On the bayesian identification of the elastic parameters of laminated structures

Numerical-experimental techniques based on vibration data allow to estimate the properties of complex structural components, such as laminated structures, whose coupled behavior is difficult to characterize through traditional tests. The estimates are obtained by minimizing in some sense the difference between the measured structural response and the behavior predicted by an assumed model. Within this context, some important choices must be made concerning both the structural model and the objective function to be minimized (estimator). In this work, the dynamic identification of the elastic constants of thick laminated composite plates from natural frequency data is considered within a Bayesian framework. In the context of fibre reinforced composites, initial estimates for the elastic constants can be obtained by knowing the properties of the fibres and matrix. This information should be used to provide a good starting point to simplify the search, but it should not influence too strongly the final estimates. With this in mind, two different estimators which differ in the way in which they account for the a priori information are compared. The structure is modeled by finite elements based on Reddy's third order theory. Numerically generated natural frequencies and real experimental data are used as input in different case studies which highlight the similarities and differences between the estimators. A unified interpretation of the two estimators is given and a modified strategy is proposed in order to improve reliability and convergence of the estimation process.