Enhanced Accuracy of a Fuzzy Logic Model for Rapid Seismic Damage Prediction of RC Buildings

Rapid Visual Screening (RVS) is a method of assessing a building's vulnerability to an earthquake. Although RVS accurately classifies the damage states of the studied buildings, the parameter values assigned by this method have intrinsic uncertainties that can highly reduce the accuracy of its predictions. Accordingly, using a fuzzy logic model is known to mitigate these uncertainties and offer a more reliable level of damage prediction. A recently-proposed Dual-Fitness Particle Swarm Optimization (DFPSO) algorithm is applied to fine-tune the hyperparameters of this model, named the FLM-DFPSO model. Furthermore, the 1994 Northridge earthquake data is used to benchmark the performance of the proposed model against a well-known model in the literature. The results suggest that the proposed model improves the performance criteria by 7.46% and 27% in the training and validation stages, respectively. The model also boosts the prediction accuracy by a rate of 53% in the validation step, confirming the FLM-DFPSO as a highly reliable model to learn and also generalize the relationship between its inputs and output to other cases when assessing seismic damage.