ABSTRACT
This paper shows a preliminary work on the proposal of a probabilistic damage detection method for truss structures in a system level. It uses the Selective Expansion Scheme and Bayesian updating methods to update the probabilities of the system damage states of truss structures in a computationally efficient manner based on observed dynamic responses, by considering only selected system damage states. The method is applied to a nine member truss example with two sets of virtually generated observation data. The analysis results identify the most important members successfully in terms of the probability of damage, and the uncertainties in system quantity predictions are well considered. The current limitations of the proposed method are listed for consideration in future work for the further development of the proposed method.
