Challenges in Model Identification for Seismic Performance Evaluation of Building Accounting for Uncertainties

Abstract

In evaluating the performance of a building, it is important to properly treat various types of uncertainties involved in the evaluation process. In the field of performance-based earthquake engineering, several attempts have been made to assess the uncertainties involved at each stage. For example, the ground motion intensity measure of input ground motion is treated as a random variable in PSHA (Probabilistic Seismic Hazard Analysis) and occurrence of damage is also treated probabilistically as component fragility curves. A response analysis model, however, does not explicitly consider uncertainties. To systematically evaluate the performance of a building by capturing the probability distribution of the building performance, i.e., best estimate plus uncertainty, the effects of uncertainties in the response analysis model also need to be considered.

When evaluating the seismic performance of an existing building, a response analysis model should be developed by considering both information obtained during design phase and that from the actual building. Though a fully Bayesian approach for model identification is considered to be effective, the benefits of applying Bayesian approach are not fully understood. This paper discusses the challenges to take into consideration the uncertainties in model identification of existing buildings using a fully Bayesian approach considering both information obtained during design phase and observed building response to moderate earthquakes during its lifetime.