A systematic sensitivity analysis workflow for hysteretic systems: applications to biaxial Bouc–Wen models
Year: 2026
Type: article
Abstract: This study provides a practical, systematic sensitivity analysis (SA) workflow for evaluating complex hysteretic models. The workflow sequentially combines multiple SA approaches, including Morris screening, Sobol’ variance-based indices, and distribution-based methods, such as the Kullback–Leibler divergence (KLD), symmetric Kullback–Leibler divergence (SKLD), and the Jensen-Shannon divergence (JSD), into a structured four-step process. These steps are used to define valid parameter bounds, screen out unimportant parameters, apply the variance-based Sobol’ sensitivity, and leverage the distribution-based methods for additional insights. The proposed workflow is applied to three biaxial hysteretic models from the Bouc–Wen family, namely the conventional, degrading, and generalized Bouc–Wen models. The SA for each model is performed with respect to full-scale experimental data of passive steel yielding dampers that were subjected synthetic and historical excitations. The results show that while shape parameters constitute the dominant factor for conventional models, the strength degradation parameter plays an important role in highly nonlinear scenarios for the degrading model. The generalized Bouc–Wen is analyzed from a phase-aware perspective that maps each SA parameter to a particular hysteretic phase, better illuminating the behavior dependence of the SA results and improving interpretability. The results also demonstrate how the JSD metric, in contrast to the KLD and SKLD measures, consistently captures not only variance but also distortion in the output distribution, providing further insight into model parameter sensitivity.
Source: Nonlinear Dynamics
Authors: Reza Farzad, Patrick Brewick
Institution: University of Notre Dame
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,790
Funder: University of Notre Dame
OpenAlex ID: W7160530828