PSI - Issue 75

Marco Bonato et al. / Procedia Structural Integrity 75 (2025) 677–690 / Structural Integrity Procedia (2025)

682

6

2.3. Latin Hypercubic Sampling for FEA runs The simulation test plan based on the LHS method has been generated in the R language [Carnell (2024)]. The first R coding package “Latin Hypercube Samples” version 1.2.0 runs on R 4.3.3. The main parameters considered were: - the Damping (see Figure 3, left side) - The fatigue exponent of the SN curve (Basquin’s b) (Figure 3, right side) - The ratio between the SN curve interact to UTS (Figure 3, right side)

- Figure 3: Graphical representation of the FEA fatigue parameters. Damping (left) and SN curve with slope, intercept and UTS (right).

Initially, the variability of these parameters was kept intentionally broad. We wanted to evaluate the predicted scatter of the vibration fatigue time-to-failure by considering the typical values of the above mentioned parameters, encountered on a large range of similar materials. For such calculations, the LHS inputs were the average of the recorded parameters of several aluminum alloys

currently part of the company products material: ● Damping : uniform distribution, [0.9 ; 1.2] ● Basquin's b: normal distribution, mean = 8.0 ; standard deviation 0.5 ● Intercept/UTS: uniform distribution, [4.7 ; 5.1 ]

Note that the fatigue exponent (the Basquin’b) highly depends on the type of loads to which the prototypes undergo (e.g. a SN curve defined only by tension-compression tests, as opposed to vibrational dynamic loads). In particular, for the type of material investigated, the authors have noticed significant differences between the b values obtained from the fatigue machine during material characterization (in this case b = 13) and values originated from vibration tests (in this case, for aluminum alloy, b is always close to 8). The resulting test plan included 20 simulations runs. The graphical allocations of the FEA parameters are shown in Figure 4.