Fuzzy-enhanced Convolutional Neural Network for Predicting Structural Responses to Seismic Excitations

This study tackles the challenge of accurately predicting structural behavior under seismic excitation—a critical need in designing and assessing resilient buildings and infrastructure. Traditional approaches, including numerical modeling and differential equation solvers, often encounter substantial computational challenges, particularly when addressing nonlinear hysteretic behavior or large-scale structures. To address these issues, we present a novel Fuzzy-Based Convolutional Neural Network (FuzzyCNN) model, which combines fuzzy logic with convolutional neural networks to effectively capture and manage the uncertainties and complexities of soil-structure interaction under seismic loading. The model's efficacy is validated through both numerical simulations and experimental data from a mid-rise concrete building subjected to seismic events. Comparative analysis reveals that the FuzzyCNN outperforms a conventional Physics-Informed CNN (PhyCNN) model, offering superior accuracy and robustness in predicting seismic responses. Key findings indicate that the FuzzyCNN not only enhances predictive accuracy but also better handles uncertainties than the PhyCNN. This advancement has significant implications for engineering design, seismic risk assessment, and the development of resilient infrastructure, promising improved efficiency and accuracy in dynamic response prediction.

کلیدواژه ها:

Fuzzy logic-based network ، convolutional neural network ، Seismic response prediction ، Seismic Excitation