Speaker
Description
Vibrothermography (VT) has emerged as an effective non-destructive testing technique for detecting barely visible impact damage (BVID) in composite materials. This work investigates vibrothermographic inspection of impact damage in carbon-based composites using fixed-frequency and frequency-sweeping ultrasonic excitation, enhanced by machine learning algorithms. Experiments were conducted on specimens with low-energy impact damage down to 5 J. Ultrasonic excitation was applied via piezoelectric transducers across multiple frequency ranges, while infrared thermography captured thermal responses. Machine learning models were trained on thermal image datasets to automate defect detection and classification, significantly improving detection accuracy and reducing inspection time. The results demonstrate that frequency sweeping, combined with AI-driven image analysis, substantially improves defect activation compared to fixed-frequency excitation by exciting local defect resonance modes. ML algorithms enhanced signal-to-noise ratio through intelligent feature extraction and enabled prediction of defect severity. Narrowband sweeps proved most effective, providing superior thermal contrast. This study validates the synergy between sweep-based vibrothermography and artificial intelligence for reliable, automated detection of low-energy impact damage in composite materials.
