Accounting for Modeling Errors and Inherent Structural Variability through a Hierarchical Bayesian Model Updating...

Area of Science:

• Structural Engineering
• Computational Mechanics
• Data Science

Background:

• Mechanics-based dynamic models are crucial for structural design and performance assessment.
• Integrating measured data enhances model accuracy but requires robust uncertainty quantification.
• Existing methods often struggle to comprehensively address various uncertainty sources.

Purpose of the Study:

• To provide an overview of hierarchical Bayesian model updating.
• To demonstrate its capability in probabilistically integrating models with measured data.
• To highlight its effectiveness in accounting for uncertainties and modeling errors.

Main Methods:

• Overview of hierarchical Bayesian model updating framework.
• Probabilistic integration of mechanics-based models with measured data.
• Explicitly accounting for sources of variability (temperature, excitation amplitude) and modeling errors.

Main Results:

• Applications to a footbridge, a 10-story RC building, and a damaged 2-story RC building.
• Demonstrated capability to account for temperature effects and excitation amplitude.
• Highlighted the impact of considering bias for prediction error.

Conclusions:

• Hierarchical Bayesian framework is essential for accurate structural identification.
• The approach offers improved, realistic predictions compared to classical methods.
• Proven advantages and performance over deterministic and standard Bayesian model updating.