doi:10.3850/9789628014194_0050

ABSTRACT

This paper explores the possibility of using a Bayesian probabilistic methodology for the crack detection of thin plate structures utilizing measured dynamic responses at only a few points on the plate. Owing to the problems of measurement noise, modelling error and “incomplete” measurement (i.e., the lack of measured DOFs when compared to the DOFs of the finite element model and the lack of higher mode information from the measured time domain signal), the results of crack detection as an inverse problem is uncertain in nature. In order to explicitly handle this uncertainty problem, the proposed crack detection methodology follows the Bayesian statistical system identification framework. Instead of pinpointing the crack parameters (i.e., the crack location and the corresponding length and depth), the posterior probability density function of the crack parameters is calculated for quantifying the confidence level of the identified results, which are extremely important for engineers to make judgments on remedial works. One of the outstanding advantages of the proposed methodology is that only dynamic measurement at a few points on the plate is required. As a result, the method is applicable even when the crack is obstructed and measurement at or close to the crack cannot be obtained. This paper reports the theoretical development in the modelling of a cracked plate and the crack detection method together with the numerical verification of the proposed methodology. The case study results are very encouraging showing that the proposed methodology is feasible.

Keywords: Bayesian approach, Crack detection, Plate structures, System identification.

© 2009 Research Publishing Services