Abstract
Composite materials have been widely used to fabricate highly reliable composite structures. Since the material constants of the composite structures are important parameters for the reliability assessment of the structures, it is thus desired to have an efficient and effective technique to determine the actual material constants of the constituent materials. In this paper, a novel sensitivity-based multi-level optimization method, which is composed of several level-wise optimization stages, is presented to identify the actual material constants of structures using measured natural frequencies. In the proposed method, the natural frequency sensitivity information for a structure is used to establish the objective functions and conduct the selection of appropriate design variables at different optimization levels. In each level-wise optimization, the number of design variables is properly reduced to simplify the optimization so that the solution can be attained easily and efficiently. The solutions of the level-wise optimization problems produce the expected values and coefficients of variation for the estimates of the material constants. An acceptance criterion established on the basis of the coefficient of variation has been used to assist the identification of the actual material constants. The accuracy verification and applications of the proposed method have been demonstrated by means of several numerical and experimental examples on the identification of material constants for composite plates with different lamination arrangements.