PSI - Issue 79

A. Della Rocca et al. / Procedia Structural Integrity 79 (2026) 475–484

480

meshing process. To manage computational cost, the dimensions of the cubes chosen for the FEA simulation were kept small. Each voxel dimension was set to 0.1 mm. This size choice resulted in a manageable number of elements: a small cube with 64 voxels per side contained approximately 65,000 elements, while scaling up to 100 voxels per side would increase the element count to 250,000. The simulations were set up to mimic a uniaxial compression test along the y-axis. This was achieved by applying a displacement equivalent to 1% of global strain to the structure, corresponding to a total displacement of 0.064 mm for the modeled cubes. The application of boundary conditions involved a structured sequence in the input file generated for each iteration. First, the bottom and top surfaces of the structure were identified by gathering all corresponding node points. Reference center points along the y-axis were also defined for both the bottom and top. The bottom surface was fully constrained (fixed) by linking its nodes to the bottom center point using a structural coupling with a distributed approach (Fig. 4). The top surface was similarly coupled to its reference center point, and the 1% compressive displacement was then applied to this top center point. Finally, the lateral surfaces of the structure were left unconstrained to simulate free lateral expansion under compression. The effective stiffness (E eff ) of the structure was calculated from the resulting reaction force (F) and displacement ( δ ) data output from the ABAQUS simulation.

Fig. 4. In red the bottom surface set and in yellow, RP1, the selected reference point

2.5. Statistical Analysis for effective stiffness prediction To establish a quantitative link between the structural morphology and the mechanical response of the spinodal structures, a series of regression analyses was performed. The goal was to correlate the effective stiffness, determined by Finite Element Analysis (FEA), with the quantitative structural descriptors extracted from the skeletonization and 2D analysis. The Pearson correlation coefficients were the criteria used to optimize the linear dependencies between individual structural parameters and the computed effective stiffness. Specifically, the coefficient of determination that represents the proportion of variance in a dependent variable that is predictable from an independent variable in a polynomial regression model was employed to predict the effective stiffness values using 14 input parameters (i-indicator reported in Table 1) derived from the structure analysis: � �� � �� � ��̄��� � ��̄� � ��� �� �� � ��̄� � � ��� � �� � ��̄� � � ��� � � (2) These parameters can be conceptually clustered into four predominant morphological categories that drive the prediction: (1) Surface area to volume ratio (SA/V); (2) Total skeleton length; (3) Counts of different node types: end,