Real-time Model Uncertainty Quantification and Hyper-robust Structural Identification

摘要

In structural health monitoring, there is always uncertainty in which model class to choose to represent a structural system.This paper aims to provide a real-time algorithm for quantification of modeling uncertainty using measured response data.The key is to recursively update the probability of each model class within a proposed set of models for the structural system, which can be done by embedding the Bayesian model class selection component into the extended Kalman filter.A hyper-robust structural identification based on model class averaging is proposed.A simulated example is presented to illustrate the effectiveness and robustness of the proposed method.