A Practical Prediction Model for Surface Deformation of Open-Pit Mine Slopes Based on Artificial Intelligence

Abstract

To solve the problems of large prediction error, slow convergence speed, and poor generalisation ability of traditional models in predicting surface deformation of open-pit mine slopes, this paper proposes a new intelligent prediction model based on the Mayfly algorithm-optimised support vector machine (MA-SVM). In this method, the MA is used to optimise the SVM parameters to reduce the uncertainty of the model and avoid time-consuming parameter adjustment. To evaluate the proposed prediction model, real-world deformation data of the north slope of the Anjialing open-pit mine in Pingshuo city, China, are collected using the microdeformation monitoring radar and used to investigate the deformation prediction performance of the proposed method. The results of the analysis demonstrate that the proposed method is able to accurately predict the deformation of the surface of the mine slope and outperforms three existing popular methods, including SVM, genetic algorithm (GA)-SVM, and particle swarm optimisation (PSO)-SVM). The mean absolute error (MAE) of the proposed MA⁃SVM is 2.52 % while 6.56 %, 4.95 %, and 5.16 % for the SVM, GA-SVM, and PSO-SVM, respectively; the root mean square error (RMSE) of the proposed MA⁃SVM is 10.21 % while 30.79 %, 17.38 %, and 22.54 % for the other three methods. Because the proposed MA⁃SVM model is able to predict slope deformation using actual monitoring data, it is of practical importance in real-world applications for early warning on landslides of mine slopes.