Robust Principal Component Thermography for Defect Detection in Composites

Area of Science:

• Non-destructive testing
• Materials science
• Signal processing

Background:

• Pulsed Thermography (PT) data often suffers from noise, hindering defect detection.
• Principal Component Analysis (PCA) is effective for defect enhancement in PT but sensitive to data corruption.
• Robust PCA (RPCA) offers an alternative for handling noisy data by separating low-rank and sparse components.

Purpose of the Study:

• To introduce Robust Principal Component Thermography (RPCT) for improved defect detection in PT data.
• To evaluate RPCT's performance against standard Principal Component Thermography (PCT).
• To quantify improvements in defect contrast and segmentation using RPCT.

Main Methods:

• Application of RPCA to PT data, termed RPCT.
• Comparison of RPCT with PCA-based PCT on CFRP samples with artificial defects.
• Quantitative evaluation using Contrast-to-Noise Ratio (CNR) and Jaccard similarity coefficient.

Main Results:

• RPCT achieved an average 40% higher CNR compared to PCT.
• RPCT showed a slightly higher Jaccard index (0.7395) than PCT (0.7010).
• PCT was 11.5 times faster than RPCT in computational time.

Conclusions:

• RPCT demonstrates superior defect detectability and segmentation potential in PT compared to PCT.
• Further research is needed to optimize RPCT's computational efficiency and test its efficacy on diverse materials.