Research On The Key Technology Of Tailings Pond Risk Warning Based On Improved LSTM Model

Tailings pond is mainly used to store tailings or industrial waste generated from mines.It is a major hazard source with high potential energy.Once a dam break accident occurs,it will cause immeasurable losses.Early warning of tailings dam break risk is of great significance for timely grasping the safety status of tailings dams,preventing dam break accidents in tailings dams,and ensuring the safety of people’s lives and property.Therefore,based on the long and short-term memory(LSTM)neural network,this paper conducts research on the risk early warning method for tailings dam failure,and the specific work includes:1.Analysis of risk factors for dam break of tailings pond.Analyzing the daily and weekly displacement changes of the tailings dam,the results show that the deformation data of the tailings dam have strong nonlinearity,frequent fluctuation,and no obvious periodicity.As the frequent fluctuation of tailings dam deformation is influenced by various external factors,Pearson correlation coefficient is used to analyze the correlation between risk factors of tailings dam failure,determine the main influencing factors of tailings dam failure,and reduce the input dimension of the model.2.Risk prediction of tailings dam break based on LSTM.Aiming at the nonlinear characteristics of deformation data of tailings dams,an LSTM model structure is designed and an LSTM dam break risk prediction model is constructed.Experiments have shown that this model can achieve the prediction of nonlinear displacement time series data,but the prediction effect is poor when there are strong nonlinear changes in displacement.Aiming at the strong nonlinearity and random volatility of deformation data of tailings dams,this paper constructs an EEMD-LSTM hybrid prediction model based on ensemble empirical mode decomposition.Through experimental comparative analysis,it is proved that using the EEMD method to stabilize displacement data can improve the prediction accuracy of LSTM model for dam break risk of tailings dams.3.Prediction of dam break risk of tailings pond under the coupling effect of multiple factors.Aiming at the problem of complex coupling between tailings dam break risk factors affecting the prediction accuracy of the model,a risk prediction method of tailings dam break based on dual attention is proposed,and an EEMD-DA-LSTM hybrid prediction model is constructed.Comparative experiments based on engineering examples show that this model is superior to the baseline model.The addition of dual attention improves the generalization ability and memory ability of the EEMD LSTM model for long-term sequence information.This paper takes a valley type tailings pond as an example to design a tailings pond risk early warning system.The system combines the dam break risk prediction model to achieve functions such as online monitoring information management and dam break risk early warning management.It can timely initiate an alarm response when the predicted displacement value of the tailings pond reaches the threshold,reminding the safety management personnel of the tailings pond to take dam break prevention measures as soon as possible.