The XVIII International Conference on Mine Surveying

34 / 2023-03-29 15:16:46

Research on GNSS CORS mining area monitoring data processing based on the deep learning model

prediction of surface subsidence in mining area ; automated monitoring data ; CEEMDAN ; CNN-BiLSTM

Abstract: To make more effective use of the advantages of GNSS in long-term real-time monitoring of mining subsidence in mining areas, realize the deep mining of hidden features and hidden information in the data, and improve the prediction accuracy of surface subsidence in mining areas. In this paper, we propose a hybrid surface subsidence prediction method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), convolutional neural network (CNN) and A mixed mining surface subsidence prediction method with bidirectional long-term and short-term memory network (BiLSTM). Firstly, the morphological data differences between the working face subsidence and industrial monitoring areas are compared and analyzed. The CEEMDAN is used to reconstruct the elevation component of the station, and it is input into the CNN model to extract the hidden information of the component data. Then, the short-term prediction of station elevation data is realized by constructing the BiLSTM model. The experimental results show that this model has a better prediction effect than the traditional CNN, LSTM, and CEEMDAN-LSTM models.