An experimental and numerical study on application of modified Bouc-Wen model for viscoelastic dampers
Seungho Chun,Mohammad Mahdi Javidan,Yang Xiang,Jinkoo Kim
Abstract
This study presents and studies the application of normalized Modified Bouc-Wen model to capture the nonlinear hysteretic behavior of viscoelastic dampers under cyclic loading. Six different tests were conducted on viscoelastic dampers with different frequencies and amplitudes to evaluate their energy dissipation performance. The model considers strength and stiffness degradation effects using a normalized form to capture nonlinear hysteretic behavior based on applied loading and energy dissipation. Since identifying Modified Bouc-Wen parameters from the experimental results is a challenging task due to the complexity and interdependence of the parameters in the model formulation, the Particle Swarm Optimization (PSO) algorithm was applied. The first stage involved exploring a wide range of parameter values to identify reasonable value ranges, and the second stage applied refined bounds to improve accuracy. The calibrated model was validated by comparing its results with experimental hysteresis curves, confirming good agreement in both stress-strain response and energy dissipation. Furthermore, various machine learning regression models were trained using measurable input parameters. The output variables were the remaining Modified Bouc-Wen parameters derived from PSO-based optimization. Among the tested machine learning models, Gradient Boosting achieved the best performance, effectively estimating the Bouc-Wen parameters and reliably predicting the overall hysteretic behavior. SHAP analysis was conducted to interpret the model, indicating that stiffness and amplitude were the most influential features in estimating the Modified Bouc–
Wen parameters.
