Heterogeneous information phase space reconstruction and stability prediction of filling body-surrounding rock combination

The traditional research believes that the filling body can effectively control the stress concentration, while ignoring the problem that the stress distribution of the filling body-surrounding rock combination under high stress conditions is complex and changeable, and the stability is unknown. The current monitoring data processing methods can not fully consider the complexity of monitoring objects, the diversity of monitoring methods, and the dynamics of monitoring data. To solve this problem, this paper proposes a phase space reconstruction and stability prediction method for heterogeneous information of backfill-surrounding rock combination. The three-dimensional monitoring system of large-area filling body-surrounding rock combination in Longshou Mine was constructed by using drilling stress, multi-point displacement meter and inclinometer, and the multi-information such as stress and displacement of filling body-surrounding rock combination was continuously obtained. Combined with the average mutual information method and the false nearest neighbour point method, the phase space of the heterogeneous information of the filling body-surrounding rock combination is constructed. In this paper, the distance between the phase point and its nearest point is used as the index evaluation distance to evaluate the stability of the filling body-surrounding rock combination. The evaluated distances show a high sensitivity to the stability of the filling body-surrounding rock combination. The new method is applied to calculate the time series of historical evaluated distances for 12 measuring points at Longshou Mine. The moments of mutation in these time series are at least 3 months ahead of the roadway return dates. In the evaluated distance prediction experiments, the ARIMA model has a higher prediction accuracy than the deep learning models (LSTM and Transformer), and its root-mean-square error distribution of the prediction results peaks at 0.26 and outperforms the no-prediction method in 70% of the cases.