Comparative study of processing methods for analyzing the structure of carbon epoxy composite materials

Abstract

The nondestructive evaluation of composite materials is needed to warrant their mechanical viability. Several classical ultrasonic signal processing methods may be used to perform this task, such as resonance spectroscopy or feature extraction from time echograms. An alternate method may be based on neural net classification of the structural parameters (ply thickness, interface quality). The aim of this work is to implement the above techniques and to apply them to theoretical models and industrial samples in order to get meaningful comparisons and to assess the validity of the neural net method. The geometry and structure of the composite materials is first recalled (unidirectional or crossed plies) and simplified models are given. For each studied method, the physical principles and related mathematics are briefly presented. Theoretical results are derived using the models. Then the electronic setup is described and experimental results obtained using several industrial samples are shown and compared to the previous theoretical results. Finally, the coherence of the results obtained from the three methods is checked and a reasonable fit is evidenced. The neural net based method appears very flexible and less mathematically involved than the other two methods. It gives a precision sufficient for most purposes, and so constitutes a good candidate for composite material (and more generally multilayer) structure evaluation.