| Since the impoundment of the Three Gorges Reservoir in 2003,numerous landslides with characteristics of step-like deformation have formed in the reservoir area,which not only affects the stable operation of the Three Gorges Reservoir but also poses a serious threat to the safety of nearby residents’ lives and property.Studying the deformation mechanism of step-like landslides can help us understand the evolution law of landslide deformation and facilitate the adoption of effective landslide control measures and the development of landslide prediction and warning systems.Relevant studies have shown that the step-like deformation characteristics of landslides are closely related to the periodic variation of soil water content.An increase in soil water content weakens the shear strength of the landslide body,causing it to become unstable and slide.Based on the changes in the strength of soil under different water content,this study establishes a functional network big data nonlinear adaptive regression algorithm based on particle swarm optimization,revealing the big data law between soil strength parameters and soil water content,and developing a comprehensive calculation model for soil strength indicators.The step-like deformation evolution law of landslides under dynamic changes in water content is obtained through numerical simulation,revealing the deformation mechanism of step-like landslides.The main work and conclusions of this study are as follows:(1)The functional network is used in nonlinear regression analysis,and the adaptive selection of neural functions and the adaptive optimization of network structure are realized through the establishment of a standard basis function pool and a neural adjustment mechanism.The optimal solution of regression model parameters is obtained by combining the functional network regression model with the particle swarm optimization algorithm.Finally,a functional network big data nonlinear adaptive regression algorithm based on particle swarm optimization is established by setting several initialization network structures and combining the AIC model evaluation criteria.(2)A large amount of soil physical and mechanical index data is collected from authoritative domestic and foreign journals.The core correlation factors of soil effective strength indicators are analyzed by introducing the maximum mutual information coefficient.The results show that there is an obvious correlation between soil effective strength indicators and water content and liquidity index.A comprehensive calculation model for effective strength indicators considering multiple influencing factors is established by selecting liquidity index,plasticity index,and void ratio as model parameters.Verification results show that the data fitting degree of the comprehensive calculation model is higher than that of the single-factor calculation model.(3)The basic framework of a viscoelastic-plastic constitutive model is established,and the stress update process in each incremental step is detailedly derived based on the implicit stress integration algorithm.The corresponding UMAT subroutine is developed to realize the secondary development of the viscoelastic-plastic constitutive model in ABAQUS.Finally,the correctness of the secondary development model is verified through a simple example.(4)Based on hydrogeological data and displacement and rainfall monitoring data,the deformation evolution law of the Tanjiawan landslide in the Three Gorges Reservoir area is analyzed.The results show that the Tanjiawan landslide is a typical ancient landslide,and it first experienced significant sliding deformation under the influence of three heavy rainfall events in 2014,and then there were displacement mutations every rainy season,showing obvious step-like deformation characteristics.Combined with the big data calculation model of soil effective strength indicators established above,the displacement and plastic zone evolution laws of the Tanjiawan landslide under different soil water content are analyzed by numerical simulation.The results show that the Tanjiawan landslide is a translational landslide in terms of failure form,and the rear of the deformation zone of the landslide experiences displacement deformation first,followed by the overall sliding deformation of the deformation zone. |
