首页|Improved feature extraction of guided wave signals for defect detection in welded thermoplastic composite joints
Improved feature extraction of guided wave signals for defect detection in welded thermoplastic composite joints
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Elsevier
? 2022 Elsevier LtdMachine learning has shown to be a powerful diagnostic tool for interpreting guided wave signals towards modeling damage indexes. However, few works propose the use of machine learning techniques to build an automatic health diagnosis framework for composite welding. This work aims to devise an automatic diagnostic tool based on the propagation of ultrasonic guided waves for composite weld defect detection. Different feature extraction methods and machine learning modeling paradigms are investigated to highlight the most suitable strategy for performing the task. The support vector machine with autoregressive features resulted in the best overall performance. The accuracy was improved by 123.04% compared to other methods present in the literature, and satisfactory computational complexity was achieved. We have extended the previously reported results, showing that feature engineering enhances the model's effectiveness, making it possible to devise an automatic diagnosis framework that uses multi-frequency data and enables small data evaluation using resampling.
NSTL
Elsevier
