Fig. 1. Experimental setup

Fig. 2. Multi-feature fusion benchmark model for stress prediction

Fig. 3. Laser ultrasonic signal. (a) Laser signal and original laser ultrasonic signal; (b) laser ultrasonic signal after filtering

Fig. 4. Envelopes of laser ultrasonic signal from time and frequency domains. (a) Original signal and envelope of laser ultrasonic in time domain; (b) frequency spectrum and envelope of laser ultrasonic in frequency domain

Fig. 5. Stress dependent laser ultrasonic wave. (a) Laser ultrasonic signals under different stresses; (b) relationship between delay time of ultrasonic wave and applied stress

Fig. 6. Cumulative contribution of principal components of laser ultrasonic features

Fig. 7. Outputs of stress prediction by using different kernel functions. (a) Linear kernel function; (b) polynomial kernel function;

Fig. 8. R² and RMSE of stress prediction results by using different kernel functions

Fig. 9. Stress prediction results by using different regressive stress prediction models. (a) Single feature; (b) multiple linear regression; (c) Bayes; (d) random forest; (e) SVM

Fig. 10. Prediction results at different stress levels by using different regressive models

Fig. 11. R² and RMSE of stress prediction results in training and test sets by using different regressive stress prediction models. (a) R²; (b) RMSE