Generalized Integral Transform-based Structural Damage Identification via Bayesian Parameter Estimation and Model Class Selection
Abstract Purpose This study presents a Bayesian framework for structural damage identification combining the Generalized Integral Transform Technique (GITT) with Bayesian parameter estimation and model class selection. Method The GITT is employed to obtain the time-domain structural response directly from the governing partial differential equations, avoiding the need for spatial discretization. The structural damage field is represented continuously in space through a Gaussian basis expansion, allowing the system integrity to be characterized by a reduced number of uncertain parameters compared to conventional finite element formulations. Bayesian inference is then used to estimate the posterior probability density functions of the uncertain parameters and to perform model class selection based on the computed evidences. The Transitional Markov Chain Monte Carlo method is adopted for efficient sampling and model evidence evaluation. Results Numerical investigations performed on an Euler–Bernoulli beam demonstrate the proposed method’s capability to detect, locate, and quantify damage under different scenarios and noise levels. Conclusion The results show that the proposed framework accurately identifies the true damage configuration, provides meaningful uncertainty quantification, and naturally penalizes overparameterized models, thus ensuring robust and physically consistent damage assessment.
Authors
- Leonardo Tavares Stutz (ORCID: https://orcid.org/0000-0003-3005-765X)
- Diego C. Knupp (ORCID: https://orcid.org/0000-0001-9534-5623)
- R. P. de Andrade (ORCID: https://orcid.org/0000-0001-6688-0344)
Publication Details
- Journal
- Journal of Vibration Engineering & Technologies
- Published
- 2026-11-14
- DOI
- https://doi.org/10.1007/s42417-026-02886-x
- Primary Topic
- Structural Health Monitoring Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00