Multi-scale analysis of GNSS monitoring data of open-pit mine slopes based on EEMD-LSTM model network
GAO Han
WANG Keke
MA Ke
WANG Yong
ZHANG Kai
Abstract：The EEMD-LSTM model is proposed to enhance the prediction accuracy and stability of GNSS monitoring data for open-pit mine slopes,addressing challenges in traditional methods such as mode mixing,noise interference,and insufficient capture of long-term dependent features in handling complex nonlinear time series data.This model integrates ensemble empirical mode de-composition(EEMD)and long short-term memory(LSTM)networks.Firstly,the EEMD algorithm adaptively decomposes raw monitoring signals into multiple intrinsic mode functions(IMFs),effectively separating noise and sudden anomaly information to re-solve the mode mixing issue inherent in traditional empirical mode decomposition(EMD).Secondly,the LSTM network extracts temporal features from decomposed IMFs and enhances long-term dependency modeling through its gate control mechanism.Lastly,an improved data isolation procedure(involving repeated decomposition and independent predictions)prevents information leakage,while multi-dimensional error evaluation metrics(MAE,MAPE,RMSE)validate model performance.Experimental validation util-ized GNSS monitoring data from an open-pit mine in Heilongjiang Province,processing 6 727 displacement datasets to predict 30-day deformation trends.Results demonstrated that EEMD successfully isolated high-frequency noise(IMF1,IMF2)and low-fre-quency trend components(IMF8,IMF9),significantly reducing anomaly-induced prediction interference.The model exhibited op-timal performance in 3D displacement prediction,achieving the highest precision in the z-direction(RMSE=0.017),though systemat-ic bias in the y-direction requires further optimization.The improved isolation process notablely reduced errors in 2D/3D predictions,confirming the model's resistance to information leakage.Correlation with slope stability calculations showed safety factor improve-ment to 1.21 after treatment,with monitoring point displacements stabilizing within 50 mm.By combining signal decomposition with deep learning,the EEMD-LSTM framework enables multi-scale analysis of complex time series features,providing a high-pre-cision dynamic prediction tool for slope disaster early warning systems in open-pit mining environments.
Keywords：EEMD-LSTMslope monitoringGNSSslope stability analysisslope managementlandslide
Publication Date：2025-09-30
Online Publishing Date：2025-09-12(First online date of this platform, not the publication date of the document)
Pages：8( 44-51 )