基于FCM-STGCN模型的露天矿边坡形变时空预测方法OA

Open-pit coal mining easily induces slope instability and deformation,to investigate an ef-fective monitoring and prediction methods,an open-pit mine in Qinghai Province was taken as the study area.Small baseline subset interferometric synthetic aperture radar(SBAS-InSAR)was used to obtain time-series slope deformation data for the study area,and monitoring points with different deformation magnitudes were clustered into subregions by fuzzy C-means(FCM).The optimal number of clusters was determined by the DB index and silhouette coefficient.Under the optimal number of clusters,de-formation points with similar deformation characteristics were divided into the same subregion by the FCM algorithm.The FCM-STGCN model was then constructed by combining the clustering algorithm with a spatiotemporal graph convolutional network(STGCN)to predict slope deformation trends.The results show that under three different training-set and prediction-set ratios,the FCM-STGCN model a-chieves the best prediction performance compared with the long short-term memory(LSTM)model and the FCM-LSTM model.Under the optimal scenario,the root mean square error,mean absolute error,weighted mean absolute percentage error,and coefficient of determination of the FCM-STGCN model are 4.2 mm,3.1 mm,6.4%,and 0.996,respectively.The minimum difference between Moran's I of the predicted values and that of the true values is as low as 0.007,indicating a high consistency in spa-tial distribution characteristics.This study achieves high-precision prediction of slope deformation and can provide a theoretical basis for slope deformation prediction and disaster prevention in open-pit coal mines.