# Bayesian System Identification based on Hierarchical Sparse Bayesian Learning and Gibbs Sampling with Application to Structural Damage Assessment

**Authors:** Yong Huang, James L. Beck, Hui Li  
**arXiv:** [1701.03550](https://arxiv.org/abs/1701.03550)  
**Date:** 2017-01-13  
**Summary page:** https://www.summarizepaper.com/en/paper/1701.03550v1/

## Key points

- Bayesian system identification based on noisy and incomplete modal data
- Focus on detecting and assessing spatially-sparse stiffness reductions caused by structural damage
- Proposal of two Gibbs sampling algorithms building upon a hierarchical sparse Bayesian learning model
- Algorithms sample from conditional posterior probability density functions (PDFs) for structural stiffness parameters and system modal parameters
- Computational efficiency as they do not require solving a nonlinear eigenvalue problem
- Low effective dimension for Gibbs sampling, involving iterative sampling from three conditional posterior PDFs corresponding to three parameter groups
- One algorithm samples the equation-error precision parameter separately from another conditional posterior PDF
- Application of algorithms to IASE-ASCE Phase II simulated and experimental benchmark studies using incomplete modal data before and after possible damage
- Contribution to strengthening vibration-based structural system identification capabilities in Bayesian inference of structural stiffness
- Broader applicability for inverse problems in science and technology where system matrices need to be inferred from noisy partial information about eigenquantities

## Summary

This paper focuses on Bayesian system identification based on noisy and incomplete modal data, specifically addressing the problem of detecting and assessing spatially-sparse stiffness reductions induced by structural damage. The authors propose two Gibbs sampling algorithms, building upon a hierarchical sparse Bayesian learning model from their previous work. These algorithms sample from conditional posterior probability density functions (PDFs) for the structural stiffness parameters and system modal parameters. The key advantage of these algorithms is that they do not require solving a nonlinear eigenvalue problem of a structural model, making them computationally efficient. The effective dimension for Gibbs sampling is low as it involves iterative sampling from only three conditional posterior PDFs corresponding to three parameter groups. Additionally, one of the algorithms samples the equation-error precision parameter separately from another conditional posterior PDF. To demonstrate the effectiveness and robustness of their proposed algorithms, the authors apply them to the IASE-ASCE Phase II simulated and experimental benchmark studies. They use incomplete modal data identified before and after possible damage to detect and assess spatially-sparse stiffness reductions caused by any damage. The authors emphasize that their research contributes to strengthening vibration-based structural system identification capabilities by addressing challenges in Bayesian inference of structural stiffness. Furthermore, they highlight that their methods have broader applicability for inverse problems in science and technology where system matrices need to be inferred from noisy partial information about their eigenquantities. Overall, this paper presents novel Gibbs sampling algorithms for Bayesian system identification using incomplete modal data, with specific application to detecting and assessing spatially-sparse stiffness reductions induced by structural damage.