Investigations on pullout behavior of geogrid-granular trench using CANAsand constitutive model

Abstract

Experimental and numerical investigations have been carried out on behavior of pullout resistance of embedded circular plate with and without geogrid reinforcement layers in stabilized loose and dense sands using a granular trench. Different parameters have been considered, such as the number of geogrid layers, embedment depth ratio, relative density of soil and height ratio of granular trench. Results showed that, without granular trench, the single layer of geogrid was more effective in enhancing the pullout capacity compared to the multilayer of geogrid reinforcement. Also, increasing the soil density and embedment depth ratio led to an increase in the uplift capacity. When soil was improved with the granular trench, the uplift force significantly increased. The granular trench improved the uplift load in dense sand more, as compared to the same symmetrical plate embedded in loose sand. Although it was observed that, in geogrid-reinforced granular trench condition, the ultimate pullout resistance at failure increased as the number of geogrid layers increased up to the third layer, and the fifth layer had a negligible effect in comparison with the third layer of reinforcement. Finite element analyses with hardening soil model for sand and CANAsand constitutive model for granular trench were conducted to investigate the failure mechanism and the associated rupture surfaces utilized. The response of granular material in the proposed model is an elastoplastic constitutive model derived from the CANAsand model, which uses a non-associated flow rule along with the concept of the state boundary surface possessing a critical and a compact state. It was observed that the granular trench might change the failure mechanism from deep plate to shallow plate as the failure surface can extend to the ground surface. The ultimate uplift capacity of anchor and the variation of surface deformation indicated a close agreement between the experiment and numerical model.