Sensitivity-Based Model Updating for Structural Damage Identification Using Total Variation Regularization

TL;DR

This paper introduces a total variation regularization approach for sensitivity-based finite element model updating to improve damage identification accuracy in structures, especially under noisy measurement conditions.

Contribution

It proposes and compares a total variation regularization method with traditional interpolation-based regularization for structural damage detection.

Findings

Total variation regularization outperforms interpolation in damage localization and severity estimation.

The method is validated on a reinforced concrete plate with multiple damage scenarios.

Regularization improves robustness against measurement noise in model updating.

Abstract

Sensitivity-based Finite Element Model Updating (FEMU) is one of the widely accepted techniques used for damage identification in structures. FEMU can be formulated as a numerical optimization problem and solved iteratively making automatic updating of the uncertain model parameters by minimizing the difference between measured and analytical structural properties. However, in the presence of noise in the measurements, the updating results are usually prone to errors. This is mathematically described as instability of the damage identification as an inverse problem. One way to resolve this problem is by using regularization. In this paper we investigate regularization methods based on the minimization of the total variation of the uncertain model parameters and compare this solution with a rather frequently used regularization based on an interpolation technique. For well-localized damages the results show a clear advantage of the proposed solution in terms of the identified location and severity of damage compared with the interpolation based solution. For a practical test of the proposed method we use a reinforced concrete plate. Measurements and analysis were repeated first on an undamaged plate, and then after applying four different degrees of damage.