Subway settlement predication method based on empirical mode decomposition and BP neural network

Abstract

The invention discloses a subway settlement predication method based on empirical mode decomposition and BP neural network. The method includes S1, utilizing EMD decomposition for decomposing non-stable settlement signals x(t) into fluctuations of different frequencies and scales; S2, establishing BP neural network models of different frequency bands; S3, selecting parameters of the BP neural network models by utilizing a cross validation method; S4, reconfiguring a predication sequence of original signals by utilizing predication values of the components. According to the invention, a problem of modal aliasing of an EMD single model and problems of local minimization and low convergence rate of the BP neural network are solved. Compared with the single model, non-linear data processing precision is improved greatly in the invention.

Description

technical field [0001] The invention relates to a subsidence prediction method during subway operation, in particular to a subway subsidence prediction method based on empirical mode decomposition and BP neural network. Background technique [0002] On-site monitoring and measurement of subway is an important part of subway construction. In the process of subway construction, whether it is internal or corresponding to the deformation of the upper part, it is a complex nonlinear and non-stationary dynamic system. How to extract the inherent things from the obtained data has become a difficult task for us, and the essence of this problem is data mining. Based on this, many researchers have conducted various analyzes on the measured data of subway deformation, such as artificial intelligence analysis and conventional analysis, and have achieved fruitful results in this field. Empirical Mode Decomposition EMD (Empirical Mode Decomposition) is a very good method for dealing with...

AI Technical Summary

Problems solved by technology

Empirical Mode Decomposition EMD (Empirical Mode Decomposition) is a very good method for dealing with nonlinear and non-stationary signals. Adaptive wavelet decomposition method, but when processing data, the problem of mode aliasing often occurs

BP (Back Propagation) neural network theory has the characteristics of self-organization, fault tolerance, self-adaptation, etc. It is suitable for dealing with problems with unclear inference rules, fuzzy information, and complex background knowledge, but there are problems of local minimization and slow convergence.

Images