Uncertainty Analysis of Damage Identification Results Based on Finite Element Model Updating

This paper aims to investigate uncertainties in damage identification results due to errors in modal parameter estimation results. Structural damage is simulated as regional stiffness loss at the column(s) and beam ends of a numerical frame type structure. In the damage identification stage, the first 4 modal parameters are used. Two different levels of noise are added to them to simulate uncertainty in modal parameter estimations. Noise levels are controlled by the coefficient of variation (C.O.V.). In order to quantify the uncertainty of the identified damage due to the variability of modal parameters, a full factorial analysis of variance (ANOVA), resulting in 16 combinations of input factors, is used. For each combination of input factors, 20 noise realizations are generated using Gaussian normal distributions with standard deviations scaled to the level of modal parameters. The results are presented in two formats: (1) Spread of the identified damage factors for all 320 identification runs with their statistical measures and (2) R² values of the mean and standard deviation of the identified damage factors due to the variability of each input factor. The results of this investigation demonstrate that specific modal parameters have only influence on specific damage factors.