (27, 56). This finding was due to the fact that the application of the geogrid also reduced the vertical compressive stresses in the base layer and subgrade because of the membrane effect, thereby reducing the permanent deformation of the pavement structure. Figure 4.10. Effect of Type of Geosynthetic on Permanent Deformation of UGM Figure 4.11. Effect of Location of Geosynthetic on Permanent Deformation of UGM 0 0.002 0.004 0.006 0.008 0.01 0 2000 4000 6000 8000 10000 A cc um ul at ed P la st ic S tr ai n Number of Load Cycles Unreinforced TX-Geogrid Middle BX-Geogrid Middle 0 0.002 0.004 0.006 0.008 0.01 0 2000 4000 6000 8000 10000 A cc um ul at ed P la st ic S tr ai n Number of Load Cycles Unreinforced Geogrid Middle Geogrid One-Quarter Below the Middle

30 Analytical Model for Quantifying the Influence of Geosynthetics The repeated load triaxial tests indicated that the placement of geosynthetics influenced the cross-anisotropic properties (i.e., the vertical and horizontal modulus) and the permanent deformation properties of the UGM. An analytical model was proposed to predict the vertical and horizontal moduli and the permanent deformation of the geosynthetic-reinforced UGM when it was subjected to a triaxial load. Figure 4.12a shows a schematic plot of a geosynthetic- reinforced UGM specimen in the triaxial load test. The geosynthetic-reinforced specimen was compressed in the axial direction and normally expanded in the lateral direction due to the plastic and resilient deformation. As shown in this figure, the lateral movement of the UGM was restrained by the geosynthetic. The shear stress was generated due to the relative lateral displacement between the geosynthetic and aggregate, which resulted in the stretch of the embedded geosynthetic. Note that the lateral movements of the aggregate and geosynthetic were identical. Figure 4.12b shows the difference in lateral movement between the geosynthetic and aggregate during the test. A coefficient  was employed to account for the difference in radial displacement between the geosynthetic and aggregate, as shown in Equation 4.10. a rr g rr   (4.10) where arr is the aggregate radial tensile strain at the interface between the geosynthetic and aggregate; and grr is the geogrid radial tensile strain. Note that the value of  was normally larger than 1, which meant that the aggregate had a larger lateral movement than the geosynthetic. The analytical solution to determine the coefficient  is shown in Equations 4.11 and 4.12 (61). 0 1 3 2 2 2 D DJ J D                 (4.11)    1/222 1 1a gG M           (4.12) where  iJ x is the Bessel function of order i; D is the diameter of the aggregate specimen (i.e., D = 6 inches); and aG is the shear modulus of the aggregate. Equation 4.12 is an implicit equation for the coefficient  . The stretch of the geosynthetic generated a reinforcement force T to confine the UGM specimen through the aggregate particle interlock and interface friction (17). Figure 4.12c shows that the reinforcement force T was equivalent to a triangularly distributed additional confining stress, 3 , which only acted on a 6-inch geosynthetic- reinforced influence zone (19). This distribution took into account the phenomenon that the influence of the geosynthetic reinforcement decreased with the distance between the aggregate and geosynthetic, and the geosynthetic reinforcement was negligible when the material was far away from the geosynthetic.

31 (a) Displacement Pattern of UGM Restraint by Geosynthetic (b) Difference in Radial Movement of Geosynthetic and Aggregate (c) Equivalence of Reinforcement Force to Additional Stress Δσ3 Figure 4.12. Schematic Plot of Geosynthetic Reinforcement on UGM Specimen Aggregate Before test After test Geosynthetic Deformed Reinforcement Force T

32 Equation 4.13 was used to calculate the maximum equivalent additional stress 3max .            3 3max 33 3 3max13 1 0 2 13max 2 0.851 m nNg H V H M e J I K E E E                         (4.13) where 1 is the axial stress applied to the specimen; 3 is the initial confining pressure; 13 is the Poisson’s ratio to characterize the effect of axial stress on lateral strain; 33 is the Poisson’s ratio to characterize the effect of lateral stress on lateral strain; HE is the horizontal modulus of the specimen; VE is the vertical modulus of the specimen; gv is the Poisson’s ratio of the geosynthetic; and  is the thickness of the influence zone (i.e.,  = 6 inches). In Equation 4.13, the only unknown parameter was the maximum additional confining stress, 3max . An iteration method was utilized to solve for this parameter. Since the thickness of the influence zone  was a constant, the calculated maximum additional confining stress, 3max , could be used to determine the distribution function of the equivalent additional confining stress,  3 z , along the depth, z , of the specimen. The determined equivalent additional confining stress distribution,  3 z , was then input into Equation 4.14 to calculate the modified vertical modulus of the base course,  V ModifiedE z , in the influence zone.     2 31 31 ( 1) k koct V Modified a a a I z E z k P P P          (4.14) where 1I is the first invariant of the stress tensor; oct is the octahedral shear stress; aP is the atmospheric pressure; and 1k , 2k , and 3k are regression coefficients. The effective vertical modulus of the entire geosynthetic-reinforced UGM specimen, V EffectiveE  , was calculated using Equation 4.15, which took into account the variation of the location of the geosynthetic in the UGM specimen.

33         0 2 0 2 0 2 2 2 2 2 2 V UGM V Modified l V UGM V Modified V Effective h l V UGM V Modified E h E z dz l h h E h l E z dz E l h E h l E z dz l h h                                                               (4.15) where V UGME  is the vertical modulus of the unreinforced base course; h is the thickness of the base course; and l is the distance between the geosynthetic layer and the bottom of the base course. The effective horizontal modulus of the geosynthetic-reinforced UGM specimen, H EffectiveE  , was calculated using Equation 4.16. H Effective V EffectiveE n E   (4.16) where n is the ratio of the horizontal modulus to the vertical modulus, which is determined from the repeated load test. Similarly, inputting the determined equivalent additional confining stress distribution,  3 z , into Equation 4.7 allowed for prediction of the permanent deformation of the geosynthetic-reinforced UGM at any given stress level. The detailed derivations of the above analytical models are presented in Appendix D. Figure 4.13 shows the comparison of the resilient moduli of geogrid-reinforced UGMs predicted by the proposed analytical models and those measured from the laboratory tests. The horizontal and vertical resilient moduli predicted by the analytical models matched the measured values with R-squared values of 0.96 and 0.98, respectively. This finding indicated that the proposed analytical models were able to accurately predict both the horizontal and vertical moduli of geogrid-reinforced UGMs.

34 (a) Predicted Horizontal Moduli vs. Measured Horizontal Moduli (b) Predicted Vertical Moduli vs. Measured Vertical Moduli Figure 4.13. Comparison of Resilient Moduli Predicted by Analytical Models with Measured Values LST Test on Pavement Layers with Geosynthetics A comprehensive experimental program, complemented with a detailed numerical model, was designed to capture the mechanism of the interaction between the geosynthetic and the surrounding unbound materials. The measured data obtained from the LST experimental program were used to validate and improve the numerical model. y = 1.0856x − 1.8013 R² = 0.9514 0 10 20 30 40 0 10 20 30 40 Pr ed ic te d H or iz on ta l M od ul us (k si ) Measured Horizontal Modulus (ksi) y = 0.984x + 0.8721 R² = 0.9789 0 30 60 90 0 30 60 90 Pr ed ic te d V er tic al M od ul us (k si ) Measured Vertical Modulus (ksi)

35 Experimental Plan and Setup Test Matrix Experimental Setup One of the largest tank containers in the United States was used to execute the experimental program. This modular cylindrical container, which measured 8 ft in diameter, was divided into three segments; each segment was 3 ft high. Two of the three segments, measuring 6 ft high, along with the base plate were assembled in the laboratory. A 12-inch-diameter actual FWD loading plate (see Figure 4.14) was used to apply the load on the surface of the pavement layer to better simulate actual tire loading conditions. The ratio of the diameter of the LST to the diameter of the loading plate was deemed sufficient to minimize the interference from the LST boundaries. Since the experimental program would include dynamic loading applied onto a pavement structure prepared in a steel container, there was concern about introducing measurement errors in the data collected from the sensors due to reflection of the waves at the boundary. A common technique to minimize such error was to install wave-absorbing material on the inside walls of the steel tank container. A field experiment was performed by the research team to determine the best commercially available wave-absorbing material. The team tested four damping materials (insulation foam, cushion pad, fiberglass, and bubble wrap). It was concluded that fiberglass provided the best absorbing mechanism, so it was selected for this project. The kraft-faced fiberglass insulation was installed in one layer with the kraft side facing inside. After installing the fiberglass material, researchers placed a plastic sheet on the inside of the LST (see Figure 4.15). This sheet provided a frictionless boundary similar to what is expected in the field as well as the numerical model. A hydraulic ram capable of delivering 60,000 lb was used to apply the dynamic and the static loads. The ram was modified by attaching a Moog-252 spool valve that could be electronically controlled to provide just the required flow to the ram to achieve the target dynamic load at the target frequency. The control mechanism also allowed researchers to apply the static load in a controlled manner. The system was connected to a hydraulic pump along with accumulators to ensure adequate flow of hydraulic fluid necessary for the repeated cycles of loading. The ram was mounted onto a stiff beam connected between two vertical columns comprising the reaction frame. A computer running a real-time operating system connected to a National Instrument (NI) four-slot SCXI-1001 chassis populated with two NI SCXI-1320 conditioners was used to control the servo valve. A 20,000-lb interface pancake-type load cell along with a Unimeasure string pot were attached to the ram and electronically connected to the controller. The controller design was a proportional-integral-derivative controller. This control loop feedback mechanism was used to control the ram in either force or displacement control mode. Careful calibration of the gain was essential to ensure the proper operation of the entire loading system. Figure 4.16 shows the completed test setup for a selected flexible pavement experiment.

Figure F 4.14. FWD igure 4.15. (a) Loading Pl Plastic She ate Used in et Covering 36 the LST Ex View the Wave- (b) periments: Absorbing (a) Top Vi Material in ew; (b) Bot the LST tom

Figu Sensors a V pavemen 3500, we the pavem 600 kPa. over a sm to captur one LVD drilled th of the ba concrete loading a and defor YMFLA Micro-el 5 g in thr geosynth geosynth recorded the surfac of the ge In each e LST for f re 4.16. Co nd Instrume arious senso t response to re used to ca ent. These Two Kyow all area. Fo e surface de T was used rough the as se layer only (PML-60) w s well as to mation) wh -2 (2-mm) st ectro-mecha ee direction etic and surr etic materia from the LV e accelerom osynthetic m xperiment, a lexible and mpleted La ntation Pla r types wer loading. N pture the to cells were 4 a 1-inch load ur Novotech flection of th to measure phalt layer, . Texas Me ere used to provide a co en combined rain gauges nical system s were used ounding un l. This meas DT with ba eters. The c aterial as w n extensive rigid pavem rge-Scale T ns e used in thi on-vibrating tal vertical a inches in d cells were nik TR-100 e pavement the top defle to provide d asurements capture the mprehensiv with surfac were used t s-based acc to capture th bound mater urement wa ck-calculate alibrated sc ell as the su network of ents. Table 37 est Setup f s project to -wire Geok nd horizont iameter with also used to LVDTs wit layer. Duri ction of the ata that cou strain gauge response of e picture of e deformati o capture th elerometers e possible i ials as well s achieved b d deflection heme was th rrounding un sensors was 4.8 describe or Flexible capture the g on Total Ear al stresses a capacities t measure th h a range fr ng the flexib base materi ld be used f s for asphalt the pavemen the paveme on data. Tex e strain in th capable of m nterface slip as the defor y calibratin using acce en used to b bound mat installed at s the genera Pavement E eosynthetic th Pressure t different lo hat ranged b e total horiz om 0 to 4 in le pavemen al through a or assessing (PMFLS-6 t under dyn nt responses as Measure e geogrid an easuring a page at the mation shap g the surface leration data ack-calcula erials at diff different lo l instrument xperiment mechanism Cells, Mode cations wit etween 250 ontal stresse ches were u t experimen hole, which the deforma 0-50) and amic and st (stress, stra ments d the geote cceleration u interface of e of the deflection recorded fr te the deflec erent locatio cations in th ation plan u and l hin and s sed ts, was tion atic in, xtile. p to the om tion ns. e sed

38 in the various experiments. Figure 4.17 and Figure 4.18 illustrate a typically instrumented experiment for a reinforced flexible and rigid pavement, respectively. Specific instrumentation diagrams for each of the completed experiments are provided in detail in Appendices E and H. Table 4.8. General Description of the Instrumentation Plan in LST Instrumentation Location Type of Measurements Installation Techniques Strain Gauges for Geosynthetics Geogrid Strain distribution All surfaces were prepared before attaching the strain gauges. For geogrids: the strain gauges were glued on the ribs using epoxy adhesive. For geotextiles: the strain gauges were attached using the silicone adhesive impregnation technique, which involved impregnating the geosynthetic filaments with a thin film of elastic silicon adhesive before attaching the strain gauge. Geotextile Total Earth Pressure Cells Above and below geosynthetic, at the centerline and at the edge of the loading plate Vertical and horizontal stress distributions Earth pressure cells (EPCs) were installed with the flat surfaces horizontal to measure vertical stresses. These EPCs rested on a thin layer of fine sand to ensure uniform pressure distribution. Accelerometers Geosynthetic Dynamic deformation of geosynthetic Accelerometers were embedded in place. Interface between geosynthetic and base or subgrade Slip at the interface Strain Gauges for AC or PCC Bottom of AC or PCC Tensile strain Strain gauge was placed on top of the base prior to the placement and compaction of the AC layer or PCC. LVDTs Pavement surface Pavement surface deflections Linear position sensors were attached to a stationary beam, acting as a reference frame. The moving tip rested on the top of the pavement layer.

Figure 4.17. Instrumentation P X = lan for Fle 0 inch; (b) P 39 (a) (b) xible Pavem lan View a ent Experi t Z = 0 inch ment 4: (a) Profile Vie w at

Figure 4.18. Instrumentation Y = Plan for Ri 0 inch; (b) P 40 (a) (b) gid Paveme lan View a nt Experim t Z = 0 inch ent 9: (a) P G rofile View at

T cells plac done afte excavate carefully material Figure 4. T attention were plac compacte mechanic desired d strings w point of t pressure the vario with appr T accelerom base (for the geosy displacem accelerat In geogrid, measure the geosy to specifi an ultra-f cyanoacr once the applying he instrume ed at a dept r compactin around the c on a leveled was placed c 19 shows th Figure 4.1 he instrume . Five earth p ed in the CA d using the ally compac epth and pla ere tied alon he strings w cells to prot us instrumen opriate wat o measure th eters were 8- and 10-in nthetic imm ents of the ion. Differen addition to or the fabric strain induc nthetic was cations. For ine sandpap ylate adhesi locations we Liquid Nail ntation for th h of 6 inche g the subgra enter of the surface of arefully on e placement 9. Placemen ntation in th ressure cell B. Due to t mechanical ting the CA cing the ear g the diame as used as th ect against s tation senso er content an e differenti embedded i ch aggregat ediately abo subgrade/ba tial displac the accelero of the geote ed in the rein in place to a the geogrid er and then ve, the strain re identified to the geote e subgrade s and 2 inch de to the fin tank to the a thin layer top of the c of the press t of the 4-in e CAB was s, three 4-in he sensitivi rammer. Th B, and then th pressure c ter of the tan e direct cen harp, coarse rs were pla d recompac al displacem n the subgra e base), wh ve the three se and the g ements were meters, stra xtile, next t forcement ssure accur , once the lo cleaned with gauges we , the area w xtile and im 41 layer consis es below the ish grade an desired dept of compacte ell and comp ure cell in th ch Earth P the most ext ch cells, tw ty of the ins erefore, this excavating ells. To loc k, each orie ter. A bed o rocks punc ced, the exca ted manuall ent between de (for 6-inc ile three oth subgrade/b eosynthetic assessed by in gauges w o the accele from surface ate placeme cations wer fast-evapo re attached a as first clea mediately s ted of two 4 subgrade s d then usin h. The pres d fine mater acted by ha e subgrade ressure Ce ensive and t o 1-inch cell trumentation stage was a around the l ate the direc nted differe f fine aggre turing the ce vated base y using a ste the base an h aggregate er accelerom ase accelero were assess subtracting ere placed d rometers in loading. Th nt. All gaug e selected, t rating aceton nd allowed ned with fas ecuring the -inch total e urface. This g a pickaxe sure cells we ial. Next, th nd using a s . ll in the Sub herefore req s, and four , the CAB c ccomplishe ocations of t center of t ntly, and th gates was p lls during te material wa el tamper p d the geosy base) or in eters were p meters. Abs ed by integr the absolut irectly on th the X and Y e installatio es were inst he ribs were e. Using th to dry. For t-evaporatin gauge in pla arth pressur placement w and a shove re then plac e subgrade teel tamper grade uired the m acceleromet ould not be d by placing placement t he tank, thre e intersectio laced around sting. Once s then mixed late. nthetic, thre the aggrega laced on to olute ating the ou e displacem e ribs of the directions t n was done alled accord smoothed u e Tokyo Sok the geotexti g acetone b ce. Once the e as l to ed plate. ost ers and o the e n the all e te p of tput ents. o after ing sing ki– le, efore

glue drie while pla One gaug direction T temperatu heated as After coo This was the strain on top, an mixture w strain gau installatio concrete concrete Care was attached was pour Figure Data Acq A with 20 N channel s a system displacem involving double-in for furthe acquired, imported d, the gauge cing the CA e was place s. he instrume re sensors u phalt binder ling, a smal used to ensu gauge and t d a static pr as then pla ge on top o n of the PC and complet was placed taken to sec lead wire ca ed carefully 4.20. (a) In uisition n NI data ac I SCXI-132 ystem was c was applica ent transdu dynamic lo tegration al r details). D the data we and manipu s were cove B. The locat d directly on ntation used sed for mon was placed l amount of re (a) a pro he asphalt l essure was ced directly f the compa C strain gau e strain tran on top of the ure the stra ble before th over the ga (a -Place Asph quisition sy 0 condition apable of sa ble for acqu cers, load ce ading were gorithm for ata from ex re stored loc lated by mo red with rub ions of the s each rib ad in the aspha itoring tem over the are a fine-grade per support ayer. Once t used to com over the stra cted CAB la ge to ensure sfer was occ CAB, on w in gauge in t e concrete w uge before p ) alt Strain with Tem stem compr ers Hz was mpling data iring data fr lls, pressure acquired at assessing th periments w ally on the st software 42 ber cement train gauge jacent to th lt concrete perature dur a of applica d asphalt m for the strain he strain gau pact the gau in gauge. F yer. A simil that the gau urring from hich the em he desired l as poured. ouring the e Gauge; (b) perature S ised of two used to acqu at frequenc om a wide r cells, and a 1024 Hz to e displacem ith static loa computer ha packages fo to protect fr s were decid e accelerom consisted of ing placeme tion and allo ix was place gauge, and ge was plac ge into the a igure 4.20 s ar practice w ge was com the PCC sl bedded con ocation and A small pat ntire slab. (b) Final In-Pl ensor 12-slot SCX ire the sens ies that rang ange of sens cceleromet accommoda ents and slip ding were a rd drive in C r data analy om damage ed based on eter in the lo one strain g nt. A small wed to coo d over the a (b) a good ed, a steel p sphalt patch hows the ins as followed pletely enc ab. A fresh a crete stain g orientation ch of the co ace Asphalt I-1001 chas or data. This ed from 1 t ors, includi ers. Data fro te the requir page (refer cquired at 3 SV format sis. that might o consistency cal X and Y auge and tw amount of l for 20 min sphalt binde bond betwe late was pla . The aspha talled aspha for the apsulated in mount of auge was p along with t ncrete mater Strain Gau sis populate 80-data- o 3000 Hz. ng strain gau m experime ements for t to Appendix 2 Hz. Once , which coul ccur . o utes. r. en ced lt lt laced. he ial ge d Such ges, nts he I d be

43 Data Analysis Methodologies The laboratory testing program for flexible and rigid pavements included a series of instrumentation that included accelerometers, LVDTs, earth pressure cells, and strain gauges. The instrumentation program was designed to assess several aspects of the influence of the base reinforcement on pavement responses under a variety of realistic pavement loading conditions. A database of pertinent pavement responses with and without reinforcement collected under dynamic and static pavement loading conditions was assembled. The pavement response database was used to assess the validity and applicability of the finite element numerical modeling of reinforced pavement structures. In particular, the instrumentation plan focused on the mechanisms associated with the interaction between the geosynthetic and the unbound materials including (a) assessment of the deflection profile of the geosynthetic; (b) investigation of the slippage at the interface between the unbound material and the geosynthetic; (c) stress transfer across the geosynthetic; and (d) load-induced strains in the geosynthetic. While the last two aspects could be addressed based on direct measurements from pressure cells (vertical and horizontal) and strain gauges, the first two aspects needed to be evaluated based on the deflections at many interior locations within the pavement. The slippage investigation at the interface required measurements of the deflections in the geosynthetic and in the adjacent unbound material to examine the relative movements between the two. As mentioned earlier, the role of the geosynthetic affecting the load transfer across the geosynthetic itself was generally referred to as shell/membrane action. The deformed shape of the geogrid or geotextile located within the unbound pavement layers during the application of the pulse loading was important to evaluate the shell/membrane action of the reinforced layer. The dynamic (instantaneous) deformation of the geosynthetic could be related to the change in vertical stress that could occur across the reinforced CAB layer. Accordingly, high-gain accelerometers were used, with the recording measurements being twice integrated to get the displacement under dynamic loading. It was important to find the best methodology for the double integration of accelerometer readings to get the displacement. Subsequently, these displacements obtained from the integration could be used to assess shell/membrane action of the embedded geogrid or geotextile. The detailed data analysis methodologies are presented in Appendix I. Flexible Pavement The database of pavement responses generated from the LST testing was substantial and covered many aspects of the geosynthetic-unbound material interaction. Therefore, a recap of pertinent key factors that had significant influence on the measured data is provided below. Such information was very useful when navigating through the collected data and during the interpretation process of the various results.  Experiments 1, 3, and 5 included a 6-inch CAB and represented, respectively, the testing for the control (i.e., no base reinforcement), geogrid-reinforced base, and geotextile-reinforced base. The geosynthetic was located in both Experiments 3 and 5 at the bottom of the CAB layer (i.e., 12 inches below the pavement surface).  Experiments 2, 4, and 6 included a 10-inch CAB and represented, respectively, the testing for the control (i.e., no base reinforcement), geogrid-reinforced base, and

44 geotextile-reinforced base. The geosynthetic was located in both Experiments 4 and 6 at the middle of the CAB layer (i.e., 11 inches below the pavement surface).  The stiffness properties of the geogrid and geotextile differed depending on the direction of testing. Overall, at the strain levels expected under the LST loading conditions, the average stiffness on a per-foot basis of the geotextile was considerably larger (as much as 80 percent) than the geogrid.  Unlike the geogrid, the geotextile used in the experiments was a woven continuous carpet-like roll and acted as a separator between the unbound materials present above and below the reinforcement. Geogrid reinforcement, on the other hand, had a cell- like configuration, which provided a better interlocking of the unbound materials across the interface.  The presentation of the results in this section included the earth pressure cell data measured above and below the geosynthetic. The locations of the earth pressure cells were not at the same equal distance above and below the geosynthetic for the 6- and 10-inch aggregates bases.  Unlike earth pressure cells, measurements made by the accelerometers, strain gauges, and LVDTs could be considered as 1-point measurements. The majority of the pressure cells used in the LST experiments were 4 inches in diameter (a couple of the pressure cells were 1 inch in diameter and were mainly used for measurements of the horizontal pressure), and the fluid present within the flexible diaphragm gave the average induced pressure within the entire surface area of the cell.  Studies showed that in many typical pavement configurations without reinforcements, there could be a reduction in horizontal stresses in the CAB, especially near the centerline of the loaded area and directly below the AC surface layer. This phenomenon was governed by the thickness and stiffness of the various pavement layers.  The earth pressure cells used to measure horizontal stresses in the CAB were located laterally away from the axis of loading and at 8 inches from the centerline (under the edge of the plate) in all experimental tests.  While an AC layer of 6 inches was targeted for all the LST experiments, measurements from the post-test core specimens revealed variations in the in-situ AC layer thickness within an individual experiment and among the various experiments. Such variation in the AC layer thickness could have an influence on the various measured pavement responses and should be kept in mind when interpreting the LST test results. Stress Distributions Across the Reinforcements First, the collected earth pressure cell data were reviewed. Figure 4.21 presents an abridged version of the LST configuration showing only the earth pressure cells in the CAB for the pavement with the thin (i.e., 6-inch) CAB (Experiments 1, 3, and 5). The locations of the earth pressure cells are provided with modified identifications to facilitate the interpretation of the results. For example, the subscripts “Center” and “Edge” refer to the pressure cells on the centerline and under the edge (8 inches from the centerline) of the loading plate, respectively.

45 The superscripts “Above” and “Below” refer to the pressure cells above and below the geosynthetic. The horizontal stress in the CAB under the edge of the loading plate is referred to as B-Horiz. A review of the assembled results with a focus on assessing the influence of the reinforcement indicated many noteworthy observations. A summary of the important results and interpretations are presented below. Figure 4.22a and Figure 4.22b show the load-induced vertical stress measurements along the centerline of the load for above and below the reinforcement location, respectively. Similarly, Figure 4.23a and Figure 4.23b show the vertical stress measurements along the edge of the load for above and below the reinforcement location, respectively. Figure 4.22c shows the difference in the vertical stresses above and below the reinforcement location at the centerline and edge of the loading plate, respectively. As expected, the vertical stresses above and below the reinforcement location consistently increased with the increase in the applied load level. The form of the relationship between the vertical stress and the load level could be used to investigate the presence and extent of the nonlinearity in unbound materials. This check for the possible presence of nonlinearity could be readily undertaken with all datasets generated in the LST testing program. At the centerline of the load, lower vertical stresses (above and below the reinforcement location) were observed in the experimental tests with reinforcements when compared to the control experiment. These reductions in vertical stresses due to CAB reinforcement were slightly higher with the geogrid reinforcement. On the other hand, except for the vertical stress above the geogrid, higher vertical stresses (above and below the reinforcement location) were observed at the edge of the loading plate in the experimental tests with reinforcements when compared to the control experiment. At the centerline of the load, the vertical stresses above the reinforcement location (i.e., in the middle of the 6-inch CAB) were consistently found to be lower than those measured below the reinforcement location (i.e., 2 inches below the subgrade surface), even though the pressure cells located above were closer to the pavement surface. This observation may be attributed to a substantial reduction in the compressive horizontal stresses due to bending of the pavement layers under loading and, accordingly, a reduction in the CAB stiffness at this location. The difference in the vertical stresses at the centerline of the load and across the geosynthetic location was found to be the highest in the case of no reinforcement (i.e., control experiment). The lowest difference was observed when the geotextile was used, and it was more pronounced at the higher load level of 16 kip (see Figure 4.23c). Unlike the observations at the centerline of the load, the vertical stresses along the edge of the loading plate and above the reinforcement location were found to be higher than those measured below the reinforcement location (i.e., in the subgrade layer) for both the control and geotextile-reinforced base. A different behavior was observed with the geogrid when compared to the geotextile-reinforced base, where the vertical stresses at the edge of the loading plate were slightly lower below the geogrid when compared to the vertical stresses above it. The difference in the vertical stresses at the edge of the loading plate and across the geosynthetic location was found to be significant for the control and geotextile-reinforced base, with a higher difference being observed in the latter case.

46 Figure 4.24 shows the horizontal stresses in the CAB layer measured along the edge of the loading plate. Noticeably lower horizontal stresses were observed with the reinforced base layer when compared to those measured in the control experiment (i.e., no reinforcement). The difference in behavior between the geogrid and geotextile was minor. Figure 4.25 presents an abridged version of the LST configuration showing only the earth pressure cells in the CAB for the pavement with the thick (i.e., 10-inch) CAB (Experiments 2, 4, and 6). Similar to the thin pavement case, the locations of the earth pressure cells are provided with modified identifications to facilitate the interpretation of the results. As noted before, it should be kept in mind that the relative offsets of the pressure cells above the reinforcements are not exactly the same between the experiments with thin and thick CAB. Figure 4.26a and Figure 4.26b show, for the thick CAB layer, the load-induced vertical stress measurements along the centerline of the load for above and below the reinforcement location, respectively. Similarly, Figure 4.27a and Figure 4.27b show the vertical stress measurements along the edge of the load for above and below the reinforcement location, respectively. Figure 4.27c shows the difference in the vertical stresses above and below the reinforcement location at the centerline and edge of the loading plate for the thick CAB layer. Similar to the case of the thin CAB, the vertical stresses above and below the reinforcement location consistently increased with the increase in the applied load level. At the centerline of the load, higher and lower vertical stresses were observed, respectively, above and below the reinforcement location in the experimental tests for the thick CAB with reinforcements when compared to the control experiment. On the other hand, noticeably lower vertical stresses above the reinforcement location were observed at the edge of the loading plate in the experimental tests with reinforcements when compared to the control experiment in the thick CAB layer. The reduction in the vertical stress at the edge of the loading plate was more significant in the case of the geotextile when compared to the geogrid-reinforced base. The reduction in the vertical stress at the centerline of the load and across the geosynthetic location was substantial in the case of the geogrid when compared to the geotextile. Unlike at the locations along the centerline of the load, the differences in stresses across the geosynthetics were not substantial, especially in the case of the geogrid. It should be noted that the experiment on the thin CAB layer with geogrid (i.e., Experiment 3) did not show any noticeable difference in vertical stresses across the geogrid at all load levels. The horizontal stress measurements in the thick CAB layer measured along the edge of the loading plate are shown in Figure 4.28. Unlike the case of the thin CAB layer, substantially higher horizontal stresses were observed in the reinforced base layer when compared to those measured in the control experiment (i.e., no reinforcement). The observed horizontal stresses were also higher in the case of the geogrid when compared to the geotextile-reinforced base. It should be noted that the measurements of the horizontal stresses in the case of the control experiment (i.e., Experiment 2) were made with a 4-inch-diameter pressure cell as opposed to a 1-inch-diameter cell, as in the case of the reinforced experiments (Experiment 4 and 6). This difference in the type of pressure cells might have influenced the horizontal stress measurements in the control experiment. As noted before, the average induced pressure within the entire surface area of the pressure cell was being measured, which might partially explain the reason behind the low observed magnitudes for the horizontal stresses in the control experiment.

Figure 4 .21. LST C and 5 onfiguratio ) Showing O n for Flexib nly Earth 47 le Paveme Pressure C nts with Th ells across in CAB (Ex Geosynthet periments ic 1, 3,

48 (a) (b) (c) Figure 4.22. Vertical Stresses at the Centerline of the Loading Plate for Thin CAB Layer (Experiments 1, 3, and 5) 7.4 10.7 15.5 5.4 8.1 12.4 6.4 9.2 15.3 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 1, 3, and 5 Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ େୣ ୬୲ ୣ୰ ሺps iሻ 9.6 13.1 18.0 7.6 10.3 14.1 8.3 11.2 16.3 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 1, 3, and 5 Control Geogrid Geotextile Ve rti cal St res s,σ ୗୋ ି୆ ୣ୪୭ ୵ େୣ ୬୲ ୣ୰ ሺps iሻ -2.2 -2.4 -2.5-2.2 -2.2 -1.8-1.9 -2.0 -1.1-5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 1, 3, and 5 Control Geogrid Geotextile ሾσ ୆ ି୅ ୠ୭ ୴ୣ େୣ ୬୲ ୣ୰ ‐σ ୗ ୋି ୆ୣ ୪୭୵ େୣ ୬୲ ୣ୰ ሿ ሺp siሻ

49 (a) (b) (c) Figure 4.23. Vertical Stresses at the Edge of the Loading Plate for Thin CAB Layer (Experiments 1, 3, and 5) 9.5 12.8 17.4 6.0 8.0 11.0 13.4 17.9 25.2 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 1, 3, and 5 Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ ୉ୢ ୥ୣ ሺps iሻ 4.9 6.6 9.1 6.7 9.0 12.1 7.5 9.8 13.5 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 1, 3, and 5 Control Geogrid Geotextile Ve rti cal St res s,σ ୗୋ ି୆ ୣ୪୭ ୵ ୉ୢ ୥ୣ ሺps iሻ 4.6 6.2 8.3 -0.72 -0.96 -1.0 5.9 8.1 11.7 -5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 1, 3, and 5 Control Geogrid Geotextile ሾσ ୆ ି୅ ୠ୭ ୴ୣ ୉ୢ ୥ୣ ‐ σ ୗୋ ି୆ ୣ୪୭ ୵ ୉ୢ ୥ୣ ሿ ሺp siሻ

Figur Figure 4 e 4.24. Hor .25. LST C and 6 izontal Stre onfiguratio ) Showing O 0 1 2 3 4 5 Ho riz on tal St res s,σ ୆ି ୌ୭ ୰୧୸ .ሺp siሻ sses at the E (Experim n for Flexib nly Earth 2.2 1.00.81 9 kips Experim Control 50 dge of the ents 1, 3, a le Pavemen Pressure C 3.0 1.3 1.0 12 kips Target Load Lev ent No. 1, 3, an Geogrid Ge Loading Pl nd 5) ts with Thi ells across 4.0 1.8 1.4 16 kips el d 5 otextile ate for Thin ck CAB (E Geosynthet CAB Lay xperiments ic er 2, 4,

51 (a) (b) (c) Figure 4.26. Vertical Stresses at the Centerline of the Loading Plate for Thick CAB Layer (Experiments 2, 4, and 6) 3.9 5.9 9.0 16.6 22.8 32.6 10.3 14.8 21.3 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ େୣ ୬୲ ୣ୰ ሺps iሻ 11.7 16.0 22.6 3.9 5.7 8.97.5 11.7 18.7 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ େୣ ୬୲ ୣ୰ ሺps iሻ -7.8 -10.1 -13.6 12.7 17.2 23.7 2.8 3.0 2.6 -15 -10 -5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile ሾσ ୆ ି୅ ୠ୭ ୴ୣ େୣ ୬୲ ୣ୰ ‐σ ୆ ି୆ ୣ୪୭ ୵ େୣ ୬୲ ୣ୰ ሿ ሺp siሻ

52 (a) (b) (c) Figure 4.27. Vertical Stresses at the Edge of the Loading Plate for Thick CAB Layer (Experiments 2, 4, and 6) 11.6 15.7 21.7 8.6 11.5 15.9 3.2 4.4 6.2 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ ୉ୢ ୥ୣ ሺps iሻ 7.0 9.0 12.0 7.6 11.0 16.9 4.9 6.7 9.3 0 5 10 15 20 25 30 35 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୉ୢ ୥ୣ ሺps iሻ 4.6 6.7 9.7 1.0 0.48 -1.0 -1.8 -2.2 -3.1 -15 -10 -5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile [σ ୆ ି୅ ୠ୭ ୴ୣ ୉ୢ ୥ୣ ‐σ ୆ି ୆ୣ ୪୭୵ ୉ୢ ୥ୣ ሿሺp siሻ

53 Figure 4.28. Horizontal Stresses at the Edge of the Loading Plate for Thick CAB Layer (Experiments 2, 4, and 6) Strain Measurements in the Geosynthetic and at the Bottom of the AC Layer Figure 4.29 and Figure 4.30 present abridged versions of the LST configuration showing only the instrumentations on and around the geosynthetics for thin and thick CAB layers, respectively. The instrumentations shown in the figures are the strain gauges on the reinforcements (designated as SG1 through SG3) and the accelerometers located on and below the geosynthetics. A number of factors that can influence the strain response are listed below and should be considered when interpreting the strain gauge results.  The strain gauges were glued to the top of the ribs in the case of geogrids and on the top surface of the geotextile. The X- and Y-axes represent the tangential and radial directions, respectively.  When the embedded geosynthetics deform under the applied surface load, tensile strains are expected because of the membrane effect. However, locally, when the geogrid, for instance, undergoes bending, there will be tension and compression induced at the bottom and top of the geogrid, respectively. The strain measurements made at the top of the geogrid rib will reflect the net effect of these two deformation mechanisms (membrane and bending).  The laboratory testing of the geosynthetic materials revealed an anisotropic behavior for both the geogrid and geotextile with a noticeable difference in stiffness between the main and cross direction. The stiffness ratio between the two perpendicular directions can be as much as 2.5. Accordingly, the assumption of the existence of an axisymmetric condition for the flexible pavement may not be strictly true.  In Experiments 2 and 4, the geosynthetic was at the interface between the subgrade and the CAB layer, while in Experiments 4 and 6, it was fully surrounded by the CAB. When the all-around crushed aggregates from the base are interlocked with the geogrid, a significant interaction between the ribs and the aggregates is expected. Accordingly, the strain measurements are expected to reflect this localized complex nonlinear interaction, which is also expected to be influenced by the loading cycles and load levels. 0.03 0.04 0.05 2.4 3.3 4.6 1.0 1.3 1.6 0 1 2 3 4 5 9 kips 12 kips 16 kips Target Load Level Experiment No. 2, 4, and 6 Control Geogrid Geotextile Ho riz on tal St res s,σ ୆ି ୌ୭ ୰୧୸ .ሺp siሻ

 T measurem geosynth load leve compared not result T of the lon the AC la respectiv consisten The load strain me bottom o (Experim Figure 4 While a 6 variations among the layer thick measured he above fac ents. Figur etics for bot ls, respectiv to those m in a propor he horizonta g-term perf yer are pres ely. Unlike t and the me -induced ten asurements f the AC lay ents 3 and 4 .29. LST C and 5) Sh -inch AC lay in the thick various ex ness values tensile strai tors indicat e 4.31 and F h thin (6-inc ely. Higher easurements tional increa l strain at th ormance of ented in Fig in the case o asurements sile strains f from the rei er were con ) compared onfiguratio owing Only er was targ ness of the A periments. I will influen ns at the bot ed that care igure 4.32 s h CAB laye strains were at locations se in the me e bottom of flexible pav ure 4.33 and f the strains increased p or the contr nforced base sistently hig to the geote n for Flexib the Instrum 54 eted for all o C layers w t is anticipat ce the meas tom of the A should be ta how the me r) and thick observed at away from asured strai the AC laye ements. The Figure 4.3 in geosynth roportionall ol experime experimen her in the ca xtile-reinfor le Paveme entations f the LST e ithin the ind ed that the d ured respon C layer. ken when in asured horiz (10-inch CA the centerli the load. Th ns. r was a crit strain meas 4 for thin an etics, the A y with the in nt were obse ts. The mea se of the ge ced base (E nts with Th on and arou xperiments, ividual expe ifferences i ses, and in p terpreting t ontal strain B layer) pa ne of the loa e increase i ical input to urements at d thick pave C strain resp crease in ap rved to be b sured tensile ogrid-reinfo xperiments in CAB (Ex nd Geosyn there were riments and n the in-situ articular th he strain dat data on the vements for d when n load level the assessm the bottom ments, onses were plied load l etween the strains at th rced base 5 and 6). periments thetic AC e a all s did ent of very evels. e 1, 3,

Figure 4.30. LST C and 6) Sh onfiguratio owing Only n for Flexib the Instrum 55 le Pavemen entations ts with Thi on and arou ck CAB (E nd Geosyn xperiments thetic 2, 4,

56 (a) (b) Figure 4.31. Horizontal Strains in the Geosynthetic Reinforcements—Flexible Pavements: (a) Experiment 3 (geogrid); (b) Experiment 5 (geotextile) -600 -400 -200 0 200 400 600 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain (X Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -600 -400 -200 0 200 400 600 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain (Y Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -600 -400 -200 0 200 400 600 0 12 24 Radial Distance (inch) Geotextile Horizontal Strain (X Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -600 -400 -200 0 200 400 600 0 12 24 Radial Distance (inch) Geotextile Horizontal Strain (Y Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s)

57 (a) (b) Figure 4.32. Horizontal Strains in the Geosynthetic Reinforcements—Flexible Pavements: (a) Experiment 4 (geogrid); (b) Experiment 6 (geotextile) Figure 4.33. Tensile Strains at the Centerline of the Load and at the Bottom of the AC Layer (Experiments 1, 3, and 5) -800 -400 0 400 800 1200 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain (X Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -800 -400 0 400 800 1200 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain (Y Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -800 -400 0 400 800 1200 0 12 24 Radial Distance (inch) Geotextile Horizontal Strain (X Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -800 -400 0 400 800 1200 0 12 24 Radial Distance (inch) Geotextile Horizontal Strain (Y Direction) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) 0 100 200 300 400 5 10 15 20 Target Load Level (kips) Tensile Strain at the Bottom of the AC Layer Control Geogrid Geotextile Te ns ile S tra in (m icr on s)

58 Figure 4.34. Tensile Strains at the Centerline of the Load and at the Bottom of the AC Layer (Experiments 2, 4, and 6) Deformation of the Geosynthetic and Interface Slippage The vertical deformed shape of the geosynthetic could be determined using the vertical displacements at various locations of the geosynthetic. As mentioned previously, the vertical displacements at the interior locations could be obtained using the proposed calibrated double- integration procedure. However, since these computed displacements were not direct measurements, caution should be exercised when synthesizing the observed measurements, particularly if a lack of proportionality in load level increase was present for any type of response. Figure 4.35 shows the computed load-induced vertical displacements for all load levels in Experiments 3 and 5. The accelerometers were located on the geosynthetic and in adjacent underlying unbound material; hence, two displacement values were available at each of the radial distances. A review of the data in the figures revealed that all computed displacements indicated proportionality with the applied load level except the displacements from the experiment with the geotextile at 9- and 12-kip load levels. Therefore, the results of this test could not be directly used in the data interpretation. The computed vertical displacements at various radial distances on the geosynthetic describe the deformed shape of the geosynthetic itself under loading. In general, minimal differences in vertical displacements between the geosynthetic and the adjacent unbound material were observed at the various load levels, thus lending credibility to the implemented double-integration procedure. Figure 4.36 shows the horizontal displacements of the geosynthetics as a function of radial distance and for all load levels in the case of the thin CAB layer. To evaluate the role of geosynthetics on the horizontal movement, the results of the control experiment (Experiment 1) are also included. The results showed that there was substantial outward movement under the edge of the plate in the case of the control experiment, while much lower or negligible inward movements were computed with the geosynthetics. This was an important observation since the geosynthetics reduced the horizontal displacement under the edge of the plate substantially. An important aspect of the geosynthetic-unbound material interaction was the possible slippage at the interface. Figure 4.37 shows the observed horizontal slippage in the thin CAB experiments, computed as the difference in horizontal displacements between the geosynthetic 0 100 200 300 400 5 10 15 20 Target Load Level (kips) Tensile Strain at the Bottom of the AC Layer Control Geogrid Geotextile Te ns ile S tra in (m icr on s)

59 and the adjacent unbound material. The potential horizontal slippage was calculated at all applied load levels. As noted above, the results for the geotextile-reinforced base under a 12-kip load should be omitted. The data in Figure 4.37 indicated the potential presence of slippage between the geosynthetic and the underlying subgrade material, especially under the edge of the loading plate. This slippage was also observed to be higher with the increase in the applied load level. The slippage between the geogrid and the underlying subgrade material under the 9-kip load was negligible and increased with the 12- and 16-kip load levels. On the other hand, slippage was detected at the 9- and 16-kip load levels for the geotextile case. Similar plots are presented for the experiments with the thick CAB layer (i.e., Experiments 2, 4, and 6) in Figure 4.38 to Figure 4.40. Although the vertical displacements of the geogrid at 16 kip did not follow the proportionality criterion noted above, the localized nonlinear interaction between the geogrid and the surrounding aggregates could explain the anomaly. Similar to the experiments with the thin CAB layer, minimal differences in vertical displacements between the geosynthetic and the adjacent aggregate base material were observed at the various load levels. The results for the horizontal displacements (see Figure 4.39) were analogous to those observed in thin CAB layer cases, where a substantial outward movement under the edge of the plate was seen for the control experiment, while much lower or negligible inward movements were computed with the geosynthetics. On the other hand, the data in Figure 4.40 indicated the potential presence of slippage between the geotextile and the adjacent aggregate base material, especially under the edge of the loading plate. However, the slippage estimates were negligible with the geogrid reinforcement when placed in the middle of the 10-inch CAB layer. The measured slippage from the LST test showed that the maximum relative displacement between the geosynthetic and aggregate was less than 0.04 inch. This suggested that the interface slippage that normally occurred in the geosynthetic-reinforced aggregates was in the linear stage (see Figure 4.3). In general, the observed difference in the behavior between the geogrid and geotextile as a function of load level, pavement structure, and geosynthetic location within the CAB layer warranted the appropriate modeling of the geosynthetic-unbound material layer interface to better capture any influence that the reinforcement may have on pavement responses.

60 (a) (b) (c) Figure 4.35. Vertical Displacements of the Geosynthetic and Adjacent Unbound Material in Experiments 3 and 5 for Various Load Levels: (a) 9 kip; (b) 12 kip; (c) 16 kip -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 9 kips Geogrid Subgrade (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 9 kips Geotextile Subgrade (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 12 kips Geogrid Subgrade (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 12 kips Geotextile Subgrade (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 16 kips Geogrid Subgrade (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 16 kips Geotextile Subgrade (Below) Ve rtic al Di sp lac em en t ( mi ls)

61 (a) (b) (c) Figure 4.36. Horizontal Displacements of the Geosynthetic and Adjacent Unbound Material in Experiments 1, 3, and 5 for Various Load Levels: (a) 9 kip; (b) 12 kip; (c) 16 kip -10 0 10 20 30 40 50 60 0 12 24 36 Radial Distance (inch) 9 kips Control Geogrid Subgrade (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 0 10 20 30 40 50 60 0 12 24 36 Radial Distance (inch) 9 kips Control Geotextile Subgrade (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 0 10 20 30 40 50 60 0 12 24 36 Radial Distance (inch) 12 kips Control Geogrid Subgrade (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 0 10 20 30 40 50 60 0 12 24 36 Radial Distance (inch) 12 kips Control Geotextile Subgrade (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 0 10 20 30 40 50 60 0 12 24 36 Radial Distance (inch) 16 kips Control Geogrid Subgrade (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 0 10 20 30 40 50 60 0 12 24 36 Radial Distance (inch) 16 kips Control Geotextile Subgrade (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward

62 Figure 4.37. Horizontal Slippage of the Geosynthetic and Adjacent Unbound Material in Experiments 3 and 5 for Various Load Levels—Flexible Pavements -8 -6 -4 -2 0 2 4 6 8 0 12 24 36 Ax is Tit le Radial Distance (inch) 9 kips 12 kips 16 kips Ho riz on tal D isp lac em en t D iffe re nc e ( mi ls) ሺD isp ୋୣ ୭୥ ୰୧ୢ െD isp ୗ୳ ୠ୥ ୰ୟ ୢୣ ሺ୆ ୣ୪୭ ୵ሻ ሻ -8 -6 -4 -2 0 2 4 6 8 0 12 24 36 Ax is Tit le Radial Distance (inch) 9 kips 12 kips 16 kips Ho riz on tal D isp lac em en t D iffe re nc e ( mi ls) ሺD isp ୋୣ ୭୲ୣ ୶୲୧ ୪ୣ െD isp ୗ୳ ୠ୥ ୰ୟ ୢ௘ ሺ୆ ୣ୪୭ ୵ሻ ሻ

63 (a) (b) (c) Figure 4.38. Vertical Displacements of the Geosynthetic and Adjacent Unbound Material in Experiments 4 and 6 for Various Load Levels: (a) 9 kip; (b) 12 kip; (c) 16 kip -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 9 kips Geogrid Base (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 9 kips Geotextile Base (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 12 kips Geogrid Base (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 12 kips Geotextile Base (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 16 kips Geogrid Base (Below) Ve rtic al Di sp lac em en t ( mi ls) -25 -20 -15 -10 -5 0 0 12 24 Radial Distance (inch) 16 kips Geotextile Base (Below) Ve rtic al Di sp lac em en t ( mi ls)

64 (a) (b) (c) Figure 4.39. Horizontal Displacements of the Geosynthetic and Adjacent Unbound Material in Experiments 2, 4, and 6 for Various Load Levels: (a) 9 kip; (b) 12 kip; (c) 16 kip -10 -5 0 5 10 15 0 12 24 36 Radial Distance (inch) 9 kips Control Geogrid Base (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 -5 0 5 10 15 0 12 24 36 Radial Distance (inch) 9 kips Control Geotextile Base (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 -5 0 5 10 15 0 12 24 36 Radial Distance (inch) 12 kips Control Geogrid Base (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 -5 0 5 10 15 0 12 24 36 Radial Distance (inch) 12 kips Control Geogrid Base (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 -5 0 5 10 15 0 12 24 36 Radial Distance (inch) 16 kips Control Geogrid Base (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward -10 -5 0 5 10 15 0 12 24 36 Radial Distance (inch) 16 kips Control Geogrid Base (Below) Ho riz on tal D isp lac em en t ( mi ls) Outward Inward

65 Figure 4.40. Horizontal Slippage of the Geosynthetic and Adjacent Unbound Material in Experiments 4 and 6 for Various Load Levels—Flexible Pavements Rigid Pavement Similar to the flexible pavement results section, a recap of pertinent key factors that had significant influence on the measured data is first provided below. Such information was very useful when navigating through the collected data and during the interpretation process of the various results.  Experiments 7, 9, and 10 included a 6-inch PCC layer on top of an 8-inch CAB layer and represent, respectively, the testing for the control (i.e., no base reinforcement), geogrid-reinforced base, and geotextile-reinforced base. Both the geogrid and geotextile were located at the middle of the CAB layer (i.e., 10 inches below the pavement surface).  With the focus given to the edge of the PCC slab, the surface layer was constructed in two segments with a gap of 6 inches in between, and the loading was applied near the edge of one of the slabs (referred to as loaded slab). Various instrumentations (earth pressure cells, strain gauges, and accelerometers) were positioned in an attempt to characterize many aspects of pavement responses at the edge of the slab.  Since pumping of the CAB layer through the joints was an important design concern, especially in the presence of moisture, load pulses resumed after introducing additional moisture to the CAB (referred to as wet experiments) and after the tests with dry CAB (i.e., at the optimum moisture content) were completed. The purpose of this activity was to investigate the roles of geosynthetics and excess moisture on pumping of the CAB at the joints.  The presence of the gap between the loaded and unloaded slabs along with the application of the load at the edge of the loaded slab represented a complex loading condition. This loading condition was non-axisymmetric and was coupled with the loaded slab independently undergoing possible rocking because of its high stiffness around the axis that was parallel to the edge of the slab (along the gap direction). Therefore, the responses under such circumstances were expected to be difficult to illustrate. -8 -6 -4 -2 0 2 4 6 8 0 12 24 36 Radial Distance (inch) 9 kips 12 kips 16 kips Ho riz on tal D isp lac em en t D iffe re nc e ( mi ls) ሺD isp ୋୣ ୭୥ ୰୧ୢ െD isp ୆ୟ ୱୣ ሺ୆ ୣ୪୭ ୵ሻ ሻ -8 -6 -4 -2 0 2 4 6 8 0 12 24 36 Ax is Tit le Radial Distance (inch) 9 kips 12 kips 16 kips Ho riz on tal D isp lac em en t D iffe re nc e ( mi ls) ሺD isp ୋୣ ୭୲ୣ ୶୲୧ ୪ୣ െD isp ୆ୟ ୱୣ ሺ୆ ୣ୪୭ ୵ሻ ሻ

66  Unlike with the tests on flexible pavements, a number of instruments malfunctioned. This problem limited the scope of the data interpretation described below.  The presentation of the results of the stress distribution across the geosynthetic was based on earth pressure cell data measured above and below the geosynthetic. The locations of the earth pressure cells were 2 inches above and below the geosynthetic.  Unlike earth pressure cells, measurements made by the accelerometers, strain gauges, and LVDTs could be considered as 1-point measurements. The majority of the pressure cells used in the LST experiments were 4 inches in diameter (a couple of the pressure cells were 1 inch in diameter and were mainly used for measurements of the horizontal pressure), and the fluid present within the flexible diaphragm indicated the average induced pressure within the entire surface area of the cell.  The earth pressure cells used to measure horizontal stresses in the CAB were located laterally away from the axis of loading and at approximately 8 inches from the centerline (under the edge of the plate) in all experimental tests.  The PCC layer of 6 inches was targeted for all the LST experiments; measurements from the core specimens showed consistently uniform thickness for the surface layer. Stress Distributions Across the Reinforcements First, the collected earth pressure cell data were reviewed. Figure 4.41 (section through the Y-axis) and Figure 4.42 (plan view at a depth of 8 inches) present the abridged versions of the LST configuration showing only the earth pressure cells in the CAB layer for the rigid pavement in Experiments 7, 9, and 10. The locations of the earth pressure cells are provided with modified identifications to facilitate the interpretation of the results. For example, the subscripts “B-Above” and “B-Below” refer to the pressure cells in the base above and below the geosynthetic, respectively. The superscript “Loaded-Center” refers to the pressure cells at the centerline of the load on the loaded side of the concrete slab, while “YD” represents the Y-direction. The horizontal stress is referred to as B-Horiz and was measured at only one location in the CAB layer (above the geosynthetic location and in the X-direction). A review of the assembled results with a focus on assessing the influence of the reinforcement indicated many noteworthy observations. A summary of the important results and interpretations is presented below. Figure 4.43a and Figure 4.43c show the load-induced vertical stress measurements along the centerline of the load for above and below the reinforcement location, respectively. Figure 4.43e shows the difference in the vertical stresses above and below the reinforcement location. The results for the CAB at the optimum moisture content (referred to as dry CAB) and partially saturated (referred to as wet CAB) are presented side by side on the left and right of the figure, respectively. The corresponding figures for the wet CAB are Figure 4.43b, Figure 4.43d, and Figure 4.43f, respectively. As expected, the vertical stresses above and below the reinforcement location consistently increased with the increase in the applied load level. As noted earlier, the form of the relationship between the vertical stress and the load level could be used to investigate the extent of the nonlinearity in the unbound materials. The vertical stresses for the dry tests above the reinforcement location were highest in the control experiment, followed by the geotextile-reinforced base and then the geogrid-reinforced

67 base. In contrast, a reverse trend was observed for the vertical stresses below the reinforcement location. A substantial reduction in the vertical stress (from above to below the reinforcement location) was observed in the control experiment, while the lowest reduction was observed across the geogrid, thus revealing the impact of the geogrid on the stress transfer across the reinforcement. In all cases, the vertical stresses in the wet CAB were similar to those in the dry CAB, indicating a very limited influence on the level of moisture introduced to the CAB during testing. Data were available in the CAB at two similar locations to assess load transfer across the concrete joint (i.e., the 6-inch gap/discontinuity in the concrete slab). These locations were below the reinforcement location at a depth of 12 inches from the surface and at 8 inches on the opposite sides of the centerline. Figure 4.44 presents the load-induced vertical stresses in the CAB layer at these specific locations for all load levels. It should be noted that some data were not available because of instrumentation malfunction. As expected, much higher vertical stresses were induced in the location under the loaded slab when compared with the results from beneath the unloaded slab. The stresses in the case of the control experiment (i.e., no reinforcement) were consistently in between those measured with the geogrid- and geotextile-reinforced base. The highest stresses were observed within the geogrid-reinforced base layer. Similar to the stress transfer across the geosynthetic, only a very limited influence was seen with the introduction of the specific moisture level to the CAB layer. Vertical and horizontal stresses in the CAB layer were also measured at the edge of the loading plate in the X-direction (parallel to the edge of the slab) and above the geosynthetic location (see Figure 4.45). The data indicated higher vertical stresses in the geotextile than in the CAB without any reinforcement (i.e., control experiment). Though the horizontal stresses were lower than the vertical stresses in all cases, significantly lower horizontal stresses were measured in experiments with a geotextile-reinforced base. The difference in behavior when additional moisture was introduced to the CAB layer was noticeable in the experiment with the geotextile.

Figure Figure Only E 4.41. LST C Only 4.42. LST C arth Press onfiguratio Earth Pres onfiguratio ure Cells ac n for Rigid sure Cells n for Rigid ross Geosy 68 Pavement across Geos Pavement nthetic—Pl Surface s (Experim ynthetic—P s (Experim an View at ents 7, 9, an rofile View ents 7, 9, an 8 inches be d 10) Show d 10) Show low Paveme ing ing nt

69 (a) (b) (c) (d) (e) (f) Figure 4.43. Vertical Stresses at the Centerline of the Loading Plate for Rigid Pavements (Experiments 7, 9, and 10—Dry and Wet) 13.5 16.4 20.8 6.9 9.1 12.110.8 14.0 17.6 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ሺps iሻ 11.0 14.7 18.8 6.9 9.3 12.2 10.2 12.8 19.0 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (wet) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ሺps iሻ 0.05 0.06 0.07 1.8 2.3 3.6 1.9 2.8 4.0 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ሺps iሻ 0.04 0.06 0.39 1.8 2.4 3.6 1.8 2.6 4.3 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ሺps iሻ 13.5 16.3 20.7 5.1 6.8 8.58.9 11.1 13.6 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextileሾσ ୆ ି୅ ୠ୭ ୴ୣ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ‐σ ୆ ି୆ ୣ୪୭ ୵ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ሿ ሺp siሻ 10.9 14.6 18.5 5.0 6.9 8.78.4 10.2 14.7 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextileሾσ ୆ ି୅ ୠ୭ ୴ୣ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ‐σ ୆ ି୆ ୣ୪୭ ୵ ୐୭ ୟୢ ୣୢ ିେ ୣ୬ ୲ୣ୰ ሿ ሺp siሻ

70 (a) (b) (c) (d) (e) (f) Figure 4.44. Vertical Stresses at Two Similar Locations in the CAB across the Joint and at 8 inches from the Centerline of the Loading Plate for Rigid Pavements (Experiments 7, 9, and 10—Dry and Wet) 10.1 13.7 19.4 15.6 18.7 24.2 4.9 6.6 7.9 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୐୭ ୟୢ ୣୢ ିଢ଼ ୈ ሺp siሻ 10.2 13.6 18.1 15.6 19.0 23.9 4.5 5.9 8.7 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୐୭ ୟୢ ୣୢ ିଢ଼ ୈ ሺp siሻ 3.7 4.7 6.5 0.7 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୙୬ ୪୭ୟ ୢୣ ୢି ଢ଼ୈ ሺps iሻ 3.8 5.0 6.6 0.3 0.5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୆ୣ ୪୭୵ ୙୬ ୪୭ୟ ୢୣ ୢି ଢ଼ୈ ሺps iሻ 6.4 9.0 13.0 7.1 -5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile ሾσ ୆ ି୆ ୣ୪୭ ୵ ୐୭ ୟୢ ୣୢ ିଢ଼ ୈ ‐σ ୆ି ୆ୣ ୪୭୵ ୙୬ ୪୭ୟ ୢୣ ୢି ଢ଼ୈ ሿ ሺp siሻ 6.3 8.7 11.5 4.2 5.4 -5 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextile [σ ୆ ି୆ ୣ୪୭ ୵ ୐୭ ୟୢ ୣୢ ିଢ଼ ୈ ‐σ ୆ି ୆ୣ ୪୭୵ ୙୬ ୪୭ୟ ୢୣ ୢି ଢ଼ୈ ሿ (p siሻ

71 (a) (b) (c) (d) Figure 4.45. Vertical and Horizontal Stresses at the Edge of the Loading Plate in the X-direction (Parallel to the Edge of the PCC Slab) for Rigid Pavements (Experiments 7, 9, and 10—Dry and Wet) Strain Measurements in Geosynthetic The normal strain measurements in the geosynthetics were made in two radial directions: X (parallel to the edge of the slab) and Y (perpendicular to the edge of the slab). Figure 4.46 (elevation view) shows the strain gauges SG1 through SG3, which measured the strains in the Y-direction. Figure 4.47 is a plan view at a depth of 10 inches below the pavement surface and shows the remaining strain gauges (SG4 and SG5), which measured normal strains in the X-direction. 4.0 4.5 6.87.2 11.1 12.9 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ ୐୭ ୟୢ ୣୢ ିଡ଼ ୈ ( ps i) 4.1 4.9 6.65.8 8.3 16.4 0 5 10 15 20 25 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextile Ve rti cal St res s,σ ୆ି ୅ୠ ୭୴ ୣ ୐୭ ୟୢ ୣୢ ିଡ଼ ୈ ( ps i) 1.7 1.8 2.1 2.4 2.7 3.5 1.0 1.4 1.5 0 1 2 3 4 5 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Dry) Control Geogrid Geotextile Ho riz on tal St res s,σ ୆ି ୌ୭ ୰୧୸ ሺps iሻ 1.7 1.8 2.0 2.4 2.8 3.3 1.0 1.2 1.8 0 1 2 3 4 5 9 kips 12 kips 16 kips Target Load Level Experiment No. 7, 9 and 10 (Wet) Control Geogrid Geotextile Ho riz on tal St res s,σ ୆ି ୌ୭ ୰୧୸ ሺps iሻ

Figure Figure Only S F wet CAB 4.46. LST C 4.47. LST C train Gaug igure 4.48 sh as a functio onfiguratio Only Strai onfiguratio es on Geos ows the str n of the rad n for Rigid n Gauges on n for Rigid ynthetic—P ain measure ial strain me 72 Pavement Geosynth Pavement lan View a ments obtain asured from s (Experim etic—Profil s (Experim t 10 inches ed for all lo the centerl ents 7, 9, an e View ents 7, 9, an below Pave aded levels ine of the lo d 10) Show d 10) Show ment Surfa with dry an ad. It should ing ing ce d be

73 noted that the measurements on the geotextile were not available due to malfunction of the strain gauges. The normal strains on the geogrid in the X-direction were consistently higher under the edge of the loading plate (i.e., at a radial distance of 12 inches in Figure 4.48a and Figure 4.48b) for all load levels. Compressive strains were indicated at 24 inches away from the centerline of the load, which is the farthest point where measurements were made. The strains in the Y-direction, measured by SG1 through SG3, are given in Figure 4.48c and Figure 4.48d. No substantial difference was observed between the dry and wet tests for the strain measurements below the edge of the loading plate for the various load levels. In general, for both dry and wet testing conditions, the strains in the X-direction at 12 and 24 inches from the centerline of the load were observed to be higher than those in the Y-direction at the same distance from the load. Furthermore, while tensile strains were observed in the X-direction, compressive strains were measured in the Y-direction at the 12- and 24-inch radial distances. (a) (b) (c) (d) Figure 4.48. Horizontal Strains in the Geogrid Reinforcement (Experiment 9)—Rigid Pavement (Dry and Wet) -200 0 200 400 600 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain, X-Direction (Dry) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -200 0 200 400 600 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain, X-Direction (Wet) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -200 0 200 400 600 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain, Y-Direction (Dry) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s) -200 0 200 400 600 0 12 24 Radial Distance (inch) Geogrid Horizontal Strain, Y-Direction (Wet) 9 kips 12 kips 16 kips Ho riz on tal S tra in (m icr on s)

74 Concrete strain gauges embedded at the bottom of the PCC slab were used to assess the normal strains developed due to surface loading. A positive strain reflected a tensile response associated with a beam action, while a negative strain reflected a compression response due to reversed beam action. Figure 4.49 shows the measured normal strains in the X-direction at the bottom of the PCC slab under the center of the loading plate for both dry and wet conditions. The data obtained from the embedded gauges in the dry structure indicated that adding a geogrid layer to the pavement structure increased the tensile strain when compared to a structure without any geosynthetic reinforcement (i.e., control). This behavior was reversed in the wet structure, reflecting lower tensile strains due to the presence of the geogrid layer. Meanwhile, adding a geotextile layer reversed the tensile strain to a compressive strain, in the dry as well as the wet structure, indicating a reversed beam action. Since these data were not expected and were somewhat controversial, care should be taken when interpreting the concrete strain gauge results. There was no evidence that the captured data were indeed a behavior of the reinforced structure and not an error due to installation. It was believed that a couple of possible installation scenarios might have caused the reversed beam action in Experiment 10. The first scenario was if the gauge accidentally rested on a piece of aggregate of relatively significant size within the concrete mixture, which happened to be in the middle of the gauge. This aggregate would have forced the gauge to arch upward when the pavement was loaded, especially due to the discontinuity in the PCC slab. A second scenario was if the gauge moved during concrete placement and changed its orientation or levelness. It should be noted that these scenarios were just speculation, and no solid data were available to prove or disprove them. (a) (b) Figure 4.49. Tensile Strains in the X-direction (Parallel to the Edge of the PCC Slab) at the Centerline of the Load and at the Bottom of the PCC Layer (Experiments 7, 9, and 10) PCC-CAB Interface Slippage While the technique of double integrating the acceleration records collected from paired accelerometers embedded at the same location in two different materials produced good results in the flexible pavement experiments, it was not as successful in the rigid pavement experiments when used to assess the slippage at the interface between the bottom of the slab and the -400 -200 0 200 400 600 800 1000 1200 5 10 15 20 Target Load Level (Kips) Tensile Strain at the bottom of the PCC layer (Dry) Control Geogrid Geotextile Te ns ile S tra in (m icr on s) -400 -200 0 200 400 600 800 1000 1200 5 10 15 20 Target Load Level (Kips) Tensile Strain at the bottom of the PCC layer (Wet) Control Geogrid Geotextile Te ns ile S tra in (m icr on s)

75 supporting base layer. The data collected were not reliable to carry out the integration algorithm, which might be attributed to difficulties associated with the installation of the accelerometers. Accordingly, two LVDTs were added in each experiment to provide a direct measurement of such potential slippage. These LVDTs were placed in the 6-inch gap between the two slabs and were attached in a way that allowed measuring the relative movement between the PCC and the CAB. This was achieved by attaching the LVDT body to the PCC while securing the tip of its shaft to a fixed point in the CAB. Figure 4.50 (elevation view) and Figure 4.51 (plan view) show the locations of the LVDTs at 12 and 24 inches from the centerline of the load along the edge of the loaded slab. While one of the two LVDTs (L7 at 24 inches from the centerline of the load) did not consistently produce dependable data due to loss of anchor of the LVDT shaft tip, good- quality data were produced with L7 and carefully examined. Figure 4.52 shows the PCC-CAB slippage measurements at the various load levels for both dry and wet testing for all experiments. Based on the data in Figure 4.52, it could be generally concluded that potential slippage between the PCC and the CAB occurred but at different rates and amplitudes. The negative sign associated with all the data collected from L6 indicated that the LVDT shaft moved inward toward the LVDT body. This finding showed that the relative movement between the CAB and the PCC was inward. In other words, the CAB was moving away from the center of the loading more than the PCC slab due to loading. The following specific observations were also made:  For the dry condition of the pavement structure without any geosynthetics, the same small amount of slippage was observed at the 9-kip and the 12-kip load level. However, this amount was slightly more than doubled when the load increased to 16 kip. In the case of the wet condition, almost the same moderate amount of slippage was observed at all load levels with a slight increase as the load increased.  For the dry and wet conditions of the pavement structure with geogrid-reinforced base, similar minor slippage was observed and remained almost unchanged as the load increased from 9 to 16 kip.  For the dry condition of the pavement structure with geotextile-reinforced base, significant slippage was observed and increased modestly as the load increased from 9 to 12 kip. Unfortunately, data at 16 kip were not available. In the case of the wet condition, minor slippage was observed at 9 kip, and the slippage increased significantly when the structure was subjected to 12 kip. Again, data at 16 kip were not available.  For the dry condition, the addition of the geogrid either kept the same magnitude of slippage or slightly reduced it when compared to the control experiment, while the addition of the geotextile significantly increased the magnitude.  For the wet condition, the addition of the geogrid reduced the slippage at all load levels when compared to the control experiment, while the addition of the geotextile reduced the slippage at 9 kip but significantly increased it at 12 kip.

Figure O Figure Only t 4.50. LST C nly the LV 4.51. LST C he LVDT U onfiguratio DT Used fo onfiguratio sed to Asse n for Rigid r Assessing n for Rigid ss Slippage Pavem 76 Pavement Slippage a Pavement at the PCC ent Surfac s (Experim t the PCC E s (Experim Edge—Pla e ents 7, 9, an dge—Prof ents 7, 9, an n View at 6 d 10) Show ile View d 10) Show inches bel ing ing ow

77 (a) (b) Figure 4.52. PCC-CAB Interface Slippage at the Edge of the PCC Slab (Experiments 7, 9, and 10): (a) Dry Condition; (b) Wet Condition Finite Element Modeling of Pavements with Geosynthetics The finite element models were developed using the software ABAQUS to simulate the LST test results (42). They were constructed for pavement structures with and without a geosynthetic layer in order to determine the critical responses of the pavement to different loading conditions. These pavement responses were used to predict the pavement performance. Figure 4.53 shows a typical geosynthetic-reinforced flexible pavement structure used in the LST test. It consisted of a 6-inch hot mix asphalt (HMA) layer, a 6-inch unbound aggregate base course, a 0.08-inch geosynthetic layer, and a subgrade. The geosynthetic layer was placed between the base course and the subgrade. The pavement structure was subjected to dynamic loading cycles with loading amplitudes of 9 kip, 12 kip, and 16 kip, respectively. The loading zone was applied with a circular loading foot with a radius of 6 inches. Figure 4.54 presents the finite element mesh of the geosynthetic-reinforced flexible pavement structure in ABAQUS. The cylindrical flexible pavement structure in the LST test was simplified as an axisymmetric model. Fine mesh was used near the load. The HMA layer, base course, and subgrade were represented as 8-node biquadratic homogeneous solid elements with reduced integration. The geosynthetic layer was represented by the 3-node quadratic membrane element. The interface between the geosynthetic layer and the aggregate/soil layer was characterized by the Goodman element (43). -20 -15 -10 -5 0 5 10 15 20 Target Load Level (kips) Control Geogrid Geotextile Ho riz on tal D isp lac em en t( mi ls) -20 -15 -10 -5 0 5 10 15 20 Target Load Level (kips) Control Geogrid Geotextile Ho riz on tal D isp lac em en t ( mi ls)

Figure Fig 4.53. Typic ure 4.54. M al Geosynt eshed Geo HMA L Base Co Subgr Geosynthet P hetic-Reinf synthetic-R ayer urse ade ic Layer 78 orced Flexi einforced P ble Paveme avement S 6 inch 6 inch 0.08 inc 4.5 ft nt Structur tructure in h Ge HM Ba S e in LST T ABAQUS osynthetic A Layer se Course ubgrade est

F the LST t 0.08-inch the base the rigid reinforce shows th Figur igure 4.55 p est. The stru geosynthet course. The pavement. S d rigid pave e meshed th e 4.55. Typ P resents a typ cture consi ic layer, and same loadin ince a 6-inc ment could ree-dimensi ical Geosyn Ba Geos CC Slab 6 inch ical geosyn sted of a 6-in a subgrade g configurat h gap existe not be simul onal geosyn thetic-Rein P se Course Subgrade ynthetic Lay P 79 thetic-reinfo ch PCC lay . The geosyn ion used wi d between t ated as an a thetic-reinfo forced Rig CC Slab er rced rigid p er, an 8-inc thetic layer th the flexib he two conc xisymmetric rced rigid p id Pavemen 6 i 8 i 4. avement stru h unbound g was placed le pavemen rete slabs, th structure. F avement mo t Structure nch nch 5 ft cture used ranular base in the cente t was applie e geosynthe igure 4.56 del in ABA in LST Te in , a r of d to tic- QUS. st

Figur Characte In was a lin nonlinear material. LST test modulus compress The nonl material material estimated e 4.56. Mes rization of M this study, ear elastic m cross-aniso Table 4.9 p and the corr test was use ive strength inear cross-a were determ was measure using the C (a) Side V hed Geosyn aterials U the HMA la aterial, the tropic elasti resents the s esponding i d to charact test was em nisotropic p ined using t d by the dir alifornia be iew thetic-Rein sed in LST yer was cha base layer w c material, a elected labo nput parame erize the vis ployed to e roperties of he rapid tria ect tension t aring ratio t 80 forced Rigi Test racterized a ith and with nd the subg ratory tests ters to the fi coelastic be stimate the e the geogrid xial test. Th est. The ela est (62). d Pavemen s a viscoelas out geosynt rade was as used to char nite elemen havior of as lastic modu -reinforced e tensile she stic modulu t Structure tic material hetic was de sumed to be acterize the t models. Th phalt concre lus of the PC and unreinfo et stiffness s of the subg (b) Top View in ABAQU , the PCC sl fined as a a linear ela materials in e dynamic te. The C slab (62 rced base of the geogr rade was S ab stic the ). id

81 Table 4.9. Selected Laboratory Tests for Material Characterization Material Type Constitutive Model Lab Test Model Input HMA Viscoelastic Dynamic modulus test Prony-series parameters (Gi, Ki, and τi), Poisson’s ratio PCC Elastic Compressive strength test Young’s modulus, Poisson’s ratio Base course Nonlinear cross-anisotropic Rapid triaxial test Inputs of the developed subroutine Geosynthetic Elastic Direct tension test Tensile sheet stiffness, Poisson’s ratio Subgrade Elastic CBR test Young’s modulus, Poisson’s ratio In the software ABAQUS, Prony-series models were used to characterize the time- dependent viscoelastic behavior of the hot mix asphalt, as shown in Equations 4.17 and 4.18.    /0 1 1 1         i n t i i G t G G e (4.17)    /0 1 1 1         i n t i i K t K K e (4.18) where G(t) and K(t) are relaxation shear modulus and bulk modulus; G0 and K0 are instantaneous shear modulus and bulk modulus; and Gi, Ki, and τi are the input coefficients. The method of fitting the Prony-series parameters with the dynamic modulus test result is as follows. The relaxation modulus of a linearly viscoelastic material can be expressed as: 1 ( ) i tn a a i i E t E E e            (4.19) where E(t) is the relaxation elastic modulus; and aE , a iE , and i are the regression coefficients in the model. Accordingly, the storage and loss moduli can be expressed by Equations 4.20 and 4.21. The magnitude of the dynamic modulus is given in Equation 4.22.   2 22 2 1 1 an a i i i i EE E         (4.20)   2 2 1 1 an i i i i EE      (4.21) * '2 ''2E E E  (4.22) where  'E  and  ''E  are the storage and loss modulus, respectively;  is the angular velocity; and *E is the magnitude of the dynamic modulus. By fitting the dynamic modulus test

82 result, the unknown parameters in Equation 4.19 could be determined based on the least square error criterion. As observed in Equations 4.17 and 4.19, the form of the Prony-series model in ABAQUS was slightly different from the model used for fitting the dynamic modulus test result. Parameter conversions between Equations 4.17 and 4.19 were required and are provided in Equations 4.23–4.27. Table 4.10 lists the determined Prony-series model coefficients used to characterize the asphalt concrete in ABAQUS. Figure 4.57 compares the fitted dynamic moduli with the measured ones, showing that the fitted dynamic moduli accurately matched the dynamic modulus test result. 0 1 n a a i i E E E    (4.23) 0 a i i EE E  (4.24)  00 2 1 EG   (4.25)  00 3 1 2 EK   (4.26) i i iG K E  (4.27) where E0 is the instantaneous elastic modulus, and ν is the Poisson’s ratio. Table 4.10. Determined Prony-Series Model Coefficients for the Plant-Mixed, Laboratory- Compacted (PMLC) Asphalt Concrete Prony-Series Coefficients i Gi Ki τi 1 0.362 0.362 4.09E-06 2 0.363 0.363 2.56E-04 3 0.1765 0.1765 7.71E-03 4 0.074 0.074 2.10E-01 5 0.0165 0.0165 3.88E+00 6 0.0057 0.0057 6.53E+01 Note: Elastic parameters: instantaneous modulus = 2630 ksi; Poisson’s ratio = 0.35.

83 Figure 4.57. Comparison between the Measured Dynamic Moduli and the Fitted Dynamic Moduli As stated in the previous section, the RaTT was employed to determine the cross- anisotropic properties of the UGM used in the LST test. The test data are given in Appendix L. The constitutive models of the UGM used in this study are shown in Equations 4.28 to 4.30 (22). 321 1 ( ) ( 1) kk oct y a a a IE k P P P   (4.28) x y En E  (4.29) xy y G m E  (4.30) where 1I is the first invariant of the stress tensor; oct is the octahedral shear stress; aP is the atmospheric pressure; 1k , 2k , and 3k are regression coefficients; xE is the horizontal resilient modulus; yE is the vertical resilient modulus; and xyG is the shear modulus in the x y plane. Table 4.11 presents the cross-anisotropic properties of the UGM determined in the LST test. Table 4.11. Cross-Anisotropic Properties of the UGM Used in LST Test Parameters k1 k2 k3 n m νxy νxx Determined Values 1545 0.75 −0.1 0.45 0.35 0.17 0.43 0 1 10 100 1,000 10,000 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2 1E+3 1E+4 1E+5 1E+6 1E+7 D yn am ic M od ul us |E *| , k si Reduced Frequency, Hz Measured |E*| Fitted |E*|

T products tensile fo for mach and geote direction the cross compared found tha manufact (a) T Fig he direct ten used in the rce and the ine direction xtile had sm . The ductili -machine dir to the data t all of the d urer’s speci ensile Test ure 4.58. Di sion tests w LST tests (s tensile strain . XMD is th aller sheet ty of geosyn ection. The in the manu etermined g fications. Setup for G rect Tensio ere conduct ee Figure 4. for the test e abbreviati stiffnesses in thetics in th sheet modu facturer’s sp eosynthetic eogrid n Test for D 84 ed to determ 58). Figure 4 ed geogrid a on for cross the machin e machine d li at 2 perce ecifications sheet modu (b) eterminin ine the shee .59 shows t nd geotexti -machine di e direction irection wa nt strain and , as shown li were high Tensile Tes g Sheet Stif t stiffness o he relations le. MD is th rection. Bot than in the c s much high 5 percent s in Table 4.1 er than the d t Setup for fness of Ge f geosynthet hips betwee e abbreviati h the geogri ross-machin er than that train were 2. Research ata in the Geotextile osynthetics ic n the on d e in ers

85 Figure 4.59. Relationships between Tensile Force and Tensile Strain for Geosynthetics Table 4.12. Comparison of Geosynthetic Sheet Stiffness Values between Laboratory Test and Manufacturer’s Specifications Geosynthetic Type Mechanical Properties—Test Mechanical Properties— Specification Sheet Stiffness @ 2% Strain (lb/in) Sheet Stiffness @ 5% Strain (lb/in) Sheet Stiffness @ 2% Strain (lb/in) Sheet Stiffness @ 5% Strain (lb/in) Geogrid MD Value 2650 1840 1713 1348 Geogrid XMD Value 3608 2563 2569 2232 Geotextile MD Value 3175 3589 NA NA Geotextile XMD Value 8419 7818 7505 NA Note: NA = not available. Development of Nonlinear Cross-Anisotropic User-Defined Material Subroutine A user-defined material (UMAT) subroutine was developed to characterize the nonlinear cross-anisotropic behavior of the UGM in the software ABAQUS. The UMAT subroutine adopted the direct secant modulus approach to determine the nonlinear resilient modulus solution in each iteration. The trial vertical modulus was computed using Equation 4.31 in each iteration (63). 0 1000 2000 3000 4000 5000 6000 7000 0 5 10 15 20 25 30 Te ns ile F or ce (l bs /ft ) Tensile Strain (%) Geogrid MD Geogrid XMD Geotextile MD Geotextile XMD

86   11i i iy y ycomputedE E E    (4.31) where iyE is the vertical modulus output from the i th iteration; 1iyE  is the vertical modulus output from the (i−1)th iteration;  is the damping factor; and iycomputedE is the vertical modulus computed in Equation 4.28 at the ith iteration. The convergence criteria are shown in Equations 4.32 and 4.33 (64). 1 2% i i y y i i y E E Error E   (4.32)     21 1 2 1 0.5% n i i y y i c n i y i E E Error E         (4.33) where iError is the individual error for each node; cError is the cumulative error for the entire model; and n is the number of nodes in the model. The secant modulus nonlinear solution technique was less complicated than the tangent stiffness approach, but it was adequate to provide good convergence of the iterations. Figure 4.60 shows the flowchart of the developed UMAT subroutine.

87 Figure 4.60. Flowchart of the Developed UMAT Subroutine Development of Goodman Model Friction Subroutine When surfaces of the geosynthetic and aggregate/soil were in contact, they usually transmitted shear and normal stresses across their interface. In this study, the interface element between the geosynthetic surface and the aggregate/soil surface was characterized using the Goodman model (4), which is shown in Equation 4.34. 0 0 s r n n r d k du d k dv                   (4.34) where  is the shear stress; n is the normal stress; ru is the relative shear displacement; r is the relative normal displacement; sk is the shear stiffness; and nk is the normal stiffness. The interface slippage condition was quantified by the shear stiffness, sk . If the geosynthetic- aggregate/soil interface was fully bonded, the shear stiffness was assigned a large value, for example, sk = 1×10 8 lb/in. If the slippage occurred at the geosynthetic-aggregate/soil interface, Define the constitutive model inputs Get initial stress state and incremental strains Compute stresses from current strains Compute vertical moduli from current stress state Check the convergence criteria Update the cross-anisotropic Jacobian matrix Update stresses and return to main program Adjust vertical moduli using damping factor

88 the shear stiffness, sk , was determined using the pullout test data. This tangential contact behavior was defined by the user subroutine FRIC in the ABAQUS software. Numerical Modeling Techniques for Geosynthetic-Reinforced Pavement Structures As mentioned previously, the reinforcement mechanisms of a geosynthetic included the lateral confinement and the vertical membrane effect. In ABAQUS, the vertical membrane effect was simulated by assigning the geosynthetic as a membrane element. However, the numerical model could not directly characterize the lateral confinement, which effectively reinforced the base material. In the numerical model, the lateral confinement of a geosynthetic was equivalent to an additional confining stress distributed in the geosynthetic influence zone, which affected the resilient modulus of the base course. For the sake of simplicity, the lateral confinement was simulated in this study by assigning the geosynthetic-reinforced base material a higher modulus value. Figure 4.61 illustrates the schematic plot to simulate the lateral confinement in the geosynthetic-reinforced pavement structure. As shown in Figure 4.61a, the shaded area is an influence zone. Previous studies reported that the influence zone ranged from 4 to 6 inches (18, 19, 65). The range of influence zone was herein assumed to be 6 inches in height when the geosynthetic was placed in the middle of the base course. In this range, the geosynthetic- reinforced base material had a higher modulus than the unreinforced material. The analytical models shown in Equations 4.13 and 4.14 were used to determine the modulus of the base material in the influence zone. The base material outside of the influence zone was considered an unreinforced material. The findings of the laboratory test evaluation indicated that placing the geosynthetic layer in the middle of the base material affected its horizontal and vertical modulus, while placing the geosynthetic layer at the bottom exerted a minor influence on the modulus of the base material. Therefore, no influence zone was assumed in the model when the geosynthetic was placed at the bottom of the base course, as shown in Figure 4.61b. This simulation represented the modulus of the base material to be the same as that of the unreinforced material.

89 Figure 4.61. Simulation of Lateral Confinement in Geosynthetic-Reinforced Pavement Structure Effect of Geosynthetic Reinforcement on Pavement Responses The current Pavement ME Design software predicts pavement performance based on the following computed critical responses by the embedded finite element program (62).  For flexible pavements: o Horizontal tensile strain at the bottom of the asphalt layer for fatigue cracking. o Vertical compressive strains within the asphalt layer, base course, and subgrade for rutting.  For rigid pavements: o Tensile bending stress at the bottom of the slab for bottom-up transverse cracking. o Tensile stress at the top of the slab for top-down transverse cracking. o Differential deflections across a joint for faulting. In this study, the critical responses of pavement with and without geosynthetics were computed based on the developed user subroutines and the proposed modeling techniques. The pavement models selected were similar to the pavement structures shown in Figures 4.53 and 4.55, respectively. The flexible pavement structures included a 6-inch base course with the geogrid or geotextile placed at the bottom, a 10-inch base course structure with the geogrid or geotextile placed in the middle, and the corresponding unreinforced structures. The rigid pavement structures included an 8-inch base course with the geogrid or geotextile placed in the center or at the bottom of the base course, and the corresponding control structure. Figures 4.62a and 4.62b compare the surface deflections of the geosynthetic-reinforced flexible pavement models with those of the unreinforced pavement models when they were HMA/PCC Base Course Subgrade Geosynthetic Layer 6 inches (a) Geosynthetic in the Middle of the Base Course HMA/PCC Base Course Subgrade Geosynthetic Layer (b) Geosynthetic at the Bottom of the Base Course

90 subjected to a 9-kip load on a circular area with the radius of 6 inches. The figure shows that placing the geogrid and geotextile at the bottom of the base course cannot reduce the surface deflections of the flexible pavement, while placing the geogrid in the middle of the base course only slightly decreases the surface deflections. The surface deflections of flexible pavement with the geotextile in the middle of the base layer are much larger than those of the unreinforced pavement. There are two reasons for this phenomenon. One is that placing the geotextile in the middle slightly reduces the vertical modulus of the base material. The other is that the slippage that occurs between the geotextile surface and the aggregate layer decreases the bonding coefficient of the geotextile-aggregate interface. (a) Surface Deflections of Flexible Pavement Structures with 6-inch Base Course (b) Surface Deflections of Flexible Pavement Structures with 10-inch Base Course Figure 4.62. Surface Deflections of Flexible Pavement Structures with and without Geosynthetic Distributions of the vertical stress beneath the load center within the flexible base layer are plotted in Figures 4.63a and 4.63b. The geogrid and geotextile-reinforced flexible pavement 0 0.005 0.01 0.015 0.02 0.025 0 10 20 30 40 50 Su rf ac e D ef le ct io n (in ch ) Distance Away from the Load Center (inch) Control-6 inch Base Geogrid-Bottom Geotextile-Bottom 0 0.005 0.01 0.015 0.02 0.025 0.03 0 10 20 30 40 50 Su rf ac e D ef le ct io n (in ch ) Distance Away from the Load Center (inch) Control-10 inch Base Geogrid-Middle Geotextile-Middle

91 structures diminished the vertical compressive stresses within the base layer by 2–3 psi. The decrease of vertical compressive stresses within the base layer was beneficial for reducing the permanent deformation of the base materials. (a) Vertical Stress Distribution within 6-inch Base Layer (b) Vertical Stress Distribution within 10-inch Base Layer Figure 4.63. Vertical Stress Distribution within Geosynthetic-Reinforced and Unreinforced Flexible Base Layer In addition, Table 4.13 presents the computed critical strains in the geosynthetic- reinforced and unreinforced flexible pavements. As the table shows, the geosynthetic reinforcement did not reduce the tensile strain at the bottom of the asphalt layer. This finding indicated that the application of a geosynthetic did not significantly influence the fatigue cracking performance of flexible pavements. In contrast, placing the geogrid and geotextile in the middle of the base course could decrease the average compressive strain in the unreinforced base layer by 17 percent and 10 percent, respectively. The geogrid and geotextile reinforced at 0 1 2 3 4 5 6 0 5 10 15 20 D ep th o f B as e L ay er (i nc h) Vertical Compressive Stress Distribution (psi) Control-6 inch Base Geogrid-Bottom Geotextile-Bottom 0 2 4 6 8 10 0 5 10 15 20 D ep th o f B as e L ay er (m ) Vertical Compressive Stress Distribution (psi) Control-10 inch Base Geogrid-Middle Geotextile-Middle

92 the bottom of the base course did not affect the compressive strain in the base layer, but they both diminished the compressive strain at the top of the subgrade significantly. This finding demonstrated that a geosynthetic reinforced in the middle of the base course reduced the permanent deformation of the base layer, while a geosynthetic reinforced at the bottom of the base course helped decrease the permanent deformation of the subgrade. Table 4.13. Computed Critical Strains for Geosynthetic-Reinforced and Unreinforced Flexible Pavement Structures Pavement Structure Tensile Strain at the Bottom of AC (με) Average Compressive Strain in Base Layer (με) Compressive Strain at Top of Subgrade (με) Control—6-inch Base 214 719 677 Geogrid—Bottom 231 735 421 Geotextile—Bottom 243 681 353 Control—10-inch Base 215 645 516 Geogrid—Middle 205 539 462 Geotextile—Middle 255 579 524 Figure 4.64 presents the load-induced tensile bending stresses at the bottom of the PCC slab for the geosynthetic-reinforced and unreinforced rigid pavements. As the figure shows, the geosynthetics slightly reduced the tensile bending stress at the bottom of the PCC slab. The tensile bending stress at the bottom of the PCC slab was not sensitive to the location and the type of geosynthetic. Figure 4.64. Tensile Bending Stresses at the Bottom of the PCC Slab for the Geosynthetic- Reinforced and Unreinforced Rigid Pavements 0 50 100 150 200 250 Control Geogrid-Middle Geotextile-Middle Geogrid-Bottom Geotextile-Bottom B en di ng S tr es s a t t he B ot to m o f P C C (p si )

93 Figure 4.65 shows the load-induced tensile stress at the top of the PCC slab for the geosynthetic-reinforced and unreinforced rigid pavements. There was no significant difference observed among the geosynthetic-reinforced and unreinforced rigid pavements. This finding indicated that the influence of geosynthetic type and geosynthetic location on the tensile stress at the top of the PCC slab is negligible. Figure 4.65. Tensile Stresses at the Top of the PCC Slab for the Geosynthetic-Reinforced and Unreinforced Rigid Pavements Parametric Study of Material Properties on Pavement Performance The sensitivity analysis of the pavement responses predicted by the finite element model was conducted by varying the material properties, such as the subgrade modulus and the geosynthetic sheet stiffness, and the thickness of the base course. Researchers found that the primary advantage of geosynthetic reinforcement was the reduction of the vertical compressive strain in the base course and at the top of the subgrade. Therefore, the pavement responses studied in the sensitivity analysis specifically referred to these two critical strains. The unreinforced pavement structure consisted of a 4-inch HMA layer and a 6-inch base course; the subgrade was analyzed as the control group, which was reinforced by a geosynthetic (geogrid or geotextile) placed in the middle or at the bottom of the base course. Figures 4.66a and 4.66b show the sensitivity of the model-predicted pavement responses to the variations in the subgrade modulus. The selected subgrade moduli were 5 ksi, 15 ksi, and 25 ksi, which represented the poor, fair, and good quality of the subgrade, respectively. The increase in subgrade modulus remarkably decreased the vertical strain at the top of the subgrade but slightly increased the vertical strain within the base layer. The placement of the geosynthetic was effective at reducing these two critical strains. The reduction of the critical strains due to the geosynthetic reinforcement was normalized using Equation 4.35. 0 10 20 30 40 50 Control Geogrid-Middle Geotextile-Middle Geogrid-Bottom Geotextile-Bottom T en si le S tr es s a t t he T op o f P C C (p si )

94 _ _ 100% _ Strain Control Strain Geosynthetic Normalized reduction of strain Strain Control   (4.35) where _Strain Control is the computed critical strain in the control model; and _Strain Geosynthetic is the computed critical strain in the geosynthetic-reinforced model. Figure 4.66c indicates that the reduction of the vertical strain at the top of the subgrade was significant when the geogrid or geotextile was placed at the bottom of the base course. The increase in subgrade modulus did not influence the normalized reduction of the subgrade vertical strain due to the presence of the geotextile but slightly decreased the reduction percentage due to the presence of the geogrid. Figure 4.66d illustrates that the geosynthetic reinforced in the middle of the base course effectively reduced the vertical strain, while the geosynthetic located at the bottom of the base course slightly increased the base vertical strain. With the increase in the subgrade modulus, the normalized reduction of the base vertical strain due to the geosynthetic decreased by approximately 5 percent. This finding indicated that geosynthetic reinforcement was more effective when it was placed over a weaker subgrade, which normally had a lower resilient modulus. Figures 4.67a and 4.67b show the sensitivity of the pavement responses predicted by the model to the variation of the geosynthetic sheet stiffness. Both the vertical strain at the top of the subgrade and the average vertical strain within the base layer decreased with the geosynthetic sheet stiffness. This finding indicated that the geosynthetic with a higher sheet stiffness was more efficient at reducing the permanent deformation of the pavement structure. Figure 4.68 indicates that the developed geosynthetic-reinforced and unreinforced pavement models were also sensitive to the thickness of the base course in predicting the vertical strains in the base layer and the subgrade. The figure shows that increasing the thickness of the base course reduced both the vertical compressive strain at the top of the subgrade and the vertical strain within the base course. The geosynthetic reinforcement was more effective for a thin base layer in terms of the percent reduction of vertical strains in the base and subgrade.

95 (a) Computed Vertical Strain at the Top of Subgrade (b) Computed Average Vertical Strain within Base Course Figure 4.66. Sensitivity of Model-Predicted Pavement Responses to Subgrade Modulus 0 400 800 1200 1600 2000 0 5 10 15 20 25 30 V er tic al S tr ai n at th e T op o f Su bg ra de (μ ε) Subgrade Modulus (ksi) Control GG-Middle GG-Bottom GT-Middle GT-Bottom 0 400 800 1200 1600 0 5 10 15 20 25 30A ve ra ge V er tic al S tr ai n w ith in B as e C ou rs e (με ) Subgrade Modulus (ksi) Control GG-Middle GG-Bottom GT-Middle GT-Bottom

96 (c) Normalized Reduction of Vertical Strain at the Top of Subgrade (d) Normalized Reduction of Vertical Strain within Base Course Figure 4.66. Sensitivity of Model-Predicted Pavement Responses to Subgrade Modulus (Continued) 0 20 40 60 SG Modulus=5 ksi SG Modulus= 15 ksi SG Modulus=25 ksi N or m al iz ed R ed uc tio n of V er tic al St ra in a t T op o f S ub gr ad e (% ) GG-Middle GT-Middle GG-Bottom GT-Bottom -10 -5 0 5 10 15 SG Modulus=5 ksi SG Modulus= 15 ksi SG Modulus=25 ksi N or m al iz ed R ed uc tio n of V er tic al St ra in w ith in B as e C ou rs e (% ) GG-Middle GT-Middle GG-Bottom GT-Bottom

97 (a) Vertical Strain at the Top of Subgrade (b) Average Vertical Strain within the Base Course Figure 4.67. Sensitivity of Model-Predicted Pavement Responses to Geosynthetic Sheet Stiffness 300 400 500 600 700 800 1000 1500 2000 2500 3000 3500 V er tic al S tr ai n at th e T op o f Su bg ra de (μ ε) Geosynthetic Sheet Stiffness (lb/in) GG-Middle GG-Bottom 600 700 800 900 1000 1100 1000 1500 2000 2500 3000 3500A ve ra ge V er tic al S tr ai n w ith in B as e C ou rs e (με ) Geosynthetic Sheet Stiffness (lb/in) GG-Middle GG-Bottom

98 (a) Computed Vertical Strain at the Top of Subgrade (b) Computed Average Vertical Strain within Base Course Figure 4.68. Sensitivity of Model-Predicted Pavement Responses to Thickness of Base Course Comparison of Finite Element Simulations with LST Measurements The finite element simulation results of the developed geosynthetic-reinforced and unreinforced pavement models were validated by comparing them to the LST test measurements in terms of the surface deflection, tensile strain at the bottom of the asphalt layer, and vertical pressures within the base and subgrade layers. The detailed comparison processes are presented in Appendix M. Figures 4.69 and 4.70 illustrate the location of the instruments, such as the LVDTs, the tensile strain gauge, and the pressure sensors, in the flexible and rigid pavement structures, respectively. 0 400 800 1200 1600 2000 2400 2800 6 8 10 12 14 16V er tic al S tr ai n at th e T op o f Su bg ra de (μ ε) Thickness of Base Course (inch) Control GG-Middle GG-Bottom GT-Middle GT-Bottom 0 400 800 1200 1600 6 8 10 12 14 16A ve ra ge V er tic al S tr ai n w ith in B as e C ou rs e (με ) Thickness of Base Course (inch) Control GG-Middle GG-Bottom GT-Middle GT-Bottom

Figure 4 (a) Flexibl (b) Flexible .69. Locatio e Pavement Pavement n of Instru 99 with a 6-in with a 10-in ments in Fl ch Base Co ch Base Co exible Pave urse urse ment Struc tures

F element m subjected LST mea deflection deflection indicated highly ac strain at t develope and unrei geotextil predicted flexible p element m explanati behavior shown in have bee Figure igure 4.71 sh odels and t to a 9-kip l surements f by LVDT at this loca that the dev curate when he bottom o d finite elem nforced pav e-reinforced vertical pre avements. M odels, exce ons existed of the subgr Figure 4.73 n due to arch 4.70. Locati ows the com he LST test oad. The mo rom LVDTs 4 and that p tion was too eloped geos predicting f the asphal ent models ement struc pavement s ssures withi ost of the m pt the meas for these dis ade was not b, the measu ing over th on of Instru parison of measureme del-predicte 1, 2, and 3. redicted by small to be ynthetic-rei the pavemen t concrete is accurately p tures but slig tructures. Fi n the base a easured pr urement of p crepancies. taken into a red pressur e sensor. 100 ments in R the surface nts when the d surface d The deviati the finite ele accurately nforced and t surface de plotted in F redicted the htly overes gure 4.73 pr nd subgrade essure value ressure cell For instance ccount (see e being low igid Pavem deflections p flexible pa eflections w on between ment model captured by unreinforce flections. T igure 4.72. tensile stra timated the esents a com layer and th s were captu s P1 and P7 , for sensor Figures 4.7 er than the p ent Structu redicted by vement stru ere in agree the measure existed bec the LVDT. d pavement he comparis The figure s in in the geo tensile strain parison be e measured red by the d . A number P1, the stre 3a and 4.73 redicted pre res the finite ctures were ment with th d surface ause the sur This finding models wer on of the ten hows that th grid-reinfor in the tween the results for t eveloped fi of possible ss-dependen b). For sens ssure might e face e sile e ced he nite t or P7,

101 (a) Pavement Structures with 6-inch Base Course (b) Pavement Structures with 10-inch Base Course Figure 4.71. Comparison of Measured and Predicted Surface Deflections for Pavement Structures with and without Geosynthetic LVDT 1 LVDT 2 LVDT 3 LVDT 4 0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.01 0.02 0.03 0.04 0.05 L ST M ea su re m en ts (i nc h) Finite Element Simulations (inch) Control Geogrid Geotextile Line of Equality ± 10 % Equality ± 20 % Equality LVDT 1 LVDT 2 LVDT 3 LVDT 4 0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.01 0.02 0.03 0.04 0.05 L ST M ea su re m en ts (i nc h) Finite Element Simulations (inch) Control Geogrid Geotextile Line of Equality ± 10 % Equality ± 20 % Equality

102 (a) Pavement Structures with 6-inch Base Course (b) Pavement Structures with 10-inch Base Course Figure 4.72. Comparison of Measured and Predicted Tensile Strains at the Bottom of Asphalt Layer for Pavement Structures with and without Geosynthetic 0 100 200 300 400 0 100 200 300 400 L ST M ea su re m en ts (μ ε) FE Simulations (με) Control Geogrid Geotextile Line of Equality + + 0 100 200 300 400 0 100 200 300 400 L ST M ea su re m en ts (μ ε) FE Simulations (με) Control Geogrid Geotextile Line of Equality + +

103 (a) Pavement Structures with 6-inch Base Course (b) Pavement Structures with 10-inch Base Course Figure 4.73. Comparison of Measured and Predicted Vertical Stresses within the Base and Subgrade for Pavement Structures with and without Geosynthetic Figure 4.74 compares the measured surface deflections of rigid pavements with those predicted by the finite element models. The figure shows that the surface deflections predicted by the finite element models were in agreement with the measurements of LVDTs 1 and 2 but did not match the measurements of LVDTs 3 and 4. The measured surface deflections from LVDTs 3 and 4 were negative, which indicated that the PCC slab moved upward at the far end near the tank wall. This movement might be because of the boundary conditions at the edge of the PCC slab. Figure 4.75 shows the comparison of the measured and predicted vertical pressures P5 P3 P2 P1 P4 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 L ST M ea su re m en ts (p si ) Finite Element Simulations (psi) Control Geogrid Geotextile Line of Equality ± 10 % Equality ± 20 % Equality P7 P4 P6 P3 P2P5 P1 P7 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 L ST M ea su re m en ts (p si ) Finite Element Simulations (psi) Control Geogrid Geotextile Line of Equality ± 10 % Equality ± 20 % Equality

104 in the base and subgrade. Note that some of the pressure sensor data (e.g., P1, P3, P6, and P8) were removed due to poor quality. The figure shows that the geogrid and geotextile effectively reduced the vertical compressive stresses in the base course when they were placed in the center of the base course but had a negligible influence on reducing the vertical compressive stresses in the subgrade. Figure 4.74. Comparison of Measured and Predicted Surface Deflections for Rigid Pavement Structures with and without Geosynthetic Figure 4.75. Comparison of Measured and Predicted Vertical Stresses within the Base and Subgrade for Rigid Pavement Structures with and without Geosynthetic In summary, the finite element simulation results were in good agreement with the LST test measurements for both the reinforced and unreinforced pavement structures. Consideration of the paving material characterization, the geosynthetic-aggregate/soil interface characterization, LVDT 1 LVDT 2 LVDT 3 LVDT 4 -0.03 -0.01 0.01 0.03 0.05 -0.03 -0.01 0.01 0.03 0.05 L ar ge -S ca le T an k M ea su re m en ts (in ch ) Finite Element Simulations (inch) Control Geogrid Geotextile Line of Equality P2 P4 P5 P7 0 4 8 12 16 20 0 4 8 12 16 20 L ar ge -S ca le T an k M ea su re m en ts (p si ) Finite Element Simulations (psi) Control Geogrid Geotextile Line of Equality ± 10 % Equality ± 20 % Equality

105 and the reinforcement influence zone was important to develop accurate numerical models of geosynthetic-reinforced pavement structures. ANN Approach for Predicting Pavement Performance The current Pavement ME Design software predicted pavement performance based on the computed critical pavement responses from a linear isotropic and layered elastic program. In other words, the determination of critical pavement responses was the key to forecasting pavement performance. The finite element models developed in this project were sufficiently accurate to compute the critical responses of geosynthetic-reinforced pavement structures. However, these models were developed using the software ABAQUS, which was not compatible with the Pavement ME Design embedded software DARWin-ME. Furthermore, replacing the current Pavement ME Design software with the developed finite element models to compute the critical responses of the arbitrary user-inputted geosynthetic-reinforced pavement structures was impractical. Therefore, there was a need to predict the responses of any given geosynthetic- reinforced pavement structure based on computation with the developed finite element models for a wide range of geosynthetic-reinforced pavement structures. To satisfy this need, the ANN approach was used in this study to predict the critical responses of geosynthetic-reinforced pavement structures. The ANN models allowed for establishing the correlations between the input variables, iX , and the output variables, jY , through the inter-connected neurons (i.e., weight factor, jiw ) (66). Note that the input variables, iX , and the output variables, jY , were usually normalized to ix and jy , respectively, and were values between 0 and 1. In this study, the output variables, jY , represented the computed critical pavement responses, including the tensile strain at the bottom of the asphalt concrete and the compressive strain within the asphalt concrete, base layer, and subgrade. The selection of the input parameters, iX , was based on the sensitivity analysis of the developed finite element models. The identified input parameters to the ANN models included the layer thickness, the modulus of the paving material, the location of the geosynthetic, and the type of geosynthetic. The correlations developed by the ANN models between the normalized input parameters, ix , and the normalized output variables, jy , are shown in Equation 4.36. 1 n j ji i i y f w x        (4.36) where f is a transfer function, which normally uses a sigmoidal, Gaussian, or threshold functional form; and jiw is the unknown weight factors. Developing a neural network model specifically requires the determination of the weight factors, jiw , as in Equation 4.36. The ANN model determined these weight factors, jiw , through two major functions: training and validating. The training dataset was used to determine the trial weight factors, jiw , and the validating dataset was employed to examine the accuracy of the model prediction. A robust ANN model

106 normally required a large database of input and output variables (67). Thus, generating the input and output variable database was the first step in developing an ANN model. Experimental Computational Plan for ANN Models To generate the database of the numerical model inputs and the corresponding computed critical pavement responses, researchers computed multiple cases based on the developed geosynthetic-reinforced and unreinforced finite element models. Tables 4.14 and 4.15 show the selected input parameters as well as their values for the geosynthetic-reinforced pavement structures and the corresponding unreinforced pavement structures, respectively. Based on these experimental computational plans, the number of computed geosynthetic-reinforced pavement models was 5,832, and the number of computed unreinforced pavement models was 486. As shown in Table 4.14, two geosynthetic types (geogrid and geotextile) and two geosynthetic locations (middle and bottom of base course) were taken into account in the computation of the multiple cases. The pavement response database was divided into five categories, including:  The geogrid placed in the middle of the base layer (GG-M).  The geogrid placed at the bottom of the base layer (GG-B).  The geotextile placed in the middle of the base layer (GT-M).  The geotextile placed at the bottom of the base layer (GT-B).  The unreinforced one (NG). Each category of pavement response corresponded to one set of neural network models. Table 4.14. Selected Input Parameters for Geosynthetic-Reinforced Pavement Structures Influential Factors Level Input Values Load Magnitude 1 9 kip HMA Thickness 3 2, 4, and 6 inches HMA Modulus 3 300, 450, and 600 ksi Base Thickness 3 6, 10, and 15 inches Base Vertical Modulus 3 20, 40, and 60 ksi Base Anisotropic Ratio 2 0.35 and 0.45 Geosynthetic Location 2 Middle and Bottom of Base Course Geosynthetic Type 2 Geogrid and Geotextile Geogrid Sheet Stiffness 3 1200, 2400, and 3600 lb/in Geotextile Sheet Stiffness 3 1800, 3600, and 5400 lb/in Subgrade Modulus 3 5, 15, and 25 ksi Note: The number of total cases was 5,832.

107 Table 4.15. Selected Input Parameters for Unreinforced Pavement Structures Influential Factors Level Input Values Load Magnitude 1 9 kip HMA Thickness 3 2, 4, and 6 inches HMA Modulus 3 300, 450, and 600 ksi Base Thickness 3 6, 10, and 15 inches Base Vertical Modulus 3 20, 40, and 60 ksi Base Anisotropic Ratio 2 0.35 and 0.45 Subgrade Modulus 3 5, 15, and 25 ksi Note: The number of total cases was 486. Development of ANN Models A three-layered neural network architecture consisting of one input layer, one hidden layer, and one output layer was constructed, as shown in Figure 4.76. The input parameters are listed in Tables 4.14 and 4.15, except the geosynthetic location and the geosynthetic type. The output variables were the critical pavement responses, including the tensile strain at the bottom of the asphalt concrete and the compressive strains within the asphalt concrete, base course, and subgrade. The hidden layer assigned 20 neurons to establish the connection between the output layer and the input layer. In this study, the transfer function used a sigmoidal functional form, which is shown in Equation 4.37 (68).     1 1 expi i f I I   (4.37) where iI is the input quantity; and  is a positive scaling constant, which controls the steepness between the two asymptotic values 0 and 1. The constructed neural network structure was programmed using the software MATLAB R2013a (69). The training algorithm used the Levenberg-Marquardt back propagation method to minimize the mean squared error (MSE) (70). The gradient descent weight function was employed as a learning algorithm to adjust the weight factors, jiw (71).

108 Figure 4.76. Illustration of Three-Layered Neural Network Architecture The pavement response database was first randomly divided into a training dataset and a validating dataset as the ratio of 80 percent and 20 percent, respectively. The training dataset was used to determine the weight factors, jiw , and the validating dataset was employed to examine the prediction accuracy of the developed neural network. Figures 4.77–4.81 show the comparisons between the pavement responses computed by the finite element model and the responses of the pavement with a GG-M structure predicted by the ANN model. As the figures illustrate, the ANN model predictions were in good agreement with the finite element model computational results. This finding indicated that the ANN models accurately predicted all of the pavement responses from the validating dataset after the training process. The comprehensive comparisons between the pavement responses computed by the finite element model and those predicted by the ANN model are presented in Appendix N. The developed ANN models could be used to interpolate the critical responses of any given geosynthetic-reinforced and unreinforced pavement structure. Input Layer Hidden Layers Output Layer x1 x2 xn yi x3 20 Neurons Sigmoidal Transfer Function Back Propagation Error

109 Figure 4.77. Comparison of Tensile Strain at the Bottom of the Asphalt Layer

110 Figure 4.78. Comparison of Average Vertical Strain in the Asphalt Layer

111 Figure 4.79. Comparison of Average Vertical Strain in the Base Layer

112 Figure 4.80. Comparison of Vertical Strain at the Top of the Subgrade

113 Figure 4.81. Comparison of Vertical Strain at 6 inches below the Top of the Subgrade Determination of Modified Material Properties The performance of geosynthetic-reinforced flexible pavements included fatigue cracking, permanent deformation, and international roughness index (IRI) (51). The aforementioned ANN- model-predicted critical pavement responses could be used to predict the pavement performance using the distress models in the current Pavement ME Design software. However, this method ignored the influence of traffic and climate on pavement performance. To eliminate this defect, the material properties of geosynthetic-reinforced pavement structures had to first be made equivalent to a combination of modified material properties (e.g., modified base modulus and modified subgrade modulus) of an unreinforced pavement structure. The determined modified material properties were then input into the Pavement ME Design software to predict the pavement performance. In this approach, the influence of traffic and climate on the pavement performance was taken into account by the Pavement ME Design software. Figure 4.82 presents a flowchart to determine the modified material properties for a geosynthetic-reinforced pavement structure. When the user input the geosynthetic-reinforced pavement structure information (e.g., layer thickness and material properties), the program would automatically generate a control structure with the same layer thickness and the equivalent material properties. The ANN models

114 were selected to predict the responses of the geosynthetic-reinforced and the control pavement structures. Subsequently, the responses of the geosynthetic-reinforced pavement structure were compared to those of the control structure. Equation 4.38 presents the convergence criterion used in this flowchart. 10%geosynthetic control control      (4.38) where geosynthetic represents the response of the geosynthetic-reinforced pavement structure; and control represents the response of the control structure. If the responses of the geosynthetic- reinforced pavement structure did not match those of the control structure, the material properties (i.e., base material and subgrade modulus) of the control structure would be modified. The iteration would end when the comparison of the critical responses passed the convergence criterion. The program would then output the modified material properties of the control structure, which were the inputs for the Pavement ME Design software. The program was written using C# language to be compatible with the current Pavement ME Design software.

115 Figure 4.82. Flowchart of the Process of Predicting Pavement Performance Input Geosynthetic- Reinforced Pavement Structure Data (Layer Information and Material Properties) Select ANN Models for Geosynthetic-Reinforced Pavement Predict Performance- Related Geosynthetic- Reinforced Pavement Generate the Control Pavement Structure (Layer Information and Material Properties) Select ANN Models for Control Pavement Predict Performance-Related Control Pavement Responses Do Geosynthetic-Reinforced Pavement Responses Match Control Pavement Responses? Determine Modified Material Properties Predict Pavement Performance Using Pavement ME Design Models Modify the Input Material Properties

116 Case studies were performed on flexible pavements with geosynthetics (i.e., geogrid or geotextile) placed in the middle or at the bottom of the base course. Figure 4.83 presents the geosynthetic-reinforced pavement structures analyzed in this study. The material properties of the geosynthetic-reinforced pavements are shown in Tables 4.16 and 4.17. The modified material properties of the control pavement structure were determined using the aforementioned approach, as presented in Table 4.18. As shown in Figure 4.83, placing the geogrid in the middle or at the bottom of the base course was equivalent to increasing the moduli of the base course and subgrade. Placing the geotextile at the bottom of the base course was comparable to increasing the subgrade modulus. It was also noteworthy that the geotextile placed in the middle of the base course could not reinforce the pavement structure but significantly reduced the base course modulus. The determined modified material properties could serve as the inputs of the Pavement ME Design software for predicting the performance of geosynthetic-reinforced pavement structures. Figure 4.83. Geosynthetic-Reinforced Pavement Structures for Case Studies HMA Layer Base Subgrade 4 inches 10 inches Semi- infinite Geosynthetic Layer HMA Layer Base Subgrade 4 inches 10 inches Semi- infinite Geosynthetic Layer a. Geosynthetic Placed at the Bottom of the Base Course b. Geosynthetic Placed at the Center of the Base Course

117 Table 4.16. Material Properties of Geosynthetic-Reinforced Pavements for Case Studies— Material Properties of Control Pavement Material Type Thickness (inch) Vertical Modulus (ksi) Anisotropic Ratio Poisson’s Ratio Asphalt Concrete 4 300 N/A 0.35 Base Course 10 30 0.4 0.4 Subgrade N/A 10 N/A 0.4 Table 4.17. Material Properties of Geosynthetic-Reinforced Pavements for Case Studies— Material Properties for Geosynthetic Products Material Type Location Thickness (inch) Sheet Stiffness (lb/in) Poisson’s Ratio Geogrid Middle of Base Course 0.08 2400 0.3 Geogrid Bottom of Base Course 0.08 2400 0.3 Geotextile Middle of Base Course 0.08 3600 0.3 Geotextile Bottom of Base Course 0.08 3600 0.3 Table 4.18. Determination of Modified Material Properties for Case Studies Material Type Location Modified Base Course Modulus (ksi) Modified Subgrade Modulus (ksi) Geogrid Middle of Base Course 32.1 15.1 Geogrid Bottom of Base Course 30.8 20.7 Geotextile Middle of Base Course 24.3 10.0 Geotextile Bottom of Base Course 30.0 18.7 Prediction of Pavement Performance The pavement structures shown in Figure 4.83 were assumed to be constructed in College Station, Texas. The two-way average annual daily truck traffic was 2,000. The vehicle class distribution and growth followed the default values. The climate information was collected from the weather station in College Station, Texas. The Pavement ME Design software was used to predict the performance of geosynthetic-reinforced and unreinforced pavements. Figures 4.84– 4.86 show the effect of geosynthetic type and geosynthetic location with the base course on the flexible pavement performance. As shown in Figure 4.84, the greatest amount of fatigue cracking

118 occurred with the geotextile placed in the center of the base course, and it was dramatically higher than that in the unreinforced pavement section. This finding indicated that placing the geotextile in the center of the base course significantly reduced the fatigue life of the pavement structures. Compared to other pavement structures, the least amount of fatigue cracking occurred with the geogrid at the bottom of the base course, though this section only slightly outperformed the unreinforced pavement section. As shown in Figure 4.85, the greatest amount of rutting still occurred with the geotextile in the center of the base course. Compared to the control pavement, placing the geogrid in the center or at the bottom of the base course, or placing the geotextile at the bottom of the base course, could effectively reduce the accumulated permanent deformation of the pavement structure. As Figure 4.86 illustrates, the pavement with the highest IRI was the one with the geotextile in the center of the base course. The lowest IRI was provided by the geogrid at the bottom of the base course. These example calculations indicated that the major benefit of geosynthetics on the performance of flexible pavements was derived from a reduction in rutting and roughness. The placement of a geogrid at the bottom of the base course would achieve the most beneficial effect. Figure 4.84. Effect of Geosynthetic Location and Geosynthetic Type on Fatigue Cracking 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 2.0 4.0 6.0 8.0 10.0 Fa tig ue C ra ck in g (% L an e A re a) Pavement Age (Years) Control GGM GGB GTM GTB

119 Figure 4.85. Effect of Geosynthetic Location and Geosynthetic Type on Rutting Depth Figure 4.86. Effect of Geosynthetic Location and Geosynthetic Type on IRI 0.0 0.2 0.4 0.6 0.8 1.0 0.0 2.0 4.0 6.0 8.0 10.0 Pe rm an en t D ef or m at io n (in ch ) Pavement Age (Years) Control GGM GGB GTM GTB 60 70 80 90 100 110 0.0 2.0 4.0 6.0 8.0 10.0In te rn at io na l R ou gh ne ss In de x (in /m i) Pavement Age (Years) Control GGM GGB GTM GTB

120 Validation of the Proposed ANN Approach Using the proposed ANN approach, a geosynthetic-reinforced pavement with any given material properties needed to be made equivalent to an unreinforced pavement with the modified material properties to obtain the identical pavement responses. The process of validating the proposed ANN approach is illustrated in Figure 4.87 and involved the following steps:  Identify the in-service geosynthetic-reinforced pavement sections from the LTPP database and Texas Pavement Management Information System (PMIS). This study focused on the in-service pavement sections with the placement of geosynthetics in conjunction with the unbound base courses.  Collect the pavement structure data, including layer thickness, construction dates, material design information, and falling weight deflectometer data.  Collect the traffic data from the identified pavement sections, which should be compatible with the input of the traffic module in the Pavement ME Design software.  Collect the climatic data or weather station information from the identified pavement sections.  Collect the performance data from the identified pavement sections, including fatigue cracking, rutting, and IRI.  Employ the proposed ANN approach to determine the modified material properties of an unreinforced pavement.  Input the unreinforced pavement structure data, the collected traffic data and climatic data, and the determined modified material properties into the Pavement ME Design software to predict the pavement performance (i.e., fatigue cracking, rutting, and IRI).  Compare the predicted pavement performance with that measured in the field.

121 Figure 4.87. Flowchart of the Process of Validating the Proposed ANN Approach After a thorough review of the in-service pavement sections in the LTPP database and PMIS, researchers found 74 pavement sections containing geosynthetics in the LTPP database and 51 pavement sections containing geosynthetics in the PMIS. A full list of the identified pavement sections is presented in Appendix P. Of these identified pavement sections, most had stabilized base courses, which were not under consideration in this study. Of the remaining pavement sections, five were from the LTPP database and five were from the PMIS. Researchers used Pavement Section 16-9032 from the LTPP database as an example to validate the proposed ANN approach. Section 16-9032 consisted of a 6-inch hot-mixed and Identify In-Service Geosynthetic-Reinforced Pavement Sections Collect Field Data Structure Data Material Information Traffic Data Performance Data Climate Data Determine Modified Material Properties Predict Pavement Performance Using Pavement ME Design Does Predicted Pavement Performance Match Field Measurement? Finish Yes Modify Proposed ANN Approach No

122 dense-graded asphalt concrete, a 23.2-inch crushed gravel unbound base, and a semi-infinite subgrade, which was classified as AASHTO 7-5 soil. A 0.1-inch woven geotextile was placed at the interface between the unbound base and subgrade. The comparisons of geosynthetic- reinforced pavement performance between the predictions from the proposed ANN approach and the field measurements are presented in Figures 4.88–4.90. The figures show that the predicted rutting depth and IRI results were in good agreement with the field measurements. The fatigue cracking of the geosynthetic-reinforced pavement was slightly overestimated by the proposed ANN approach. These findings indicated that the proposed ANN approach was capable of accurately predicting the performance of geosynthetic-reinforced pavements. Figures 4.88–4.90 also present the predicted performance of the control pavement and demonstrate that the geotextile placed at the base/subgrade interface had beneficial effects on reducing the rutting and IRI of flexible pavements. The comparisons of the geosynthetic-reinforced pavement performance between the ANN approach predictions and the field measurements for other identified pavement sections are detailed in Appendix O. Figure 4.88. Comparison of Rutting Depth between ANN Approach Prediction and Field Measurement for Pavement Section 16-9032 0 0.1 0.2 0.3 0.4 0.5 0.6 0 1 2 3 4 5 6 7 8 R ut tin g D ep th (i nc h) Pavement Age (Year) Measured Predicted Geosynthetic-reinforced Predicted Control

123 Figure 4.89. Comparison of Fatigue Cracking between ANN Approach Prediction and Field Measurement for Pavement Section 16-9032 Figure 4.90. Comparison of IRI between ANN Approach Prediction and Field Measurement for Pavement Section 16-9032 0 4 8 12 16 20 0 1 2 3 4 5 6 7 8 Fa tig ue C ra ck in g (% L an e A re a) Pavement Age (Year) Measured Predicted Geosynthetic-reinforced Predicted Control 70 80 90 100 110 120 130 140 0 1 2 3 4 5 6 7 8In te rn at io na l R ou gh ne ss In de x (i n/ m i) Pavement Age (Year) Measured Predicted Geosynthetic-reinforced Predicted Control