Abstract This study proposes an unsupervised fuzzy neural network-based two-phase approach for updating the stiffness of shear buildings. The first phase constitutes a preliminary stiffness estimation process that employs the stiffness updating index (SUI) strategy to facilitate a rapid, albeit approximate, update of the stiffness parameters. In the SUI strategy, an inter-story deflection-based SUI is employed to determine the appropriate direction for updating the floor stiffness (whether to increase or decrease the predicted stiffness values). The inter-story deflection is computed from the flexibility matrix, which is approximated using the first-mode shape of the structure. Due to the inherent approximation in the inter-story deflection calculation, the accuracy of the SUI tends to diminish as the discrepancy between the expected and predicted stiffness values decreases. The second phase constitutes a refinement process that employs the unsupervised fuzzy network (UFN) strategy to determine the optimal stiffness by refining the preliminary stiffness obtained from the SUI strategy. The UFN strategy employs the UFN reasoning model, a pattern recognition technique that integrates an unsupervised neural network with a fuzzy computing mechanism, to update the stiffness of shear buildings effectively. Numerical and experimental examples demonstrate the feasibility of the proposed approach for updating the stiffness of shear buildings. The results indicate that the proposed approach can accurately and efficiently update the stiffness and detect damage (including single and multiple damage) to shear buildings with complete and incomplete measurements.
Kao et al. (Mon,) studied this question.