Application of an intelligent hybrid global optimization (IHGO) algorithm for enhanced seismic analysis in masonry-infilled RC frames

Abstract

This study introduces the Intelligent Hybrid Global Optimization (IHGO) algorithm to improve the predictive accuracy of neural network models for estimating the fundamental period of vibration in masonry-infilled reinforced concrete (RC) frame structures. Using a dataset of 4,026 entries, which includes critical structural parameters such as the number of storeys (ranging from 2 to 15), span length (3–8 m), opening ratio (0–50%), and masonry wall stiffness (up to 10⁵ kN/m), the IHGO algorithm optimizes neural network hyperparameters. The IHGO-optimized neural network outperforms baseline models, achieving an R² value of 0.92, a Mean Absolute Error (MAE) of 0.012 s, and a Root Mean Square Error (RMSE) of 0.017 s, compared to 0.85 R², 0.018 MAE, and 0.026 RMSE for the standard neural network. The optimization balances exploration and exploitation, enhancing precision and revealing complex nonlinear relationships between structural features and seismic behavior. The study demonstrates the critical role of accurate period estimation in seismic design, supporting better assessments of structural vulnerabilities and compliance with safety standards. This work highlights the efficacy of hybrid optimization in structural engineering and suggests future research on adaptive tuning and broader seismic applications.