Stability Study Of Slopes With Different Soil-rock Mixing Ratios In Open-pit Mine Dump Site Based On Transfer Learning

During the energy extraction process,open-pit mining waste dump slopes are often affected by rainfall,resulting in geological hazards such as large-scale landslides,debris flows,and collapses.These hazards pose a serious threat to the safety of mine workers and nearby residents,as well as negatively impacting the production and economic efficiency of the mine.Therefore,conducting stability analysis and research on open-pit mining waste dump slopes is crucial for disaster prevention and safety production in mines.However,there are many challenges,such as complex geological conditions,difficulty in obtaining unstable landslide information,and limited landslide sample data.This paper focuses on the F open-pit mining waste dump in Shaanxi Province,China and uses physical model experiments to explore the unstable failure mechanism of waste dump slopes with different soil-rock ratios under rainfall conditions.Furthermore,a transfer learning-based stability analysis method for waste dump slopes is proposed.The specific research content includes:(1)Designing laboratory rainfall-induced instability tests on waste dump slopes with different soil-rock ratios,based on the physical and mechanical properties of the material obtained from indoor soil mechanics tests.(2)Investigating the failure mechanism of waste dump slopes with different soilrock ratios under rainfall infiltration.The relationship between the deformation and failure phenomenon of the study model,water content,and stress change of the slope model is analyzed.The unstable failure mechanism of waste dump slopes is revealed under different rainfall intensities and different soil-rock ratios.(3)Constructing a stable state discrimination model for open-pit mining waste dump slopes using a transfer learning algorithm.Due to the poor performance of traditional single classification models in processing high-dimensional complex data and small sample data,a GBDT ensemble learning classification model is constructed with CART decision trees as the base learning algorithm.Furthermore,the Tr Adaboost algorithm is used to transfer the data samples between the different soil-rock ratio slopes to improve the discrimination accuracy of the GBDT model on small sample data label categories.The Tr Adaboost-GBDT model for stable state discrimination of waste dump slopes is validated using experimental data from physical model experiments.(4)Stability state discrimination of slope of open pit mine drainage field.Taking the slope of F open pit mine in Shaanxi Province as the research object,the stability of the slope of the drainage field is analyzed by numerical simulation,and the water content,safety coefficient and stress-strain data inside the slope are obtained,and the Tr Adaboost-GBDT model is used to analyze and discriminate the stability state of the slope of the drainage field.The experimental results show that the algorithm model proposed in this paper has higher accuracy and performance in discriminating the instability status of slope of drainage field in real engineering.By comparing the calculation results of the discriminative model and numerical simulation,it further verifies the feasibility and practical value of the Tr Adaboost-GBDT model in the discriminative state of slope stability of open pit mine.