Bayesian Inference to Damping Identification of Fiber-Reinforced Composites from Experimental Modal Data
* Presenting author
To study the dynamic behavior of composite structures, the identification of elastic and damping parameters is essential. A significant amount of variability arises to the modal response of such structures due to the uncertainty in the fiber orientation, geometry parameters and inter-laminar stress during manufacturing process. Furthermore, there are various sources of modeling errors, such as inaccurate boundary conditions, material constitutive behavior, simplification and linearization of the numerical model. In this paper, the Bayesian inference technique is used to treat the problem of uncertainty of the responses by updating the damping parameters of the finite element (FE) model based on experimental test data. Bayesian updating aims to minimize the probability of the error associated with the behavior of the FE model and the actual behavior from experimental data. The sampling based Markov Chain Monte Carlo method is used for the numerical Bayesian inference. This methodology identifies the probability distribution of the best possible distribution for the damping parameters to predict uncertainty of the actual structural responses using the computational FE model. The application of the method will be examined to update the FE model performing the numerical modal analysis of a twelve layers fiber-reinforced composite plate from experimental modal data.