Automatic deformation forecast and warning of a catastrophic landslide can effectively avoid significant casualties and economic losses. However, currently it has not come to a comprehensive forecast framework covering all the deformation stages of a landslide. Moreover, landslide deformation prediction possesses high error and false alarm rates. This work suggests a novel integrated framework of landslide deformation forecast and warning by coupling machine learning and physical models. The framework can relatively accurately predict all the deformation stages from creeping deformation to critical sliding and features 4 advantages. (a) The forecast indices are established by combining deformation and disaster-triggering characteristics to improve the prediction accuracy. (b) It leverages the advantage of C5.0 decision tree algorithm in knowledge interpretability to automatically extract deformation forecast criteria. (c) It capitalizes on the precision superiority of a graph convolutional network in time-series data learning to predict the four deformation stages from creeping deformation to rapidly accelerated deformation. (d) It utilizes the physical and mechanical bases of Morgenstern-Price method to forecast the critical sliding stage. Zhujiatang Landslide is a large-scale deep-seated soil landslide with significant deformation. It is selected as a case study because it has endangered 1,131 persons and may cause a direct financial loss of 100 million RMB. The validating and predicting Accuracy values attain 97.39% and 95.72%, respectively, and the Kappa values reach 0.91 and 0.93, respectively. The landslide will run out when it suffers from a rainstorm with cumulative rainfall of 79.57 mm and an earthquake with a horizontal coefficient of 0.04. A novel integrated framework of landslide deformation forecast and warning is suggested by coupling machine learning and physical models The framework can automatically extract forecast criteria and predict all the deformation stages from creepage to critical sliding Graphical convolutional network features an accuracy advantage and is for the first time employed to forecast deformation stages of a landslide

Constructing an interpretable model for the long-term deformation Structural Health Monitoring (SHM) of earthrock dams is of great significance for improving the safety state evaluation and monitoring effect. In this paper, a physics -data -driven model for the deformation SHM of earth -rock dams is proposed based on deep mechanism knowledge distillation. Firstly, the deterministic model is established based on the Finite Element Model (FEM) and outputs the hydraulic load component curve and aging component curve. Then a regression prediction model (HTSGAN) between influencing factors and deformation measurements at multiple measurement points is established based on the Graph Convolutional Network (GCN) and attention mechanism. Finally, the TeacherHydraulic -Time -Seepage Graph Attention Networks (T-HTSGAN) model is established based on the featurebased multi -teacher knowledge distillation using the knowledge of hydraulic loading physics and soil -rock creep physics of the FEM for mechanism constraints. The model effectively solves the problems of poor model interpretability and lack of physics knowledge constraints in previous earth -rock dam SHM models. The research results are applied to a project of a 185.5 -meter -high concrete -faced rockfill dam, and the predictive performance of the model is more effective and stable through the comparison of six baseline models. The comparative analysis of the component curves proves the effectiveness of the proposed knowledge distillation method for mechanism constraints and improves the interpretability of the neural network model. Therefore, the model is more suitable for engineering applications.