A hybrid ANN-metaheuristic framework for vibration-based identification of orthotropic parameters in GFRP laminated plates

Glass-fibre-reinforced polymer (GFRP) composite plates are widely used in engineering structures, yet their numerical modelling is challenging because their material properties can vary due to anisotropy, manufacturing variability, and uncertainty in input parameters. This study develops a fast numerical framework for vibration-based finite element model updating and inverse identification of orthotropic material properties in GFRP composite plates using analytical modal reference data. A baseline finite element model was developed in Abaqus and compared with modal results obtained using Classical laminate plate theory. To reduce the high computational cost of repeated finite element simulations during model updating, an artificial neural network (ANN) surrogate model was trained on parametric FE data and coupled with genetic algorithm (GA), particle swarm optimisation (PSO), and slime mould algorithm (SMA) optimisation methods. The novelty of the study lies in comparing these ANN-assisted optimisation frameworks with direct GA-Abaqus-Python updating under the same modal identification problems. Pearson correlation-based sensitivity analysis showed that the longitudinal elastic modulus E x was the dominant parameter affecting the modal response, with sensitivity values ranging from approximately 0.62 to 0.80. In the representative updated model, ANN-SMA achieved the lowest average frequency-prediction error of 2.289%, compared with 3.821% for the conventional GA-Abaqus-Python updated model. However, statistical analysis over 50 independent runs showed that ANN-PSO provided the best robustness, with the lowest mean frequency error, smallest standard deviation, and lower runtime than ANN-SMA. The ANN-assisted methods reduced the computational time from approximately 5.830 h to 6.617–42.409 s. These results show that ANN-assisted model updating provides an efficient numerical approach for vibration-based material-property identification in composite structures.