PSI - Issue 78

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

1234

Fig. 3. Mode shapes obtained by modal testing of panel 3

3.2. Modelling

All the three numerical approaches modelled within DIANAFEA environment produced very similar sti ff ness pre dictions, with di ff erences remaining well within experimental variability. This suggests that, for longitudinal sti ff ness modelling, any of these strategies can be used with acceptable accuracy. Figure 3 (b) gives an example of FE results for panel 3. The results of the di ff erent panels and property assignments are shown in Figure 4. Additional tests (not presented here) included assigning the statistical mode as a constant value for each layer, or uniformly applying a single random modulus to all lamellae within a layer. While the former produced results nearly identical to the average-based model, the latter led to larger variability due to the imposed homogeneity within layers. A more extensive Monte Carlo simulation could help quantify this e ff ect, but was beyond the scope of this study. Overall, the modelling results appear robust with respect to the choice of modulus-assignment method, supporting the use of simplified average properties in practical applications. The comparison between the numerical, experimental non-destructive (NDT), and destructive testing (DT) results from 4-point bending tests highlights important aspects of the mechanical behaviour of hybrid CLT panels. The two sets of sti ff ness results, those obtained from the numerical model and those from NDT, exhibit the same overall trend, with one set appearing as a shifted version of the other (see Figure 4). This consistent pattern suggests that the variation is systematic. Moreover, modifying the orthotropic property ratios used in the model, as shown in Table 2, can lead to noticeable di ff erences in the predicted sti ff ness values, particularly in the transverse and shear directions, where the sensitivity to these ratios is higher. In the longitudinal direction ( E X ), the finite element model reaches very good agreement with the destructive test results, showing a relative error of only 1.58%, while the NDT-based estimation deviates by 2.5%. The di ff erence between the model predictions and NDT results is 4.1%, confirming that the numerical framework accurately captures the dominant sti ff ness direction. This is attributed to the fact that the outer Ash layers, being both the thickest and the sti ff est, govern the global longitudinal sti ff ness response. 3.3. Comparison