A PCA-Based Approach for Structural Dynamics Model Updating with Interval Uncertainty

Abstract

Uncertainty widely exists in engineering structures, making it necessary to update the finite element (FE) modeling with uncertainty. An updating approach for the FE model of structural dynamics with interval uncertain parameters is proposed in this work. Firstly, the correlations between the updating parameters and the output qualities of interest are uncorrelated by the principal component analysis (PCA), and the massive samples of major principal components are obtained using the Monte Carlo simulation. Then, the massive samples of updating parameters and output qualities are obtained by reversing transformation in the PCA, and the new samples of updating parameters are re-chosen by combining the massive samples with the experimental intervals of the qualities of interest. Next, the 95% CI of the updating parameters is estimated by the nonparametric kernel density estimation approach, which are regarded as the intervals of updating parameters. Lastly, the proposed approach is validated in simple rectangular plates and the transport mirror system. The updating results evidently demonstrate the feasibility and reliability of the proposed approach.