PSI - Issue 37

8

Alexey Tatarinov et al. / Procedia Structural Integrity 37 (2022) 453–461 Alexey Tatarinov et al./ Structural Integrity Procedia 00 (2019) 000 – 000

460

Fig.8. Evaluation of factors of interest W and Th W in control specimens W2, 3 mm and W2, 25 mm by pattern recognition method. Upper row presents recognition result in full scale; lower row depicts enlarged fragments in upper plots marked by white rectangles, where crosses denote projected values and the brightest areas reveal real solutions of recognition as intersection of several decision rules.

The existence of several distant solutions, where only one of them is close to the true value as shown for the specimen W1 3mm in Fig. 9, suggests further improvement of the method. Further measures to improve evaluation accuracy may include choosing only the most informative decision rules and introducing the concept of weights for each rule. In the present study, formal statistical criteria for pattern recognition have been applied in the processing of ultrasonic signals without particular reference to the physical nature of the propagation of ultrasonic waves. The introduction of new decision rules based on the parameters of the propagation of ultrasonic waves in the time domain can improve the specificity.

Fig.9. Evaluation of factors of interest W and Th W in control specimens W1, 3 mm. Projected value is marked by cross; white arrows denote 3 probabilistic solutions as maximal intensity of intersection of several decision rules. 4. Conclusion An approach to assess two factors of interest - the degree of concrete degradation and the depth of deteriorated surface layer of concrete based on pattern recognition is proposed. 2D spatiotemporal ultrasonic waveform profiles at two ultrasonic frequencies were the source material for the mathematical model. Verification of the method by substituting data on a control series of specimens into the mathematical model built on a grid of specimens of a training series allowed a realistic determination of a specimen from the control series for both factors of interest. Application of only formal statistical criteria is not sufficient for reliable recognition in all cases requiring for the further improvement of decision rules, particularly, by introducing weights for individual rules and applying additional rules based on wave propagation in the time domain.