PSI - Issue 37

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

459

7

Fig.6. Examples of ultrasonic spatiotemporal profiles at 50 and 100 kHz in solid specimen S and 2-layer specimen W3 12mm: abscissa is ultrasonic time; ordinate is step of profiling (distance between transducers), amplitude of ultrasonic signal is coded by brightness.

3.2. Decision rules 26 decision rules for 13 criteria described in 2.3.1 were calculated for 2 ultrasonic frequencies, 50 and 100 kHz, using 2D spatiotemporal ultrasonic profiles acquired according to the specimens’ grid. Decision rules presented topological surfaces with expressed gradients and local extremums. Examples of a few characteristic decision rules are shown in Fig.7. Further verification of the model included testing of control specimens with intermediate values of the factors using the algorithm described in 2.3.2.

Fig.7. Examples of decision rules for selected criteria presented as 2D brightness plots.

3.1. Verification of the proposed evaluation method The set of specimens was divided into the training set (n=21) and control set (n=6). The control set included the set included 2 samples each from lines W1, W2 and W3 with Th W of 3 and 25 mm so as to cover the range of the both factors of interest. Ultrasonic testing data for control samples were passed by the decision rule algorithm described in 2.3.2. The error values were chosen relatively high as ε1=0,08 and ε2=0,05 taking into account the permissible error of Th W and composition of the concrete mixture in the manufacture of specimens by semi-manual way. The areas of intersection of 13 decision rules made it possible to determine, with a certain probability, both factors of interest, Th W and W in most control specimens, but sometimes offering more than one solution, of which only one was true. In successful cases, evaluation of Th W was achieved with an approximately error of 1 mm and 0.1 for the cement-sand ratio on the W scale. Examples of successful evaluation are presented in Fig.8, where the locality of the solutions corresponds to the maximum intensity of the intersection of the decision rules and stands out as the brightest zone. A combination of results obtained at two frequencies, 50 and 100 kHz, can help to refine the definition, as shown for the specimen W2, 3 mm in Fig.8. In some cases, several close solutions can be found around the true value, thus determining the region of its probable location, as for the specimen W2, 25 mm in Fig.8.