PSI - Issue 23

Hynek Lauschmann et al. / Procedia Structural Integrity 23 (2019) 107–112 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

111

5

Maximal curvature, quantile P = 0.1 ሾμmሿ

Horizontal curvature, stand. dev. along crack front direction of the stand. dev. along crack growth direction ሾ μm ሿ

Height, mean value along crack front direction of quantile P = 0.2 along crack growth direction ሾ μm ሿ

Fig. 4: Features used for solutions shown in Fig. 5. Courses of training features f i ( x ) (markers) and regressions ( ) i f x (curves) with x -axis scaled in crack growth rate . v Markers denote sections of decomposition: o = 1, ...,  = 8.

a

b Fig. 5: Comparison of crack growth rates (a) and growth history (b) estimated on the basis of 3D crack morphology with real courses known from the experiment. Features used are presented in Fig. 4. Results. Eq. (1) has been used for assessment of crack growth rates on the basis of training data. Quality of results was judged according to the value of correlation coefficient between known and estimated crack growth rates. For all three specimens, several tens of features were found to give good results - correlation of crack growth rates greater then 0.9 for both training and testing data. Examples of some of the best results are documented in Fig. 5 by graphs comparing data computed on the basis of fracture surface with data known from the experiment. In the first graph, crack growth rates are compared, the second graph compares crack growth histories (dependences of crack length on the number of loading cycles).