Data Assimilation of PSO-Kalman Filter and InSAR/GNSS for High Precision Monitoring and Stability Evaluation of Key Dam Body

Tailings reservoir is an important part of the mine. The monitoring and assessment of the stability of their key structures have always been the focus of research by many scholars. In this experiment, Interferometric Synthetic Aperture Radar (InSAR) and the Global Navigation Satellite System (GNSS) were employed to obtain 30 sets of longitudinal deformation data of 6 feature monitoring points on the surface of a tailings reservoir in Anhui Province. Considering the complementarity of InSAR and GNSS in surface monitoring, it is proposed to calculate the optimal initial parameters Q and R of Kalman filter by using Particle Swarm Optimization ( PSO ), so as to construct the optimal PSO-Kalman data assimilation model to achieve bidirectional data assimilation between InSAR and GNSS data, improve the accuracy of InSAR data, and comprehensively analyze the stability of key dams. After that, the Long Short Term Memory (LSTM) recurrent neural network is utilized to conduct time series predictions on the assimilated data and the buried depth data of the saturation line of the monitoring points of the mine dam during the same period. The experimental results show that using PSO-Kalman filter for data assimilation, compared with the original InSAR data, the overall mean absolute error ( MAE ) is reduced by 52.7 %, and the overall root mean square error ( RMSE ) is reduced by 53.6 %. By using LSTM to predict the time series data, the RMSE of the key dam deformation data test output set is less than 2.5mm, and the RMSE of the saturation line buried depth test output set is less than 1.5mm. Finally, the existing data is used to train the dam stability evaluation model. Under the condition that the correct rate of the evaluation model is 88.89 % and the AUC value of the model stability is 0.88889, the LSTM prediction data and the stability evaluation model are used to analyze the future stability of the key structure. This paper innovatively proposes to jointly evaluate the stability of the key dam structure by combining the deformation data and the saturation line height data. It also presents the first relatively complete integrated research method of ' monitoring-prediction-evaluation ' for the key dam body of the tailings reservoirs. The experimental results show that this method can provide reliable data and technical support for the monitoring, stability analysis and evaluation of the key dam body of tailings reservoirs.