Research And Application Of Rail Ultrasonic Flaw Detection Technology Based On BP Neural Network

In recent years,with the rapid development of electronic and computer technology,the field of nondestructive testing technology is also escalating in order to achieve the digital and intelligence,with a view to improve the accuracy of test results and diagnostic stability.In order to pursue low-cost testing equipment and methods,ultrasonic testing technology has been gradually applied in the field and has been greatly developed.At the same time,due to the development of rail transportation,the maintenance and testing of active rail are also being paid more attention.In this context,this thesis studies the damage detection and recognition of rails,using the digital ultrasonic flaw detection technology to collect the rail damage information and using the mature and robust algorithm to identify rail defects.In this thesis,the theory and method of ultrasonic flaw detection are analyzed and researched.The data obtained from the scene are denoised by using the wavelet decomposition method,then reducing the dimension and using a certain network to identify the diagnosis of the rail,and its application in the railway system,to get better detection.The main contents are as follows:1)Since the acquired ultrasonic data contains noise,it needs to be denoised firstly,and the noise has non-stationary property.Because the Fourier transform can not analyze the non-stationary signal,then we choose the wavelet to denoise and get useful information,in order to get more accurate and clear study.2)After the noise reduction,the data volume is too large.Through PCA and LDA method,the dimensionality reduction method can be used to obtain eigenvalues,and the fault will be better carried out.Then the result of the failure will be better for inter-class dispersion and within the polymerization.This can also make the establishment of follow-up network simple and easy and the calculation of time will be greatly reduced.3)BP neural network was used to study the data.The reduced dimension data was used as the input of the network.Then BP model was built based on MATLAB simulation platform,then trained and tested,and the parameters of the network were selected.In order to avoid the problem that the network may fall into the local minimum when the weights are modified,L-M algorithm is proposed4)The method of rail ultrasonic flaw detection is applied to the railway platform to identify the type and position of the damaged rail.