Analysis of end-restraint effects in DEM triaxial tests using linear and hertz contact models

This study investigates how end restraint influences the relative peak-response performance of Linear and Hertz contact models in discrete element simulations of sand triaxial tests. Consolidated-drained tests were conducted on Xiamen medium sand at confining pressures of 100, 200, 300, and 400 kPa. A flexible-boundary model was developed through PFC–FLAC3D coupling. Both contact models were calibrated against the same experimental deviatoric stress–axial strain curve at 100 kPa; therefore, this case represents calibration rather than independent validation. With all micromechanical parameters fixed, the 200–400 kPa tests were used for independent validation. Based on peak stress and peak strain, the Linear model produced smaller deviations at 100 and 200 kPa, whereas the Hertz model performed better at 300 and 400 kPa. At 400 kPa, the Hertz model yielded a lower relative peak-stress error (1.19% versus 2.37%) and a smaller peak-strain deviation (0.4 versus 0.6% points) than the Linear model. For the modeled interface contrast µ = 0.2 for frictional ends and µ = 0 for ideal smooth ends—frictional ends increased simulated peak strength by approximately 0.9%–8.0%. This difference decreased with increasing confinement, from 5.57% to 0.90% for the Linear model and from 7.41% to 1.59% for the Hertz model between 100 and 400 kPa. End friction also restricted radial deformation near the loading platens and altered the specimen deformation pattern. These results show that contact-model agreement depends on both stress level and end boundary condition.