Beam structure defect noncontact quantitative diagnosis method based on convolutional neural network

Abstract

The invention discloses a beam structure defect noncontact quantitative diagnosis method based on a convolutional neural network. The diagnosis method comprises the following steps: using a laser Doppler noncontact method to measure the weak vibration time-domain signals of multiple points of a beam under random excitation; adding the random vibration signals of any three points, which are not in a same line; carrying out rapid Fourier transform to generate a frequency spectrum; converting the frequency spectrum from continuous wavelet transform (CWT) into a frequency-size distribution diagram. The quantitative evaluation result of beam structure damage can be obtained by inputting frequency-size distribution diagrams of vibration data of any three points, which are not in a same line and on the surface of a beam structure, into a trained convolutional neural network (CNN). The CNN can evaluate the bean defect level.

Description

technical field [0001] The invention relates to a non-contact quantitative diagnosis method for beam structural defects based on a convolutional neural network, belonging to the fields of laser detection and measurement and image processing. Background technique [0002] Beam structures are widely used in mechanical engineering, civil engineering, aerospace and other fields. During service, due to the influence of ambient temperature, long-term mechanical loads and various corrosion conditions, it will inevitably produce defect damage, resulting in beam structure Damaged or even fractured, resulting in huge property losses and immeasurable catastrophic accidents. Rapid quantitative diagnosis of beam structures is increasingly becoming an important research topic. The traditional beam defect identification method needs to add sensors to the beam structure for measurement. The additional mass causes large errors in the measurement results of thin-walled small beams, and the i...

AI Technical Summary

Problems solved by technology

The traditional beam defect identification method needs to add sensors to the beam structure for measurement. The additional mass causes a large error in the measurement results of thin-walled small structure beams, and the information of the measurement points cannot be fully utilized.

Images