PSI - Issue 68

Nedeljko Vukojević et al. / Procedia Structural Integrity 68 (2025) 379– 385 N. Vukojević et al. / Structural Integrity Procedia 00 (2025) 000–000

384

6

variables depending on the combination of parameters is given, and insignificant influence of D x t . Fig. 3 shows the 'multiline plot' option, where the curve dependence of the output size on the treated parameters in individual combinations is given. From the position of the lines of the simultaneous influence of the combination of two factors on the objective function in Fig. 3(a), 3(b), and 3(c) it can be concluded that the interaction between the parameters R eH - t and parameters R eH - D exists, and that the interaction between the parameters D - t on the value of total impact energy is not significant.

a

b

c

Fig. 3. Diagram 'Multiline plot' (a) R eH x t multiline plot; (b) R eH x D multiline plot; (c) D x t multiline plot.

3.2. Analysis of the results for the fracture toughness As the second output function of the goal, the dependent function K JIc was analyzed - values of fracture toughness in the observing points of the experimental model. The coefficient of determination is R 2 = 0.1323. The statistical characteristics of the regression model PRESS RMSE, RMSE, R 2 and PRESS R 2 indicate a nonsatisfactory match between the real data of the experiment and the 'predicted' data. Fig. 4 shows the 'surface response' for the combination of influential treated parameters of the R eH x t experiment, where insignificant influence of D x t is shown. Fig. 5 shows the 'multiline plot' option, where the curve dependence of the treated parameters is given.

b

a

Fig. 4. 'Surface response' diagram: (a) R eH x t parameter; (b) D x t parameter relationship

4. ANALYSIS OF RESULTS For the analysis of the relationship between the input parameters ( R eH , D, t ) and the output quantity - the objective functions ( KV and K JIc ) - the same initial form of the regression model of linear form was used, and through the later 'stepwise' analysis the optimization of the regression model and the number of variables was performed. In the analysis of the first output function KV, the values of the residuals, as well as the 'predicted/observed' relationship showed a satisfactory agreement between the mathematical expression and the real values, and the regression model describing the dependence of the output function on the input parameters proved to be reliable. In the analysis of the second output functions K JIc , the values of the residuals, as well as the 'predicted/observed' relationship showed a nonsatisfactory agreement between the mathematical expression and the real values. For the K JIc parameter, due to the inadequacy of the regression model, every analysis is unreal.