Shallow Sliding Failure Prediction Model of Expansive Soil Slope based on Gaussian Process Theory and Its Engineering Application

Numerous studies on the instability and shear strength of expansive soil slopes have suggested that the fundamental reason for shallow sliding failures is insufficient shear strength caused by repeated dry-wet cycles under low stress. Therefore, this paper studied the effects of low stress and dry-wet cycles on the shear strength of expansive soil slopes. A nonlinear curve that can be well fitted using generalized power functions was observed for the shear strength of expansive soil, and a numerical model with nonlinearly distributed shear strength taken into consideration was built using FISH. In this way, the dynamic distribution of shear strength as a function of vertical stress was obtained. Compared to conventional models, the proposed model was more accurate in terms of theoretical aspects, and the obtained results exhibited improved consistency with the actual mode. In order to improve the applicability of this model, a mapping model of atmospheric action and engineering design of safety factors of slopes was established using a Gaussian Process Regression (GPR) model, which is suitable for cases characterized by high dimension, small sample population, and nonlinearity. Herein, the numerical calculations were replaced with output responses. In this way, fast prediction of the stability of expansive soil slopes in terms of a shallow sliding failure prediction model was achieved. To verify the feasibility of the proposed approach, dynamic predictions of shallow sliding failures of expansive soil slopes were carried out for Nanyou Highway under different rainfall conditions. Meanwhile, safety factor responses corresponding to different slope design schemes were also obtained using the proposed model, providing reference for slope stability analysis in expansive soil zones.