The subgrade material used in this study was a mix of kaolin clay (plasticity index, PI = 24) and non-plastic concrete sand (ASTM, 2018) at a mixing ratio 1:3 (i.e., 25% clay and 75% sand) based on their dry masses. Three subgrades at different CBR values were prepared at different moisture contents in this study. The subgrade with a nominal CBR value of 2% was prepared at a moisture content of approximately 9.8%, while the intermediate subgrades with nominal CBR values of 3% and 4% were prepared at the moisture contents of approximately 8.9% and 8.5%, respectively. The maximum dry unit weight of the subgrade was 20.3 kN/m³. The subgrade was compacted in six layers of 150 mm thick each layer at the desired moisture content and density to achieve the required CBR value.

To control the quality of the subgrade material during construction, vane shear tests were conducted according to ASTM (2016c) for each subgrade layer, and also Dynamic Cone Penetrometer (DCP) tests according to ASTM (2016d) were conducted after the completion of the subgrade layers and after the placement of the aggregate base.

The base course material used in all test sections was Virgin Granular Base (VGB) that is usually utilized in road projects and meets the specifications of the Kansas Department of Transportation (KDOT) for granular base courses under concrete pavements (KDOT, 2018). Figure 1 shows the gradation curves of the base course along with the upper and lower limits defined by KDOT (2018) . The plasticity index of the VGB particles passing No. 40 sieve was 7 and the optimum moisture content and its corresponding maximum dry unit weight according to modified Proctor tests (ASTM, 2016b) were 7.1% and 21.2 kN/m³, respectively. The nominal CBR value of the aggregate base at the moisture content of approximately 7.1% was 15% and verified by DCP tests after the preparation of each test section.

A nonwoven geotextile layer was placed between the aggregate base and the subgrade and anchored to the subgrade layer at the four corners of the test box. Nonwoven geotextile is commonly used for a variety of functions including separation, filtration, and drainage for geotechnical engineering projects. Table 1 shows the parameter values of the nonwoven geotextile used in the tests.

The geo-mat made of high-density polyethylene material consists of porous cells with the mat dimensions of 1 m long, 0.5 m wide, and 0.05 m high. Each cell is either rectangular (83 mm × 165 mm) or square (83 mm × 83 mm) to provide confinement to aggregate filled in the cell. The geo-mats were connected using U-shape steel clips to form a geo-mat system. Figure 2 shows the geo-mat system configuration in the test box.

Totally six large-scale tests were constructed inside a test box with dimensions of 2.0 m × 2.2 m × 2.0 high. In each test, six subgrade layers with a thickness of 150 mm were placed and compacted by an electrical compactor to a target CBR value (i.e., 2%, 3%, or 4%). Five vane shear tests were conducted at two depths for each compacted subgrade layer to confirm the CBR value. After the completion of the subgrade layers, four dynamic cone penetrometer (DCP) tests were conducted at different locations for each test section. In the control sections (i.e., without a geo-mat system), the base course material was compacted in two layers with a thickness of 150 mm. In the test sections with the geo-mat system, a nonwoven geotextile layer was placed on the top surface of the subgrade and anchored to the subgrade layer by stakes at the four corners of the test box before the placement of the base course layer. The base course material was placed and compacted in one layer of different thickness (i.e., 150 mm and 50 mm) as desired. Sand cone method (ASTM, 2016a) was carried out for the base course layer to confirm the relative compaction not less than 95% of the maximum dry unit weight. Three more DCP tests were carried out after the compaction of the aggregate base to confirm the CBR values for the aggregate base and the subgrade. The correlation developed by Webster et al. (1994) as shown in Eq. 1 has been commonly used in the literature to estimate the CBR values of the base and the subgrade from the DCP tests. The CBR value of the subgrade has also been correlated with its undrained shear strength in the literature. Walkenbach et al. (2019) established the correlation between the subgrade CBR value and its undrained shear strength from the vane shear test as shown in Eq. 2, specifically for the subgrade used in this study. The geo-mat units were placed on top of the base course and connected together by U-shape clips and then filled in with compacted virgin granular base material. Table 2 presents the average CBR values calculated from the DCP test and the vane shear test for the subgrade and the base course layers. CBR = 292 / D P I 1.12 (1) CBR = c u / 29.54 (2) where DPI = dynamic penetration index in mm/blow and c_u = undrained shear strength from the vane shear test in kPa.

Since the length of the geo-mat unit was 1 m, two of the geo-mat units were cut in halves in all the test sections to arrange the geo-mat units in bricklayer pattern inside the test box with the dimension of 2 × 2 m. The mat inside the test box was consisted of six complete geo-mat units and four-halves units. The complete mat unit was connected to the adjacent mats by 12 steel clips, while the cut mats were connected to others by six steel clips. The load was applied on a 300-mm diameter steel plate seated on three adjacent mats. Figure 2B shows the bricklayer pattern of the geo-mat system with the location of the loading plate.

Cyclic load increments of 17.8, 35.6, 53.4, and 71.2 kN were applied on the top surface to simulate different axle loads ranging from a small axle load to an H-20 load that has a single axle load of 142.4 kN (i.e., each tire 71.2 kN). The time for each single cycle of load was 1.3 s, i.e., a frequency of 0.77 Hz. The actuator produced cyclic loading in a trapezoidal waveform with a minimum load magnitude of 0.2 kN as shown in Figure 3. A pre-installed displacement transducer (i.e., inside the actuator) was used to measure the loading plate displacements during the test. Three more displacement transducers were placed on the surface of the base course or the geo-mat at distances of 300, 450, and 600 mm from the loading plate center. To measure the vertical stress at the interface between the subgrade and the aggregate base, six earth pressure cells were used in each test, four of them, installed at the center and at distances of 150, 300, and 450 mm from the loading plate center, and the other two placed in the transverse direction in front and behind the loading plate at the distance of 300 mm from the center. Figure 4 shows the schematics of the test sections with the locations of sensors.

Figure 5 shows the displacements of the six tests under the loading plate subjected to incrementally cyclic loading. Each load increment was applied to the maximum cycle number of 2000, or until the maximum displacement reached 75 mm or more in some tests. The permanent displacement represents the smallest displacement in the curve for each load increment, while the resilient displacement is the band thickness of the curve (i.e., the maximum displacement minus the permanent displacement). Figure 5A presents a comparison between the two control test sections with different subgrade CBR values. As expected, the control test with a strong subgrade (i.e., 4% CBR) showed less vertical displacement and higher load bearing capacity than the control test with a weak subgrade (i.e., 2% CBR). For example, at the end of the cyclic load of 17.8 kN, the maximum displacement under the loading plate center decreased by 38.5% when the CBR value increased from 2% to 4%. Also, the control test section with the weak subgrade reached a punching failure at the cyclic load of 35.6 kN while the test section with the intermediate subgrade (i.e., 4% CBR) failed at the cyclic load of 54.4 kN. To assess the performance and the benefit of the geo-mat system, comparisons between the control test sections and the geo-mat test sections were made at different CBR values of subgrade. Figure 5B shows the comparison between the control and geo-mat sections on the weak subgrade. The test results show that both tests failed before completing 2000 cycles at the cyclic load of 35.6 kN. At the cyclic load of 17.8 kN, the control section exhibited less displacement as compared with the geo-mat section, e.g., at the end of cycle number, the permanent displacement in the control section was 16.6 mm while 25.6 mm in the geo-mat section. This behavior changed when the cyclic load increased to 35.6 kN. Even though both tests failed before completing 2000 cycles of the load increment of 35.6 kN, the permanent displacements at 500 cycles of load were 82 mm for the control section and 71.1 mm for the geo-mat section, respectively. It should be pointed out that the geo-mat section had a base course layer of 150 mm thinner than the control section. Figure 5C shows the comparison between the control and geo-mat sections on 4% CBR subgrade. The test results show almost the same performance for both tests up to the cyclic load of 35.6 kN, but the geo-mat section exhibited more bearing capacity than the control section with the increasing cyclic load. For example, the control section failed at the cyclic load of 53.4 kN while the test section failed at the cyclic load of 71.2 kN. Figure 5D presents the comparison between the geo-mat test sections with different base course thicknesses (i.e., 50 mm and 150 mm) and different CBR values of subgrade. The test results show that the bearing capacity of the test section increased with the increase of the subgrade CBR value. Moreover, the thickness of the base course layer had insignificant effect on the bearing capacity of the test section on the intermediate subgrade (i.e., 4% CBR).

To quantify the benefits of the geo-mat system at different cyclic loads, permanent displacements induced by each cyclic load increment were determined and are presented in Table 3. The permanent displacements were reported at the last cycle of each load increment or at the cycle when the displacement under the loading plate reached the maximum value (approximately 75 mm or higher). Table 3 shows that the geo-mat system reduced the permanent displacement by approximately 13% at the cyclic load of 35.6 kN for the 2% CBR subgrade and 24.8% at the cyclic load of 53.4 kN for the 4% CBR subgrade.