Research On Slope Displacement Prediction Model Of Open Pit Coal Mine Based On Machine Learning

The accidents such as wall caving and landslide caused by slope instability are potential safety problems of open-pit coal mine.As one of the basis for judging slope instability,predicting its future change trend is of practical significance to ensure the safe production of open-pit coal mine.Aiming at the research on the slope displacement prediction model of open-pit coal mine,based on the slope displacement monitoring data of the south end of Tianchi energy South open-pit coal mine,this thesis establishes the support vector machine(SVM)slope displacement prediction model,the long shortterm memory neural network(LSTM)slope displacement prediction model and the modified model based on Markov chain(MC).The prediction results are compared and analyzed,and the slope displacement prediction effect is better.The main research contents are as follows:(1)This thesis analyzes the current situation and influencing factors of the south side slope displacement of the south open pit coal mine.According to the data of the slope displacement monitoring system,aiming at the frequently occurring area of the slope,the displacement data is selected based on the characteristics of stable overall change trend,large sudden change in local position and unstable overall change trend.The data can better represent the overall slope displacement and can be used to establish the slope displacement prediction model.(2)Support vector machine and long short-term memory neural network are used,they had the characteristics of good generalization and high accuracy.The displacement data collected by slope radar were took as the input value,SVM slope displacement prediction model and LSTM slope displacement prediction model are established respectively.The mean absolute error(MAE)and root mean square error(RMSE)are used as the evaluation indexes of model accuracy.The smaller the MAE and RMSE values,the smaller the error between the predicted displacement value and the real displacement value,and the more accurate the prediction result is.The four groups of slope displacement prediction accuracy values RMSE and MAE obtained by SVM model are 5.4 ～ 11.5mm and 5.0 ～ 10.8mm respectively,so the prediction effect is poor.The four groups of slope displacement prediction accuracy values RMSE and MAE obtained by LSTM model are 0.6 ～ 3.5mm and 0.4 ～ 3.0mm respectively,which are better than SVM model.(3)The Markov chain correction theory is introduced to correct the predicted value of slope displacement of SVM model and LSTM model.The four groups of slope displacement prediction accuracy values RMSE and MAE obtained by MC-SVM model are 2.8 ～ 4.6mm and 2.2 ～ 3.2mm respectively,which is up to 64.5% higher than that of SVM model,and can meet the actual accuracy requirements.The four groups of slope displacement prediction accuracy values RMSE and MAE obtained by MCLSTM model are 0.4 ～ 1.5mm and 0.7 ～ 1.0mm respectively,which is up to 75.1%higher than that of LSTM model,and still has good correction effect on the basis of high original accuracy.The displacement prediction results based on the modified Markov chain model show that MC-LSTM model has the best effect on slope displacement prediction.(4)The MC-LSTM model is applied to the case analysis of wall caving.The displacement increment is used as the judgment basis to analyze the time of wall caving.The wall caving time point recorded on site at 9:00 on May 25,2021 is selected as the reference.The result of sudden change of displacement increment predicted by the model is between 6:00 on May 25 and 12:00 on May 26,2021,which is consistent with the actually recorded wall caving time.In addition,MC-LSTM model has good prediction effect for slope displacement with timeliness of about 7 days.