Transformer-based composite material defect detection method and system

Abstract

The invention discloses a transformer-based composite material defect detection method and system, and the method comprises the steps: scanning a composite material based on Transformer, and collecting an ultrasonic signal of the composite material; dividing the ultrasonic signals into a training data set X and a verification data set Y; constructing a feature learning and classification model based on Transformer; and inputting the divided training data set X into a constructed feature learning and classification model based on Transformer to train the feature learning and classification model, and inputting the divided verification data set Y into the trained feature learning and classification model to obtain the defects of the composite material based on Transformer. According to the method, feature learning and classification are carried out on the collected ultrasonic signals of the composite material through a deep learning method, and accurate judgment on the defect type of the composite material is achieved.

Description

technical field [0001] The invention belongs to the technical field of composite material defect detection, and in particular relates to a Transformer-based composite material defect detection method and system. Background technique [0002] With the maturity of the process and the improvement of performance, composite materials have been developed from the initial application in the aerospace and military fields to the recent large-scale development in the industrial field, among which automobiles, electronics and other fields have become very promising application fields. However, various forms of defects will be formed during the production or use of composite materials. The existence of these defects will inevitably affect the performance indicators of materials, restrict the use of composite materials, and cause economic losses and safety problems. Therefore, accurate, fast, effective and timely detection of defective composite materials is of great significance to mode...

AI Technical Summary

Problems solved by technology

However, its limitations are also very large. Ultrasonic testing requires the surface to be tested to be relatively smooth. It is difficult to detect materials or workpieces with complex shapes. In addition, there are certain requirements for the thickness of composite materials.

[0004] The detection method based on the cyclic neural network must be processed in sequence according to the front and rear positions of the signal, which leads to two problems, one is the slow processing speed, and the other is that more and more signal information is lost as the forward propagation progresses, so The final extracted features are incomplete; although the detection method based on convolutional neural network can be executed in parallel, it cannot capture global features

[0005] The above reasons lead to the problem that the judgment of the defect type in ultrasonic testing is greatly affected by external factors, and the problem of difficult defect characterization has not been effectively resolved.

Images