PSI - Issue 41

America Califano et al. / Procedia Structural Integrity 41 (2022) 145–157 Author name / Structural Integrity Procedia 00 (2019) 000–000 7 The SLs represent a threshold between safe and unsafe condition areas. The SLs are dependent from the geometrical parameters of interest, namely � and � � � � . In Figure 6, the slopes of the SLs shown in Figure 5 and their quadratic fit is reported with respect to the � � � � parameter, by keeping the thickness of the gesso layer, � , constant ( � = 1.0 mm in Fig. 6a, � = 1.5 mm in Fig. 6b and � = 2.0 mm in Fig. 6c). It can be easily appreciated that the slope of each SL is non-linearly dependent from the thickness ratio; therefore, this preliminary consideration has been taken into account in the binary classification problem assessment. In the following, if not explicitly described, the evaluations are related only to the points (and data) belonging to the plane z = 6 within the BV for the twelve configurations (filled scatters in Figure 5) . 3.2. Machine learning implementation According to the aim of the work, the problem under concern is mainly a binary classification problem. As a matter of fact, the final goal is to classify between two conditions: safe (no further climate-induced damage - 0) and unsafe (further climate-induced damage - 1). 151

Figure 4 - Border Volume (BV) selected within each configuration. The blue and red scatters follow the logic of the SED criterion previously explained.