1. Nondestructive evaluation of bonding quality of dual-layer coatings based on the multi-feature ultrasonic method.
- Author
-
Fan, Shaorui, Yuan, Maodan, Xu, Jianlin, Song, Yongfeng, Chen, Yan, and Ji, Xuanrong
- Abstract
• A multi-feature ultrasonic (MFU) method was proposed to evaluate the bonding condition of dual-layer coatings. • Parametric numerical studies were carried out to study the influence of stiffness at two interfaces on different ultrasonic features. • Stacked sparse autoencoder (SSAE) was applied to automatically extract multiple features from ultrasonic waveforms and spectra sensitive to the bonding quality of coating. • The correlation between feature sets extracted by SSAE and those traditional selected features was explored. • The proposed method is effective to detect both the topcoat debonding and primer debonding in the industrial coatings. Ultrasound has been extensively utilized to evaluate the bonding quality of coatings. However, current ultrasonic methods based on single temporal or spectral feature cannot accurately characterize the bonding quality within industrial coatings, especially for potential debonding at different layers. This paper proposed an ultrasonic method with multiple temporal and spectral features to evaluate the bonding condition of dual-layer coatings. Ultrasonic features were automatically extracted based on stacked sparse autoencoder (SSAE) to compare with the conventional selected features. Firstly, numerical simulations were carried out to analyze the variation of ultrasonic signals with different stiffnesses at topcoat and primer interfaces. Two temporal and three spectral features were manually extracted and show high sensitivity to different bonding conditions of the two interfaces. Meanwhile, a SSAE network was designed to automatically extract ultrasonic features. The high correlation between the manually and automatically extracted features demonstrates its effectiveness of automatic feature extraction via SSAE. Then, a 50 MHz ultrasonic pulse-echo system was applied to collect signals from 115 coating samples of laptop shell. Four sets of ultrasonic features were extracted via manual method and SSAE networks. Principal components analysis was then employed to reduce the feature space and support vector machine was applied to distinguish topcoat debonding and primer debonding from the intact coatings. The results show that the feature set including 64 temporal and spectral features extracted by SSAE exhibit the best classification with an accuracy of 95.652 %. This proposed multi-feature ultrasonic method can accurately assess the bonding quality of different interfaces in the multi-layer coatings. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF