Fig. 1. Simulation results cloud diagram. (a) Surface temperature; (b) Surface deformation

Fig. 2. Node temperature change curve

Fig. 3. Data processing flow of defect detection using DOD feature extraction algorithms

Fig. 4. Data processing flow of defect detection using PCA feature extraction algorithms

Fig. 5. Flow chart of detection

Fig. 6. Ultrasonic thermography inspection system. (a) Detection system sketch; (b) Detection experimental platform

Fig. 7. Experimental specimens with cracks of different lengths

Fig. 8. Extraction effect of different feature extraction algorithms. (a) DOD amplitude diagram; (b) DOD phase diagram; (c) PCA first principal component; (d) PCA second principal component

Fig. 9. Feature clouds with different feature extraction algorithms. (a) Part #1 amplitude diagram; (b) Part #2 amplitude diagram; (c) Part #3 amplitude diagram; (d) Part #1 first principal component; (e) Part #2 first principal component; (f) Part #3 first principal component

Fig. 10. Micrograph of part’s notched edges (100 ×)

Fig. 11. Definition of the crack region and the normal region

Fig. 12. Effect of modulation parameters on feature contrast ratio. (a) Start frequency affects DOD amplitude contrast ratio; (b) Start frequency affects PCA feature contrast ratio; (c) Termination frequency affects DOD amplitude contrast ratio; (d) Termination frequency affects PCA feature contrast ratio; (e) Scanning period affects DOD amplitude contrast ratio; (f) Scanning period affects PCA feature contrast ratio