Foundation pit disaster prediction and early warning method based on LSTM and deep residual neural network

Abstract

The invention belongs to the technical field of time sequence model processing and geotechnical engineering disaster prediction in artificial intelligence, and particularly relates to a foundation pit disaster prediction and early warning method based on LSTM and a deep residual neural network, and the method comprises the following steps: S1, arranging a measurement point position at a key position of a measured foundation pit, the measurement point position is used for measuring data of underground water level, support axial force, diaphragm wall stress and deep horizontal displacement of a point location; and S2, collecting the data, and preprocessing the data; the method has the beneficial effects that the influence of different characteristics of the prediction data on the prediction precision is comprehensively considered, and the acquired data are predicted by using different neural network models, so that the prediction precision is improved; the foundation pit disaster prediction is carried out by combining the LSTM and the residual network model, compared with a single model, the prediction precision is greatly improved, the foundation pit monitoring data prediction is improved, and the occurrence of the foundation pit disaster is prevented and avoided to a certain extent.

Description

technical field [0001] The invention relates to the technical field of time series model processing in artificial intelligence and geotechnical engineering disaster prediction, in particular to a foundation pit disaster prediction and early warning method based on LSTM and deep residual neural network. Background technique [0002] Traditional gray theoretical models, such as gray DNGM (1,1), TPGM (1,1) and ARIMA prediction models, are affected by many aspects and are not accurate enough for the prediction of foundation pit disasters affected by multiple factors; With the development of data, the method of using deep neural network to process and predict disaster information has gradually emerged. [0003] The accuracy of the traditional BP neural network for processing time-related data is not high, and the deeper the neural network is, the accuracy rate does not improve much, and it also increases the time for training the model; the cyclic neural network can handle the sa...

AI Technical Summary

Problems solved by technology

[0005] Therefore, the purpose of the present invention is to provide a foundation pit disaster prediction and early warning method based on LSTM and deep residual neural network, to predict the impact of different characteristics of data on prediction accuracy, and to use different neural network models for each type of data collected. , which improves the prediction accuracy; combining LSTM and residual network model for foundation pit disaster prediction, compared with a single model, greatly improves the prediction accuracy, helps to improve the prediction of foundation pit monitoring data, and prevents and avoids foundation pit disasters to a certain extent. The occurrence of pit disasters solves the problem that the traditional BP neural network has low accuracy in processing time-related data, and the deeper the neural network is, the accuracy rate does not improve much, and it also increases the training time of the model; The network can handle time-related data very well, but its memory capacity is insufficient, it cannot handle data that is too long apart, and it is also prone to gradient explosion and gradient disappearance.

Images