PSI - Issue 78

Vittoria Borghese et al. / Procedia Structural Integrity 78 (2026) 1229–1236

1232

( 5802A , Dytran instrumens , USA). The accelerometers and the hammer were connected to a data acquisition device ( PCI-4472 , National Instruments , USA) to enable synchronization. A software program ( Dewesoft 7.1.3 , Dewesoft , Slovenia) was used to visualize the mode shapes in real time during testing. Equation 1 requires free-free boundary conditions, realized by the orange bands. The orange bands were used to not damage the CLT panels to ensure use in construction after testing. A comparison study indicated no significant influence of these boundary conditions, compared to (traditional) metal fasteners mounted into the top narrow side to establish free-free boundary conditions. A customizable, parametric algorithm was developed to generate 3D FEA models of the CLT panels, defining their geometry and material properties and analysis settings, and perform the modal analyses. The outputs of the analysis process include the displacements, eigenfrequencies, and associated mode shapes. The outcome of the models is compared to the above-described experimental tests to assess how three di ff erent modulus-assignment methods a ff ect the accuracy of the natural frequency and mode shape predictions. The considered modulus-assignment methods were assigning: 2.4. Parametric Modelling and Analysis DIANAFEA ( DIANA10.7 , DIANA , The Netherlands) was used to generate the FEA Models and perform the modal analysis using three distinct Python scripts ( Python 3.11.4 , Python Software Foundation , USA) as input. These scripts handle the input data generation, model assembly, and output processing. This implementation o ff ers a fully parametric CLT panel model with its coordinate system the same as depicted in Figure 1 (a). The overall panel dimensions are L , W , and H = m k = 1 t k . Each lamella in layer k at in-plane position i is denoted by Λ k , i ,where k = 1 , 2 , . . . , m is the layer index (bottom to top), i = 1 , 2 , . . . , n k is the lamella index within layer k , t k is the uniform thickness of all lamellae in layer k , l k , i and w k , i are the length and width of Λ k , i , respectively. m equals five in this study. A global Cartesian coordinate system ( X , Y , Z ) is defined for the CLT panel (Figure 1) and a local coordinate system ( x k , i , y k , i , z ) is used for each lamella Λ k , i , with z ≡ Z . The in-plane axes depend on the layer parity with x k , i ∥ X and y k , i ∥ Y , if k is odd, and x k , i ∥ Y , and y k , i ∥ X , if k is even. The densities for each specie is given in Table 1. With the framework structured as described, multiple geometries and panel configurations were generated based on assumptions tailored to the research objectives and scope considered here. The modular architecture of the code enables for straightforward future extension and implementation. The mentioned modules include, firstly, 3D struc tural modelling, with the lamellae being generated as 3D elements, here coupled together us a rigid tie connection between all lamella in, and between, the layers, reflecting the glue bond. The 3D elements accurately capture the through-thickness bending, shear and torsional sti ff ness distributions, and the resulting inter-lamellar coupling govern the eigenfrequencies and mode shapes of the panel. 2D elements or plate models inherently neglect out-of-plane ef fects and local stress gradients, leading to less reliable vibrational predictions (Reddy, 2004). Secondly, 3D elements, eight-node hexahedral solid elements ( SOLID/STRSOL ) are used, with mesh-seed control along the CLT and lamella axes, defined via GEOMET/LOCAXS , to represent the full geometry of each lamella and orthotropic material orienta tion without approximation. In literature, other typology of modelling with 1D elements are reported, for example by Huang et al. (2020). Finally, elastic material behaviour, linear-elastic orthotropic properties (longitudinal, radial and tangential moduli of elasticity, shear moduli and Poisson’s ratios) are assigned to each lamella. Three statistical and probabilistic approaches were investigated for assigning the elastic moduli to the respective lamellae or layers within the model. The framework is schematically shown in Figure (2). The input data is represented by variables and vectors. The variables are the CLT panel dimensions, L , W = w b , 1 , w b , 2 , . . . , w b , m , and H = t 1 , t 2 , . . . , t m . The actual lamella widths were computed by packing as many full segments of width w b , k as fit within the panel dimensions (either W or L ) and adding a final remainder width as needed. Material density for each layer was stored in the array ρ k , i , and the measured longitudinal modulus values for each lamella of a specific panel (e.g., the averages shown in Table 1) were stored in a matrix denoted as E raw l , organized by layer and lamella position. This matrix was used exclusively in analysis typology 1 , where the actual lamella positions within the CLT panel were preserved. Furthermore, the processed elastic modulus array E l , k , i was generated according 1. the individual lamella values as in the real hybrid CLT (real); 2. the average value of the lamellae within an individual layer (average); 3. a random modulus from all layer datasets to each lamella (random).