System identification of adaptive composites

Abstract

Modern composites are non-homogeneous materials having high strength fibers embedded in a polymeric or metal matrix, and having directional properties. Shape Memory Alloys (SMAs) such as Nitinol can be embedded in composites for active control of structures. Since the micromechanics of composites in general and adaptive composites, in particular, are poorly understood, accurate values of material elastic properties and in-plane loads generated are seldom available. It is possible to determine these parameters if natural frequencies of the structure are available. The techniques used to determine these parameters from the modal response of a structure are generically called ‘System Identification’.

This thesis reviews the various System Identification techniques applicable to vibrating structures. Three techniques are then adapted to an orthotropic laminate and implemented in FORTRAN. These techniques, being iterative in nature, require an initial estimate of the parameters to be identified as input. Because all the techniques discussed here are iterative, they are sensitive to the values of initial estimates. The robustness of these techniques in face of a scatter in the input data is tested using a randomized statistical analysis. The techniques are compared based on certain attributes and recommendations are made.