Abstract: The stability of slopes in open-pit mines directly affects the safety production of mines.Slope displacement monitoring data are important parameters to characterize the development process of slope deformation,and the early warning of landslide can be realized by analyzing and studying the monitoring data.Taking a mining slope of Hebei Jinyu Zenith Cement Co.,Ltd.as the study example,based on the displacement data collected from the monitoring points,a multi-factor slope displacement prediction model based on Conv-LSTM is established.The Long short-term memory network (LSTM) is used to extract the timing information in the displacement time series,and the deep features hidden in the displacement series are extracted by the convolutional layer.Aiming at the problem that the convolutional layer is not sufficient to extract the internal features between the data,the self-attention mechanism is introduced to fully extract the relationship features between the slope displacement data by using the self-attention mechanism.The test results show that the Conv-LSTM slope displacement prediction model with self-attention echanism has higher prediction accuracy and better correlation with the original displacement sequence,which can more truly reflect the deformation law of the slope.

Key words: open-pit slope,neural network,displacement prediction,self-attention mechanism