Laboratory evaluation of geocell-reinforced gravel subbase over poor subgrades

Transcription

1 Geosynthetics International, 23, 2, No. 2 Laboratory evaluation of geocell-reinforced gravel subbase over poor subgrades B. F. Tanyu, A. H. Aydilek 2,A.W.Lau 3, T. B. Edil 4 and C. H. Benson 5 Assistant Professor, Department. of Civil, Environmental and Infrastructure Engineering, George Mason University, 44 University Dr, Engineering Building, Suite 3, MS 6C, Fairfax, VA 223, USA, Telephone: , Telefax: , btanyu@gmu.edu 2 Associate Professor, Department of Civil and Environmental Engineering, University of Maryland, 73 Glenn L. Martin Hall, College Park, MD 2742, USA, Telephone: , Telefax: , aydilek@umd.edu 3 Engineer I-Geotechnical, Geotechnical Section, Express Rail Link, MTR Corporation, Hong Kong, China, formerly Graduate Research Assistant, Department of Civil and Environmental Engineering, University of Wisconsin, Madison, WI 5376, USA, Telephone: , alanlwl@mtr.com.hk 4 Professor, Department of Civil and Environmental Engineering and Director, Recycled Materials Resource Center, University of Wisconsin, 45 Engineering Dr, Madison, WI 5376, USA, Telephone: , Telefax: , tbedil@wisc.edu 5 Wisconsin Distinguished Professor and Chair, Department of Civil and Environmental Engineering, University of Wisconsin, 45 Engineering Dr, Madison, WI 5376, USA, Telephone: , Telefax: , chbenson@wisc.edu Received 9 March 22, revised 7 December 22, accepted 2 December 22 ABSTRACT: Large-scale experiments with cyclic loading were conducted to determine how incorporation of high-density polyethylene (HDPE) geocells affects the rutting properties of working platforms and resilient properties of a subbase in a pavement structure over soft subgrades. Four different geocells were used in this study to reinforce common subbase/base course gravel. Experiments were performed with 225 mm and 45 mm thick unreinforced and reinforced gravel and a crushed rock that is typically used for conventional cut-and-fill working platforms. Experiments were conducted to simulate loading conditions both during construction due to construction equipment and after construction due to traffic conditions over the asphalt pavement once the pavement structure is constructed. Materials used in this study were compacted to 9% relative compaction based on standard Proctor to determine the effect of geocells specifically with gravel material that is compacted to lower than typical standards. Deflections, modulus of subgrade reaction and resilient modulus of each section were evaluated. In summary, presence of geocells reduced the plastic deflection of the working platforms by 3 5%, improved the resilient modulus of the subbase by 4 5%, and the modulus of subgrade reaction by more than 2 times. KEYWORDS: Geosynthetics, Geocell, Subbase reinforcement, Poor subgrade, Deflection, Resilient modulus, Strain gauge REFERENCE: Tanyu, B. F., Aydilek, A. H., Lau, A. W. Edil, T. B. & Benson, C. H. (23). Laboratory evaluation of geocell-reinforced gravel subbase over poor subgrades. Geosynthetics International, 2, No. 2, [ INTRODUCTION Subgrade soils have a significant impact on the design, construction, structural response and performance of pavements. Problems arise during construction over poor subgrade with placement and compaction of subbase and base materials and in providing adequate support for subsequent paving operations. Once the pavement is commissioned, pavement structural responses continue to # 23 Thomas Telford Ltd be influenced by the subgrade. Therefore, proper preparation of soft subgrade prior to pavement construction has been recognized as an important issue in various states (e.g. Wisconsin where over 6% of the surficial soils are classified as silt and clay, which are considered as problem soils during construction (WisDOT 997)). For the purposes of this study, poor subgrade is defined as soils with low stiffness where the road construction 47

2 48 Tanyu, Aydilek, Lau, Edil and Benson becomes problematic due to large rutting during hauling. These soils in Wisconsin typically have California bearing ratio (CBR) of 4 or less. A common engineering practice for construction on poor subgrade is to remove the poor soil to a certain depth and replace it with a layer of material that has sufficient support value. This replaced layer is initially constructed to provide firm ground conditions to allow truck and equipment traffic during road construction (i.e. creating a working platform). If the performance of the working platform is anticipated to be sufficient for construction traffic without large repairs, subsequently it can be incorporated into pavement design to serve as a subbase layer. Examples of how the contribution of the working platform can be incorporated into flexible pavement design have been provided by Tanyu et al. (25) and Kim et al. (25). The most commonly used material to replace poor subgrade soils is crushed rock, which is a generic term used for large size aggregate and may have different particular meaning in different states. In this study, the crushed rock is referred to as breaker run, which is produced from mechanical crushing of quarried stone or reclaimed concrete that is not screened or processed after primary crushing (WisDOT 996). is selected for this study because it is commonly used in Midwestern states of the USA for improving poor subgrades. The thickness of breaker run used for poor subgrade improvement typically ranges between.3 and.9 m. In projects where breaker run is produced from quarried stone (as opposed to reclaimed concrete), natural rock resources are being used for production and, depending on the project location, sometimes the material has to be hauled over considerable distance, making the cost of using this material high. This method is also not considered to be the most sustainable solution because it requires mining and processing in the case of quarried stone with associated energy consumption and greenhouse gas emissions (Lee et al. 2). As a result, cost-effective alternatives are sought that reduce or eliminate the need for breaker run and yet provide adequate support during construction as a working platform and then provide a firm structural layer for the pavement system as a subbase. In this study, the use of four different geocells filled with aggregate is evaluated as an alternative to breaker run using large-scale model experiments conducted in a test pit. It is intended that the use of geocells will provide reinforcement to gravel aggregate and will eliminate the need for constructing thick crushed rock layers. 2. BACKGROUND ON GEOCELLS 2.. Performance of pavement systems and contribution of geocells In pavement systems, cracking (thermal and fatigue) and rutting are considered to be the principal types of distress conditions (Huang 993). In any of the pavement components, rutting stems from permanent (plastic) deformation. Rutting occurs mostly along the wheel paths where consolidation or lateral movement of the pavement materials is most likely. An alternative solution to limit rutting is to use cellular confinement systems, often known as geocells, that increase the strength and stiffness of soils and aggregates. Geocells improve the load deformation behavior of soils and aggregates through lateral confinement. Hoop stress in the cell and passive resistance by adjacent filled cells are responsible to the improvement in lateral spreading, whereas friction developed between the soil and cell wall increases the shear strength (Bathurst and Jarrett 988; Bathurst and Crowe 992; Mhaiskar and Mandal 992). These improvement mechanisms require a stiff material and therefore cellular confinement systems are mainly made of high-stiffness high-density polyethylene (HDPE) Existing research on geocell use in road construction Over the years, numerous studies have been performed to investigate the behavior of geocell-reinforced soils, but most studies are based on laboratory experiments. Some researchers such as Cowland and Wong (993), Dash et al. (23), Han et al. (28), Sireesh et al. (29), Yang et al. (2) and Zhang et al. (2) focused on studying the effect of geocell reinforcement on improving the bearing capacity of footings or embankments and determined that the geocell reinforcement is effective in increasing the bearing capacity when the geocell structure is properly designed. Other researchers, such as Bathurst and Karpurapu (993), Rajagopal et al. (999), Mengelt et al. (26), Wesseloo et al. (29) and Pokharel et al. (29, 2), studied the effect of geocell reinforcement on the deflection and modulus of the soils when the soils were subject to static and cyclic loading. These studies are relevant to the research described in this study and overall the findings showed that geocell-reinforced fill material is able to withstand 4 % higher loads based on the type of the geocell and the magnitude of applied loads. Mengelt et al. (26) determined that geocell reinforcement increased the resilient modulus of granular materials by.4 3.2% and of fine-grained soils by %. However, the studies listed above mainly focused on testing single-geocell structures and the results varied mainly due to the selected geocell material and the applied load conditions. Rajagopal et al. (999) and Wesseloo et al. (29) also studied the behavior of multiple geocells; researchers in one study concluded that at least three interconnected cells were required to produce repeatable shear strength results, and in the other study they noted that as the number of geocells increased, the peak strength of the geocell-reinforced soil decreased. To quantify the amount of decrease in peak strength as a result of change in geocell number, Wesseloo et al. (29) defined an efficiency factor. These experiments show the limitations of the bench-scale tests and the need for large-scale experiments in which the boundary conditions are simulated as closely as possible to field conditions. Bathurst and Jarrett (988) were among the early researchers who performed large-scale model tests Geosynthetics International, 23, 2, No. 2

3 Geocell-reinforced gravel subbase over poor subgrades 49 (2.4 m m 3.8 m) to study the load deformation behavior of gravel-infilled geocells. Geocells used in their study were and 2 mm deep and the results showed that incorporating geocells reduced rutting potential. Since this study, there have not been many large-scale laboratory tests to study the load deformation behavior of geocellreinforced soils; however, other researchers have performed field studies. Al-Qadi and Hughes (999) studied the effect of mm-deep geocells during a reconstruction project in Pennsylvania where the geocells were underlain by nonwoven geotextile to reinforce a subgrade with a CBR of 4. Back-analysis of deflection from falling weight deflectometer (FWD) data showed that geocell geotextile combination increased the modulus of the gravel by a factor of 2 at 3 years after construction. Latha et al. (2) performed field studies to evaluate the improvement of the load-carrying capacity and reduction in rut depth of unpaved roads when subbase layer is reinforced with variety of geosynthetics including geocells. Geocells used in this study were created from biaxial geogrid and this was determined to be the most efficient form of reinforcement compared to all other types of geosynthetics. Pokharel et al. (2) investigated the use of Neoloy polymeric-alloy geocell as reinforcement for base course for low-volume unpaved roads constructed over weak subgrade (i.e. CBR of 3). The experiments were conducted on full-scale trafficking tests using three different types of base course, 5 cm high geocells, and a non-woven geotextile as a separator between the subgrade and base layer. This research showed that geocell-reinforced sections could be constructed thinner and that the use of geocell increased the life of the unpaved road between.5 and 3.5 times depending on the aggregate used as infill. Existing studies show that geocell reinforcement can increase the modulus and strength of the infill materials. However, limited information exists on large-scale testing of geocell-reinforced aggregates placed on poor soils to draw consistent and quantitative conclusions with regards to the degree of improvement. The objective of this study was to assess improvements in rutting behavior and resilient modulus of a geocell-reinforced gravel layer when the materials were compacted to 9% relative compaction and to compare these results with that of breaker run and unreinforced gravel (i.e. no geocells). Experiments were designed for large-scale conditions to simulate the situation both during and after construction traffic loads. 3. MATERIALS Materials for this study consisted of breaker run, subbase/ base gravel, four different types of textured HDPE geocells, and expanded polystyrene (EPS) board simulating a typical soft subgrade. is large-size aggregate (usually less than 225 mm) obtained by mechanical crushing of rock or salvaged material that is not processed beyond initial crushing (WisDOT 996). The breaker run used in this study was a mix of crushed dolomite and limestone with 3% fines, coefficient of uniformity of 6 and effective grain size of.25 mm. It is classified as poorly graded gravel (GP) based on the unified soil classification system (USCS) and has a CBR of 8. The gravel used in this study as an infill to geocell was selected following the recommended gradations for subbase and base courses by the American Association of State Highway and Transportation Officials (AASHTO 993). AASHTO recommendations are set forth in the design guide to ensure adequate stability under repeated load and are such that the liquid limit of the fraction passing a No. 4 sieve must be less than 25 and the plasticity index (PI) must be less than 6. The selected gravel is called grade 2 gravel and is commonly used as aggregate for subbase/ base construction in Midwestern states (particularly in Wisconsin). It is produced from crushed or natural aggregate that is screened to meet the Gradation No. 2 requirement stated in WisDOT s Standard Specifications for Highway and Structure Construction (WisDOT 996). gravel (G2) is also occasionally used to construct working platform for truck traffic (Tanyu et al. 24). gravel used in this study was produced from crushed limestone and classified as poorly graded gravel with silt and sand (GP-GM) in accordance with USCS. The maximum dry density of the grade 2 gravel is 22.6 kn/ m 3 and the optimum moisture content is 8.2% in accordance with ASTM D 698 for standard Proctor test. The geocells used in this study were cellular confinement synthetics with a peak strength of 22 kn/m (at % strain) and residual strength of 8 kn/m (at % strain) (Table ). They consisted of mm thick strips of textured HDPE that are welded together at intervals along their length. The four different types of geocells used in this study are classified as GW(2)5, GW(2)2, GW(3)5 and GW(3)2. The number in parenthesis indicates the cell diameter in centimeters and the following number represents the cell depth in millimeters (i.e. four different geocells with diameters of 2 and 3 cm and heights of 5 and 2 mm). A 5 mm thick EPS board was employed to simulate a soft subgrade because the stiffness of the EPS board selected for this study is similar to that of soft subgrade materials in Wisconsin such as Antigo silt loam. Dynamic tests were conducted on low-density EPS specimens (.67 kn/m 3 density) having diameters of 5 mm and 3 mm with an aspect ratio of. These tests were compared with those conducted on Antigo silt loam (classified as CL-ML with PI ¼ 2). The EPS exhibited similar stress strain behavior to the Antigo silt loam and had similar modulus of slightly below kpa (Tanyu et al. 23). Nequssey and Jahanandish (993) also reported Table. Summary of the properties of geocells Sample Peak strength (kn/m) Peak strain (%) Residual stress a (kn/m) Strip Seam shear Seam peel a Residual stress at strain of %. Geosynthetics International, 23, 2, No. 2

4 5 Tanyu, Aydilek, Lau, Edil and Benson similarities in one-dimensional compressibility of EPS and clayey soils. Additionally, a needle-punched non-woven geotextile was used as a separator between the EPS and geocells. The geotextile consisted of polypropylene slit film yarns with an apparent opening size of.5 mm and tensile strength of 9 N. 4. TEST FACILITY AND EXPERIMENTAL PROGRAM Large-scale model experiments (LSMEs) were conducted in a test facility with a 3 m 3 3m m reinforced concrete test pit as shown in Figures and 2. Each test section was constructed in the LSME to simulate a working platform during construction of pavement structure and a subbase layer after the completion of the construction of the pavement structure. Each test section was constructed using either aggregate only (i.e. breaker run and grade 2 gravel) or geocell-reinforced gravel (i.e. grade 2 gravel). Geocell-reinforced and unreinforced sections were constructed to be 225 mm and 45 mm thick, respectively, and each was constructed over the simulated subgrade of EPS blocks of 45 mm thickness. A summary of the testing program is provided in Table 2. The EPS layer simulating a very soft subgrade was underlain by dense uniform sand. The sand had an effective grain size of.22 mm, a coefficient of uniformity of.8, a dry unit weight of 7.4 kn/m 3, a void ratio of.49 and a relative density of 85%. Before the EPS was placed, the sand was tamped and the surface was smoothed at an elevation 45 mm below the surface elevation of the test Position transducers MTS 9 kn Actuator Loading frame Figure 2. Photograph of large-scale model experiment set-up pit. Three layers of EPS, each with a thickness of 5 mm, were placed on top of the sand. A needlepunched non-woven geotextile was used as a separator on top of the simulated subgrade and the working platform/ subbase materials were placed on top of the geotextile. The test section layer was then formed by spreading the material using hand tools. Similar procedures were followed when geocells were used. Geocells were installed on top of the non-woven geotextile and backfilled with grade 2 gravel until the desired thickness was obtained. Each test section layer was placed in lifts and each lift was compacted with a vibratory plate compactor that Cyclic load through hydraulic actuator Ground surface Wooden walls Geocell-reinforced or unreinforced (gravel only) subbase Circular steel plate d 25 mm m EPS.45 m Sand 2.55 m Reinforced concrete pit walls 3. m Figure. Typical cross-section of large-scale model experiment (not drawn to scale). (Note: layers above 3 m could be adjusted for desired test length) Geosynthetics International, 23, 2, No. 2

5 Geocell-reinforced gravel subbase over poor subgrades 5 Table 2. Summary of testing program Profile a Layer thickness (mm) BR b 225 G2 b 225 G2 75 GW(3)5 5 G2 25 GW(3)2 2 G2 75 GW(2)5 5 G2 25 GW(2)2 2 BR 45 G2 45 G2 3 GW(3)5 5 G2 25 GW(3)2 2 G2 3 GW(2)5 5 G2 25 GW(2)2 2 a Profiles with geocell were created by placing geocell at the bottom of the profile. b BR, breaker run; G2, grade 2 gravel. delivered 3. kw at 36 rev/min with a square contact plate 45 mm wide. Compaction continued until no clear settlement was observed. After compaction, a rubber balloon test (ASTM D 267) was performed at the surface to check the dry unit weight and moisture content of each lift. To simulate the low compaction that typically occurs when materials are compacted over soft subgrade and to enhance the contribution of geocells, test sections were targeted to be compacted to 9% relative compaction (RC) based on standard Proctor test, which is lower than the typical target of 95% standard Proctor RC. A 9 kn MTS hydraulic actuator was used to apply a series of dynamic loads directly on the test materials. Two levels of loads were applied: one to simulate the limited number of construction equipment traffic load expected on working platforms during construction phase and the other to simulate the distributed traffic load on subbase after the pavement had been constructed and the pavement system was opened to service (i.e. traffic phase). Model experiments were performed first with construction load of 35 kn, which was applied through a circular load plate 25 mm in diameter directly on the test materials for cycles. The number of load cycles was determined based on typical truck traffic observed in road constructions in Wisconsin. The second load was also applied directly on the test materials using the same circular plate used for the first load and was applied immediately after the completion of the first load level. The magnitude of the second load was determined on the assumption that the working platform is incorporated into the pavement system as a subbase layer. Typical pavement layers above the subbase were not constructed in the LSME, but reduced traffic load over the subbase due to the presence of base and pavement layers was determined based on elastic analysis implemented in KENLAYER (a module within the KenPave program). The magnitude of the second load was selected to be 7 kn and was applied for cycles. A haversine load pulse was applied that consisted a. s load period followed by a.9 s rest period. Details of the selection of both construction and reduced loads are discussed in Tanyu et al. (24) and Tanyu et al. (23), respectively. Resistance type (bonded metallic foil) strain gages were used to measure strains in the geocells. Locations of the strain gages are shown on Figure 3. The strains measured by the strain gages were confirmed independently using digital image analysis (Aydilek et al. 24). Vertical deflections in the test pit were measured using Series 6 position transducers. Deflections at the surface were measured with the position transducers at 3 mm, 45 mm and 65 mm from the load. A four-point deflection basin was obtained using the data collected from the position transducers and the linear variable differential transducer (LVDT) in the actuator. A CR9 data logger was used for acquiring data from the position transducers and strain gages. Detailed description of the instrumentation is given by Lau et al. (2). 5. IMPACT ON RUTTING BEHAVIOR Rutting behavior is one of the important engineering properties of working platforms during construction. Excessive rutting affects the cost and schedule of the construction. To minimize the potential negative impacts, sometimes Department of Transportation agencies in the USA establish specific target deflection (rutting) values when evaluating materials to be used for working platforms. Once the pavement is constructed, rutting of subgrade or the layers above (such as working platform or subbase) under traffic load is still a concern as rutting of Applied load Geosynthetics International, 23, 2, No. 2 Column 4, 3, 2, Test section 4 gauges/location (ortho. direction doubled sided) 2 gauges/location (single direction, doubled sided) Figure 3. Locations of strain gages installed on geocells Test area Near side

6 52 Tanyu, Aydilek, Lau, Edil and Benson these layers may also have a detrimental effect on the layers above. Therefore, in this study the rutting behavior of the test sections was also evaluated under traffic phase loading. 5.. Rutting behavior under construction load 5... Total deflection Total deflections and applied stresses during construction phase for the 225 mm and 45 mm thick sections are shown in Figure 4. The maximum average applied stress for the 225 mm thick sections was approximately 4 kpa and the overall applied stress varied as much as 8 kpa based on the deflection behavior of the material. The applied stress was lower than the targeted stress, which was 7 kpa (i.e. 35 kn load with an approximately.5 m 2 loading plate). This is because the cross-sections were prepared with a lower target relative compaction than usual (i.e. 9% standard Proctor RC as opposed 95% or more RC) and the material could not withstand higher stresses without large deflections. Therefore, the maximum loads that could be applied became limited. However, all cross-sections with geocells accumulated less total deflection than that of grade 2 gravel and breaker run with the exception of GW(2)5. The average total deflections of geocells were approximately 7 mm (ranging from 55 to 9 mm), which is slightly higher than typically maximum desired total deflection during construction (i.e. 5 mm) (Tanyu et al. 24). For the 45 mm thick sections, all cross-sections with the exception of grade 2 gravel were able to withstand Deflection (m) Deflection (m) BR Total deflection Stress Elastic deflection G2 GW(3)5 Subbase configuration Figure 4. Total and elastic deflections in 225 mm and 45 mm subbase sections during construction phase GW(3)2 GW(2)5 GW(2)2 Applied contact stress (kpa) Applied contact stress (kpa) approximately the maximum targeted stress. The unreinforced grade 2 gravel had significant total deflections and could only sustain 5 kpa stress. Although this stress level is higher than what could be applied to 225 mm thick grade 2 gravel, without geocell reinforcement, grade 2 gravel continued to rut. The breaker run exhibited the smallest total deflection (45 mm) and all cross-sections with geocells deflected significantly less than unreinforced grade 2 gravel (,4% less). The average total deflections of geocells were approximately 95 mm (ranging from 9 to 4 mm). The significant reduction in total deflections of reinforced grade 2 gravel and the ability to withstand much higher stresses show the effectiveness of geocells for improvement over marginally compacted aggregate (i.e. 9% standard Proctor RC as opposed to 95% RC or more). The limitations on applied stresses that could be applied without total failure are believed to be mainly based on the low relative compaction of the materials tested and not a limitation of the LSME set-up. This is true because Tanyu et al. (24) showed that when grade 2 gravel and breaker run were compacted to 95% or more relative compaction based on standard Proctor, not only could the maximum targeted stress be applied but also the total deflections were accumulated to be significantly less (i.e. for 45 mm thick breaker run and grade 2 gravel sections, the total deflections were 9 mm and 22 mm, respectively) Elastic deflection The average elastic (recoverable) deflections during construction phase for both 225 mm and 45 mm thick sections are shown in Figure 4. The elastic deflections of all layers differed only by approximately only 2 mm, and geocells had less impact on elastic deflection than total deflection. However, it should be noted that the maximum stress that could be applied to 45 mm thick grade 2 gravel test section was approximately kpa less than could be applied to the test sections with geocells Plastic deflection Plastic deflection provides an indication of the rutting behavior of the various sections. Plastic deflections were obtained by subtracting the elastic (recoverable) deflections from the total deflections. Plastic deflections accumulated as a function of loading cycles during the construction phase for the 225 mm and 45 mm thick sections are shown in Figure 5. For all sections, the plastic deflections accumulated most rapidly at first, and then gradually diminished. In general, the largest plastic deflections were obtained with the grade 2 gravel. One exception for the 225 mm thick sections was GW(2)5. The reason for this discrepancy was not clear, but it is believed to be the result of differences in the test set-up. Plastic deflections for the geocell sections were comparable regardless of the geocell type and were either smaller than those for the breaker run sections (225 mm thick) or between those for grade 2 gravel and the breaker run sections (45 mm thick). At the end of cycles, plastic deflections in the geocell sections were about 5% Geosynthetics International, 23, 2, No. 2

7 Geocell-reinforced gravel subbase over poor subgrades 53 Cumulative plastic deflection (m) GW(3)5 GW(3)2 GW(2)5 GW(2)2 Deflection at cycles (mm) Edge of load plate GW(3)5 GW(3)2 GW(2)5 GW(2)2 Cumulative plastic deflection (m) GW(3)5 GW(3)2 GW(2)5 GW(2) of the unreinforced grade 2 gravel sections for both thicknesses. Zhou and Wen (28) also reported 44% reduction in plastic deflection of sands as a result of geocell reinforcement. The trend in plastic deflection accumulation for the grade 2 gravel sections differed from the trends obtained for the other sections. Plastic deflections in grade 2 gravel continued to increase with little reduction in the rate of increase at the end of the cycles. For all other sections, the rate of increase in plastic deflection approached zero by the time cycles was reached Deflection basin Deflection basins were created using the deflections measured at the contact between the load plate (radius ¼ 25 mm) and the surface of the section using the LVDT in the actuator and at other locations using position transducers. The position transducers were placed at three different locations away from the edge of the load plate: 3 mm, 45 mm and 65 mm (Figure 2). Transducers at locations 3 mm and 45 mm from the load were placed on both sides of the load plate and deflections measured at these locations were averaged to create the deflection basin. Deflection basin at the end of construction phase is shown in Figure 6. gravel had a deeper and wider 8 8 Figure 5. Plastic deflection in 225 mm and 45 mm subbase sections during construction phase Deflection at cycles (mm) Distance from center of load plate (mm) Edge of load plate GW(3)5 GW(3)2 GW(2)5 GW(2) Distance from center of load plate (mm) Figure 6. Deflection basins for the 225 mm and 45 mm subbase sections during construction phase deflection basin than that for geocell sections. The geocell sections had deflection basins that closely resembled the basin of the breaker run. At distances 25 mm (edge of load plate) and 3 mm from the load, deflections of breaker run and geocell cross-sections were 2 to 6 times less than those of unreinforced grade 2 gravel at the same locations. At distances greater than 45 mm from the load, difference in deflections became negligible among all materials tested. Bathurst and Jarrett (988) reached similar conclusions during loading of geocell-reinforced base course in a 2.4 m m 3.8 m test pit. They observed that the deformations occurred across a broader area of the pit when geocells were used, indicating a wider distribution of load relative to the distribution obtained in the unreinforced base course Strains in geocells Strain gages were attached on the geocells in the test for GW(2)2 to obtain the induced circumferential strain along and perpendicular to the direction of opening. The direction of opening is defined by the direction in which cells are opened from their collapsed form to their installation position. Strains in 225 mm and 45 mm thick sections during construction phase are shown in Figures 7a, 7b, 8a and 8b, respectively. Distance from the load is shown in each graph in terms of number of cells from the load. 7 7 Geosynthetics International, 23, 2, No. 2

8 54 Tanyu, Aydilek, Lau, Edil and Benson Cumulative strain (microstrain) cell from load 2 cells from load 3 cells from load Cumulative strain (microstrain) cell from load 2 cells from load 4 cells from load Cumulative strain (microstrain) cell from load 2 cells from load 3 cells from load Cumulative strain (microstrain) cell from load 2 cells from load 4 cells from load. (c). (d) Figure 7. Circumferential strain in GW(2)2 in 225 mm thick sections along and perpendicular to the direction of opening during the construction phase, and (c) along and (d) perpendicular to the direction of opening during the traffic phase The results show that strains decreased with distance from the load and reduced dramatically in the direction of opening and gradually in the direction perpendicular to the opening. Two hypotheses are suggested to explain this behavior: (i) the load transfer mechanism was most effective in the direction of opening due to stiffness of geocells and, as a result, the strain dissipated faster in this direction; (ii) the direction of opening of geocells was oriented perpendicularly to the direction of movement of the soil mass and, as a result, the hoop tension and the strains were lower. Geosynthetics International, 23, 2, No Rutting behavior under traffic load Total deflection Total deflections obtained during traffic phase (i.e. after the construction is completed under lower load level with load cycles) for the 225 mm and 45 mm thick sections are shown in Figure 9. During these tests, generally the applied stress was within kpa of the targeted maximum 4 kpa stress. This indicates that, although the materials were compacted only to 9% RC, if they survive the construction phase they are able to withstand the traffic loads. At the end of the traffic phase (see Figure 9) deflections from all sections were significantly less than the deflections at the end of the construction phase (see Figure 4). Also, as for the construction phase, the unreinforced grade 2 gravel had the largest total deflection when compared with reinforced sections. For the 225 mm sections, comparable total deflections were obtained with the geocells and breaker run. These deflections were typically less than one-half of that of the grade 2 gravel section. For the 45 mm sections, the total deflection of geocell-reinforced grade 2 gravel layers was 3% of the total deflection of the unreinforced grade 2 gravel layer but was 5% more than that of the breaker run. Therefore, for both 225 mm and 45 mm thick sections, geocells provided significant improvement over unreinforced grade 2 gravel in terms of total deflections. When compared with unreinforced layers, during traffic phase it appears that the reinforcement effect of geocell was more evident in thinner, 225 mm thick, sections than in thicker, 45 mm thick ones (i.e. the thinner geocellreinforced layers had similar total deflection to that of the breaker run section, but the thicker geocell-reinforced layers did not outperform breaker run). Both during construction and traffic conditions, geocell reinforcement was effective in increasing the stiffness of the unreinforced gravel. The improvement in stiffness ranged approximately between 5% and 7%. This finding is similar to that of Pokharel et al. (29). They reported

9 Geocell-reinforced gravel subbase over poor subgrades 55 Cumulative strain (microstrain) cell from load 2 cells from load 4 cells from load Cumulat ive strain (microstrain) cell from load 2 cells from load 4 cells from load.... Cumulat ive strain (microstrain) (c) cell from load 2 cell from load 3 cell from load Cumulat ive strain (microstrain) (d) cell from load 2 cells from load 4 cells from load Figure 8. Circumferential strain in GW(2)2 in 45 mm thick sections along and perpendicular to the direction of opening during the construction phase, and (c) along and (d) perpendicular to the direction of opening during the traffic phase 5% increase in stiffness of sand due to geocell reinforcement. However, their study was based on single-geocell units using sand as an infill. For direct comparison, in addition to the differences between the infill materials, other factors such as load levels, number and rate of load cycles also need to be considered Elastic deflection Elastic deflections during traffic phase for the 225 mm and 45 mm thick sections are shown in Figure 9; they are markedly lower than in the construction phase. The average elastic deflection was comparable for all sections having a thickness of 225 mm. For the 45 mm thick sections, the average elastic deflection was the lowest for breaker run section and highest for the section with grade 2 gravel. The elastic deflections for geocell sections fell between the elastic deflections of grade 2 gravel and breaker run. When compared with each other, elastic deflections of all geocell sections were similar. Thus, geocells had some impact on elastic deflections as the thickness of the layers increased Plastic deflection The larger plastic deflection of grade 2 gravel observed in the construction phase also occurred in the traffic phase (Figure ). In the traffic phase, geocell and breaker run sections underwent minimal plastic deflection, whereas plastic deflection in the unreinforced grade 2 gravel continued to increase throughout the test. At the end of the cycles of traffic load, the plastic deflection in geocell sections was 3 5% of that in grade 2 gravel (i.e. 5 7% improvement) for both thicknesses, similar to the observations made by Pokharel et al. (29) in testing of geocell-reinforced sands. In the traffic phase for both section thicknesses, geocells stabilized the rate of rutting and prevented rutting failure from occurring Deflection basin The deflection basin at the end of traffic phase is shown in Figure. The deflection basin obtained from traffic phase was similar to that obtained during construction phase. This indicates that during both tests, geocells were effectively providing support to grade 2 gravel even away from the load source (i.e. geocell-reinforced sections showed less deflection along the basin than unreinforced grade 2 gravel at same locations away from the load source) Strains in geocells As described previously, strain gages were attached on the geocells in the test for GW(2)2 to obtain the induced circumferential strain along and perpendicular to the Geosynthetics International, 23, 2, No. 2

10 56 Tanyu, Aydilek, Lau, Edil and Benson Deflection (m) Deflection (m) BR Total deflection Stress Elastic deflection G2 GW(3)5 Subbase configuration Figure 9. Total and elastic deflections in 225 mm and 45 mm subbase sections during traffic phase GW(3)2 GW(2)5 GW(2) Applied contact stress (kpa) Applied contact stress (kpa) Cumulative plastic deflection (m) Cumulative plastic deflection (m) GW(3)5 GW(3)2 GW(2)5 GW(2)2 2 GW(3)5 GW(3)2 GW(2)5 GW(2) Figure. Plastic deflection in 225 mm and 45 mm subbase sections during traffic phase direction of opening. Strains in 225 mm and 45 mm thick sections are shown in Figures 7c, 7d, 8c and 8d, respectively. Distance from the load is shown in each graph in terms of number of cells from the load. As for the construction phase, strains decreased with distance from the load for the traffic phase and reduced dramatically in the direction of opening and gradually in the direction perpendicular to the opening. Comparison of strains during the construction and traffic phases based on Figure 7 indicates that the strain during the traffic phase is dramatically lower than that during the construction phase for 225 mm thick sections. Latha et al. (2) suggest that significant strains develop only when the surcharge pressure exceeds the surcharge capacity of the unreinforced soil. In the construction phase, the strain in the geocells was larger because the load was large relative to the shear strength of the soil. In contrast, the lower surcharge pressure during the traffic phase was small relative to shear strength of the soil and, although geocells still contributed to the overall stability, their relative contribution was smaller, as indicated by their very low strains. Similar trends were observed for the 45 mm thick sections as shown in Figure 8. Rajagopal et al. (999) and Dash et al. (24) observed that a considerable amount of confinement is developed with the use of geocells. Rajagopal et al. (999) showed that three interconnected cells were required to realize the Geosynthetics International, 23, 2, No. 2 full effect of geocells. The distribution of strain in Figure 7 shows that the strain decreases significantly with distance from the load in both directions parallel and perpendicular to opening, but around the third or fourth cell from the load the strain becomes constant. Hence, from this perspective, the number of cells required to realize the full effect of geocells was the same in this study as was observed by Rajagopal et al. (999). Strains in the geocells were also used to estimate the increase in confining stress ( ó 3 ) applied by the geocells using the theory developed by Henkel and Gilbert (952) with the relationship presented in Equation. ó 3 ¼ 2Må c () d å a where M is the modulus of the membrane, å a is the axial strain of the specimen, d is the original diameter of the specimen, and å c is the circumferential strain as defined by Equation 2. å c ¼ p ffiffiffiffiffiffiffiffiffiffiffiffi å a p ffiffiffiffiffiffiffiffiffiffiffiffi (2) å a Table 3 shows the increase in confinement stress obtained during construction phase. Confinement perpendicular to the opening was consistently higher than that

11 Geocell-reinforced gravel subbase over poor subgrades 57 Deflection at cycles (mm) Deflection at cycles (mm) Edge of load plate Distance from center of load (mm) Edge of load plate Distance from center of load (mm) GW(3)5 GW(3)2 GW(2)5 GW(2)2 parallel at a given distance from the load. Bathurst and Jarrett (987) made similar observations during large-scale ( m 3.5 m) unconfined compression testing of geocellreinforced granular backfills. Their results showed that 6 GW(3)5 GW(3)2 GW(2)5 GW(2)2 Figure. Deflection basins for the 225 mm and 45 mm subbase sections during traffic phase 6 Table 3. Increase in confinement induced during construction phase for the 225 mm and 45 mm thick sections of GW(2)2 Distance from load Orientation relative to opening 7 7 Increase in confinement (kpa) 225 mm sections Parallel 4. cell from load Parallel cells from load Parallel cells from load Parallel 2.8 Perpendicular 4. cell from load Perpendicular.5 2 cells from load Perpendicular cells from load Perpendicular Not analyzed 45 mm sections Parallel.4 cell from load Parallel.9 2 cells from load Parallel.8 3 cells from load Parallel.8 Perpendicular.4 cell from load Perpendicular. 2 cells from load Perpendicular.9 3 cells from load Perpendicular.8 Geosynthetics International, 23, 2, No. 2 deflections along the direction of opening were smaller, which indicated that the composite was stiffer perpendicular to the direction of opening. Table 2 also shows that an additional covering of grade 2 gravel on top of geocells reduces the effect of the load on confinement by geocells. During traffic phase, the increase in confinement was negligible due to the small load applied (not shown here). In conclusion, geocell reinforcement provided an increase in confining stress. This increase may lead to an increase in resilient modulus as shown by Mengelt et al. (26) up to approximately 3% in grade 2 gravel. The increase in resilient modulus due to presence of geocells in the current study is discussed below. 6. IMPACT ON PROPERTIES FOR PAVEMENT DESIGN Performance of the sections in large-scale model experiments is also evaluated against the engineering properties that are of importance for pavement design. 6.. Modulus of subgrade reaction The modulus of subgrade reaction (K) is the property used in the design of rigid pavements. It is defined as the ratio of stress applied (ó) to the total deflection ( L) measured at the contact between the load plate and pavement section under a dynamic load. The estimated moduli of subgrade reaction of the unreinforced and geocell-reinforced sections investigated in this study are shown in Figure 2 as a function of number of cycles under the traffic phase loads. Table 4 shows the modulus of subgrade reaction (K) and improvement ratio (IR K ) (as defined in Table 4) for all sections at the end of traffic phase loading. The IR K for 225 mm thick sections with geocells is in the order of (excluding the data for the GW(2)5), and for 45 mm thick sections it is in the order of Therefore, it can be stated that the presence of geocells was effective in improving the K value of unreinforced grade 2 gravel Resilient modulus The resilient modulus is the main design parameter for flexible pavements. The KENLAYER computer program was used to obtain the resilient modulus of the materials tested in this study. KENLAYER allows that resilient modulus (M r ) of granular materials follows a non-linear power function model (Equation 3). M r ¼ k ó k 2 b (3) where ó b is the bulk stress and k and k 2 are empirical constants. The pavement section in the LSME was simulated using KENLAYER as described by Tanyu et al. (23, 25). The resilient modulus of the subbase layer (unreinforced or geocell-reinforced) was left as a variable and modeled as given in Equation 3, and all other parameters (e.g. subgrade modulus) and the layer thicknesses were fixed at their measured values. Rada and Witczak (98, 982) and Zaman et al. (994) indicate that k 2 is primarily a material property that falls within a

12 58 Tanyu, Aydilek, Lau, Edil and Benson Modulus of subgrade reaction (kn/m ) 3 Modulus of subgrade reaction (kn/m ) GW(3)5 GW(3)2 GW(2)5 GW(2) Figure 2. Modulus of subgrade reaction (K) for 225 mm and 45 mm thick sections during the traffic phase. narrow range. Thus, in this study initially the values of parameter k 2 of aggregate sections were fixed based on previous recommendations by Rada and Witczak (98). The initial k 2 parameters of geocell sections were also fixed using the same value as for grade 2 gravel. The parameter k, which varies over a broad range, was adjusted until the measured elastic deflection from LSME and predicted elastic deflection from KENLAYER matched for the applied plate loads. The resilient modulus is then determined using Equation 3 based on the estimated bulk stress of 28 kpa (NCHRP 24) based on Mechanistic Empirical Pavement Design Guide for a typical pavement profile. The resilient modulus corresponding to 28 kpa is called the summary resilient modulus (SRM) and is considered representative of the base course in a pavement as suggested by Section of NCHRP -28A (NCHRP 24). Uzan (985) indicates that Equation 3 neglects the effect of accumulation of shear strains and, when the equation is used, the vertical strain should be between and : Deflection data from LSME and subsequent analyses by KENLAYER indicated that vertical strain in the subbase was beyond the range described by Uzan (985). For 225 mm and 45 mm thick sections of unreinforced grade 2 gravel, the vertical strain ranged from to and from to.4 3 2, respectively. Therefore, to avoid the strain limitation, the improvement in resilient modulus with geocell sections is presented in the form of an improvement ratio rather than an actual resilient modulus value. The improvement ratio, IR M is obtained by dividing the predicted summary resilient modulus of geocell-reinforced cross-sections (SRM GR ) by the predicted summary resilient modulus of the unreinforced grade 2 gravel section (SRM UR ). Table 5 presents the estimated IR M with geocells. For both 225 mm and 45 mm thick sections with geocells the IR M is on the order of.4.5 (i.e. 4 5% higher SRM) (excluding the data for GW(2)5). The findings in this study of increase in resilient modulus with geocell reinforcement are qualitatively consistent with the findings of other researchers. Han et al. (28) also reported an increase of approximately 5% in elastic modulus (i.e. from 3.2 MPa to 6 MPa) when they tested a sand with 7% relative density with and without a single-geocell reinforcement. Mengelt et al. (26) observed approximately 3% increase in resilient modulus of geocell-reinforced grade 2 gravel when the improvement was tested in a single-cell conventional resilient modulus apparatus. However, Mengelt (2) also concluded that the axial strains increased very slowly over the course of the repeated tests, producing minimal hoop stresses in geocells. Therefore, the increase in resilient modulus was attributed to increasing particle interlock in both reinforced and unreinforced specimens and not to the presence of the geocells. Al-Qadi and Hughes (999) observed a twofold increase Table 4. Geocell improvement ratios (IR) based on modulus of subgrade (K) for 225 mm and 45 mm sections under traffic loads Materials tested 225 mm thick sections 45 mm thick sections K (kn/m 3 ) IR K K (kn/m 3 ) IR K GW(3)5 K GR GW(3) GW(2) GW(2) K UR Notes: K is determined at the end of cycles. K GR, modulus of subgrade reaction of geocell-reinforced sections; K UR, modulus of subgrade reaction of unreinforced sections; IR K ¼ K GR /K UR : Geosynthetics International, 23, 2, No. 2

13 Geocell-reinforced gravel subbase over poor subgrades 59 Table 5. Geocell improvement ratios (IR) based on summary resilient moduli at ó b sections under traffic loads. 28 kpa for 225 mm and 45 mm Materials tested 225 mm thick sections 45 mm thick sections SRM (kpa) IR M SRM (kpa) IR M GW(3)5 SRM GR GW(3) GW(2) GW(2) SRM UR Notes: ó b is bulk stress. SRM GR, summary resilient modulus of geocell-reinforced sections; SRM UR, summary resilient modulus of unreinforced sections; IR M ¼ SRM GR /SRM UR : in FWD back-calculated resilient moduli of a mm thick conventional aggregate field section as a result of geocell reinforcement. Their finding is also higher than what was observed in this study. However, their layer was much thinner than the ones tested here and possibly better compacted. The improvement due to geocell reinforcement appears to be higher in thinner layers. 7. CONCLUSIONS A series of large-scale model experiments were conducted to determine how incorporating geocells into a granular layer placed over soft subgrade affects the rutting and resilient behaviors of pavements. Comparisons were made with a granular layer without geocells as well as an alternative layer using a stronger material (i.e. breaker run) which is used commonly by highway agencies as the working platform over soft subgrades. Each profile was tested in two phases corresponding to construction and traffic load levels and numbers of cycle conditions. Deflections (total, elastic and plastic) and pavement design properties for rigid and flexible pavements, i.e. modulus of subgrade reaction and resilient modulus, respectively, were obtained for each profile. Strains in the geocells were also measured during the tests in directions parallel and perpendicular to the opening of geocells to provide data needed to understand the reinforcement mechanism. This study shows that the main benefit of geocells is in relation to the rutting behavior. The modulus of subgrade reaction and resilient modulus are also improved, but to a lesser degree. The following observations support this conclusion. In both the construction and the traffic phases of the 225 mm and 45 mm thick working platforms, plastic deflection of the geocell-reinforced grade 2 gravel was 3 5% of the plastic deflection of the unreinforced gravel. The total deflections recorded from the LSME for both reinforced and unreinforced grade 2 gravel were higher than what is typically desired during construction. This is because both reinforced and unreinforced grade 2 gravel were compacted to 9% standard Proctor relative compaction. In all experiments, geocells were effective in increasing the stiffness of the unreinforced grade 2 gravel. The difference in geocell geometry (diameter and cell height) investigated in this study did not appear to provide significant differences in rutting behavior of the tested sections within LSME. When compared with breaker run, geocell-reinforced grade 2 gravel sections were able to carry about the same stress as the breaker run section in all experiments, although the effect of reinforcement was more evident in thinner sections tested in this study. The strain gage data obtained from geocells indicated that geocell reinforcement provided increase in confining stress. This increase is believed to lead to an increase in resilient modulus. In both 225 mm and 45 mm thick sections, the presence of geocells was effective in improving the modulus of subgrade reaction of unreinforced grade 2 gravel. The current study showed that the presence of geocells improved the resilient modulus by 4 5% in both 225 mm and 45 mm thick sections, with the exception of 225 mm thick test with GW(2)5. In that test, the improvement was 3%. The conclusions presented in this paper are based on the materials tested in this study under specific loading conditions. The effects of surface water infiltration or the changes in ground temperature were not considered in this study. ACKNOWLEDGEMENTS Geosynthetics International, 23, 2, No. 2 Financial support for this project was provided by the Industrial and Economic Development Research Program administered by University-Industry Relations at the University of Wisconsin, Presto Products of Appleton, Wisconsin, US Federal Highway Administration Recycled Materials Resource Center (RMRC), and the Wisconsin Department of Transportation. The support of these parties is gratefully acknowledged. The findings and opinions expressed in this paper are solely those of the authors and endorsement by the sponsors is not implied.