Defect Detection and Identification of Thin Carbon Fiber Composites Based on Multi‐Feature Fusion

ABSTRACT Carbon fiber composite materials are challenging to manufacture and costly; moreover, internal defects can precipitate catastrophic structural failure. In this study, ultrasonic phased array inspections were conducted on carbon fiber reinforced polymer (CFRP) laminates with a nominal thickness of 3 mm. A multi‐domain signal feature extraction method, based on kernel principal component analysis, was proposed and integrated with a particle swarm optimization –optimized support vector machine. This approach achieved quantitative detection of defects in thin‐walled carbon fiber components. The recognition rate for delamination defects at all investigated depths reached 100%, while the recognition rate for defects of different sizes exceeded 95%. These results demonstrate the method's capability for effective and intelligent recognition of delamination defects at different depths and sizes in carbon fiber composites. This study provides novel methodologies and insights for the application of artificial intelligence in industrial non‐destructive testing.