Predicting the settlement of Urumqi subway based on wavelet denoising and BP neural network

Abstract

In this study, we propose a model to accurately predict the ground subsidence caused by subway excavation using the wavelet denoising model and BP neural network. First, we develop an optimal denoising model by comparing and analyzing the denoising effect of different wavelet denoising parameters. The model is used to reduce the noise of the monitoring data. Then, we utilize BP neural network to develop a prediction model in which the proposed denoising model is used. Finally, we apply the proposed model to Urumqi subway. The results demonstrate the rationality and accuracy of the proposed model.

Prediction model of land surface settlement deformation based on improved LSTM method: CEEMDAN-ICA-AM-LSTM (CIAL) prediction model

“…Traditionally, theoretical analysis and calculation, empirical formula fitting, numerical simulation, and physical model experimentation have been employed. In recent years, artificial intelligence algorithms, including BP neural network [ 1 ], recurrent neural network (RNN) [ 2 ], support vector machines [ 3 , 4 ], grey prediction model [ 5 ], and random forest method [ 6 ], have been widely used in various fields due to their impressive speed and accuracy. Backpropagation neural network (BPNN) [ 7 ], as a representative of the prediction algorithm of surface subsidence, has become mainstream.…”