Defect detection of carbon fiber deflectors based on laser infrared thermography and experimental modal analysis

• Main Authors: Zhou, Guangyu, Ong, Zhi Chao, Zhang, Zhijie, Yin, Wuliang, Chen, Haoze, Ma, Huidong, Fu, Yu
• Format: Article
• Published: Elsevier 2024
• Subjects: TJ Mechanical engineering and machinery; TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.um.edu.my/46783/; https://doi.org/10.1016/j.ymssp.2024.111763
• Institution: Universiti Malaya

In this study, we integrate two techniques, laser infrared thermography (LIT) and experimental modal analysis (EMA), to inspect carbon fiber deflector plates using a combination of online and offline modes. In the LIT phase, we employed line laser scanning to detect the defects in each row, pinpointed their locations, and compiled a dataset correlating the presence of defects with the peak normalized temperature. In the EMA phase, separate tests were conducted at the front and back of the carbon fiber deflector. The front test aims to capture the modal information of the sample, whereas the back test aims to explore the effects of local defects. To establish the relationship between defects and vibration amplitude efficiently, we propose a sensitive modal identification method based on Long Short-Term Memory (LSTM). Using this method, we constructed a dataset comprising defect parameters corresponding to the peaks of the differential amplitude response in the sensitive modes. Finally, we developed an LIT-EMA-support vector machine (LE-SVM) defect parameter prediction model based on an SVM. The results demonstrated that the prediction accuracy of the bivariate model surpassed that of the univariate model. In particular, the R2 value of the evaluated results for defect depth reached 0.99959, with a maximum prediction error of only 0.032 mm. Similarly, the R2 value of the evaluated results for the internal defect size reached 0.99969 with a maximum prediction error of only 0.18 mm. These experimental findings validate that the integration of the two methods not only enables comprehensive structural health inspection of carbon fiber deflector plates from overall to localized structural health detection but also yields more precise parameter evaluation results.