Cable-stayed bridge cable damage positioning method based on modified reverse transmittance nerve network

Abstract

Cable-stayed bridge cable damage location method based on improved back-propagation neural network is a key solution to how to effectively improve the BP neural network for cable-stayed bridge cable damage, so as to use it for cable-stayed bridge cable damage location. damage location, and establish a cable-stayed bridge cable damage location method based on the improved BP network. The present invention proposes to comprehensively adopt the "Bold Driver" (blind drive) technology, increase the momentum item, simulated annealing algorithm, and random hill-climbing algorithm to combine traditional BP The algorithm is improved. This method has the advantages of accelerating convergence, improving learning efficiency, avoiding the learning process from falling into local minimum, and can reasonably determine the network topology.

Description

technical field [0001] The invention is a method applied to structure damage location, especially a method applied to cable-stay damage location of cable-stayed bridges. Background technique [0002] BP (Back Propagation, backpropagation) neural network technology has been applied to structural damage identification and related research since the 1990s. Wu et al. [1] A numerical simulation of damage identification is carried out on a three-layer frame structure using a single hidden layer BP network. Elkordy et al. [2] The damage diagnosis of a 5-story steel frame is carried out by using BP network. Hanagud and Luo [3] Damage identification of composite panels using frequency response function data based on BP neural network; Luo and Hanagud [4] In the subsequent research, the dynamic learning rate was introduced into the above research to speed up the convergence of the neural network. For glass fiber composite beams, Jenq and Lee used the change of the first 4 order ...

AI Technical Summary

Problems solved by technology

However, the current research work generally has deficiencies in this aspect, especially for the analysis of damage characteristics of long-span cable-supported bridge structures.

[0008] 2) BP neural network is the most widely used neural network in engineering. However, BP network itself has some inherent defects, such as slow learning convergence, easy to fall into local minimum, and difficult to determine the network structure

In the application of structural damage recognition, some researchers have paid attention to the problem of slow learning convergence of BP neural network, and considered some improvement measures. The learning process of the neural network shows the unstable phenomenon of good and bad, which also makes it difficult to reproduce some research results; in addition, the neural network is a structured learning tool, and its topology has a decisive impact on the performance of the network, but The determination of the optimal topology is a thorny issue, not as many neurons as described in some literature, the better

However, in the application, the topology of the BP network is often determined subjectively by the researchers, which is obviously inappropriate

Images