A stress detection method for metal components based on eddy current thermography

Abstract

The metal components are subjected to various complex stresses during processing and use, which may lead to material failure in severe cases. Therefore, the detection of stress in these components has practical and important application value. In this study, a theoretical model of stress detection was constructed based on the principle of eddy current thermography, and the expression between temperature change and stress was deduced. Especially in the cooling stage, the relationship expression between stress and temperature change was obtained. By establishing an eddy current thermography simulation model, the change of the temperature response of the sample surface was analyzed, and a stress detection method based on the relationship between time-temperature change law-stress distribution was developed. An eddy current thermography stress detection system was built, and the stress detection equation was constructed based on the experimental data to verify the effectiveness of the method. Then, to eliminate the interference of factors such as excitation parameters, the stress detection method was optimized by using cooling data, and its measurement error is 5.37%, which further improved the stability and accuracy of stress detection. Finally, according to the idea of deep learning, a convolutional neural network (VGG19 and ResNet101) assisted method was used to evaluate the stress state. By training the temperature thermal aberration map, the stress distribution state of the structure can be quickly and effectively evaluated, the test accuracy can reach 99%, and the accuracy of the cooling data set after excluding the influence of factors such as eddy current excitation is higher. It also verified the accuracy of the stress detection method based on the time-temperature change law-stress distribution proposed in this paper. In addition, it provided technical support and data reference for subsequent quantitative stress detection and damage assessment of more complex structures.