Detection of Interface Debonding Defects Based on Principal Component Analysis of Multiple Energy Features and K-Means Clustering Analysis

The coin-tap detection (low-frequency resonance acoustic detection) method has been widely utilized in detecting the interface debonding defect of bonding structure. The frequency domain and energy features of the signals change with the presence or absence of the defect. The recognition error rate of debonding defects can be high when the computer automatically judges defects, which may be caused by the diverse types of defects, complex spectrum variations, and the influence of factors such as uneven man-made percussions, environmental noise, etc. In order to solve the above problems, a detection method for identifying the interface debonding defect in multilayer adhesive structures is proposed based on multiple energy features principal component analysis and K-means clustering analysis. In this method, the multiple energy features of low-frequency resonance ultrasonic signals are extracted while the dimensionality the signal energy features are reduced by employing principal component analysis method. Besides, K-means clustering on the principal components of the signal energy features is performed. Furthermore, the relationship between the acoustic signal and the energy features is established, and the initial clustering center of debonding detection is obtained. Data samples has been selected for verification. The results show that the detection accuracy of the method reaches 96%, which is effectively improved compared with the traditional solutions.