PSI - Issue 78

Pasquale Guarino et al. / Procedia Structural Integrity 78 (2026) 1561–1568

1565

generate the file containing the mode shapes. The positions of the sensors and the measurement directions are shown inFig. 3. After the files for emulating information derived from experimental data have been generated, the following pa rameters were chosen for the Particle Swarm Optimization algorithm:

• number of particles = 15 • max iter = 30 • v = 0.025 • w = 0.9

• c1 = 2 • c2 = 2 • α = 1 • β = 10 • η = 0

Fig. 3. Mesh representation in Revit with sensors’ location.

where ’max iter’ is the maximum number of iterations, ’v’ is the rate of change of velocity, ’w’ is the inertia weight for PSO, ’c1’ and ’c2’ are the cognitive and social components for PSO respectively, ’ α ’, ’ β ’, and ’ η ’ are weighting factors.

4. Results

The analysis reports the Young’s modulus and mass density multipliers of the fitting parameters as a function of the number of iterations performed, showing that the value stabilized after 22 iterations (Fig. 4a). Furthermore, the output includes the progression of the cost function as it changes with the number of iterations (Fig. 4b). Regarding the model parameters, the plug-in provides the Modal Assurance Criterion (MAC) values between the experimental and the numerical mode shapes (Fig. 5), as well as the comparison in terms of resonant frequencies for the calibrated and the uncalibrated models (Table 2). As it can be seen in Table 2, the calibrated frequency values show a limited error, reaching a maximum of 0.4%, which demonstrates the e ff ectiveness of the model update. This is further confirmed by the calculated values of the