Research on coal rock seepage path prediction method by fusing LSTM network

Coal-rock seepage paths are directly related to the risk of water damage in mines, which is a key issue to ensure the safety of coal mines. Traditional numerical simulation methods are difficult to accurately predict the nonlinear evolution of seepage paths under complex working conditions. In this study, we propose a coal-rock seepage path prediction method by integrating LSTM network, and use the time series data of pore pressure, stress, and seepage rate to build a deep learning model to improve the prediction accuracy and timeliness. By introducing physical constraints, a Physics-informed LSTM model is constructed to achieve physically consistent modelling of seepage paths. In the experimental and field data validation, the method outperforms the traditional model in terms of RMSE, path deviation, R², etc., and shows good generalisation ability and application prospect. The study provides new ideas for coal mine water damage prevention and control, and promotes the integration of deep learning applications in geotechnical and hydrological fields.