Prediction of resurrection deformation of old landslides at tunnel entrances based on self excavation multi-scale recognition
In order to achieve high-precision prediction of landslide deformation,taking the revival of landslides in the tunnel entrance section as the background,the modal identification of landslide deformation data is carried out using variational mode decomposition,and the trend displacement is proposed by combining the moving average method.The landslide deformation data is decomposed into trend displacement and random displacement.Based on the decomposition of landslide deformation data,a trend displacement prediction model was constructed using the predator algorithm and bidirectional long short-term memory.Then,on the basis of data self excavation and multi-scale analysis,random displacement prediction was achieved through BP neural network or support vector machine.The results showed that the deformation characteristics of the landslide at the tunnel entrance was significant due to the construction of the tunnel entrance section,and the prediction results from three monitoring points were statistically analyzed.The EMAP value of the prediction results ranged from 2.01％to 2.05％,and the Tt value ranged from 162.45 ms to 185.45 ms,which not only had excellent prediction accuracy,but also had strong stability,and could effectively grasp the deformation law of the landslide,laying a certain theoretical foundation for subsequent landslide prevention and control.
landslidedata decompositiondecomposition of variational patternsdeformation prediction
万方数据
维普
