Probabilistic Updating of Structural Models for Damage Assessment Using Approximate Bayesian Computation.

Area of Science:

• Structural Engineering
• Computational Mechanics
• Bayesian Inference

Background:

• Accurate structural damage assessment is crucial for safety and maintenance.
• Traditional model updating methods often rely on explicit likelihood functions, which can be computationally intensive or intractable.
• Modal data (frequencies and mode shapes) are sensitive indicators of structural changes.

Purpose of the Study:

• To propose a novel probabilistic approach for structural model updating using approximate Bayesian computation with subset simulation (ABC-SubSim).
• To enhance the accuracy and efficiency of damage assessment in structures by leveraging modal data.
• To address limitations of likelihood-based methods in Bayesian model updating.

Main Methods:

• Development and application of a likelihood-free Bayesian approach (ABC-SubSim) for structural model updating.
• Introduction of new stopping criteria for determining the tolerance level in ABC-SubSim.
• Employment of a hybrid optimization scheme for refining model parameter estimation.
• Adoption of an iterative approach to optimize weighting factors for modal frequency and shape residuals.

Main Results:

• The proposed ABC-SubSim method effectively updates structural models using modal data.
• Novel stopping criteria improve the robustness of the tolerance selection process.
• Hybrid optimization and iterative weighting factor determination lead to more accurate parameter estimations.
• Demonstrated effectiveness across three diverse structural examples.

Conclusions:

• ABC-SubSim offers a powerful and flexible framework for structural model updating and damage assessment.
• The introduced enhancements significantly improve the performance and applicability of the ABC-SubSim approach.
• This method provides a robust alternative for structural health monitoring when likelihood functions are challenging to define.