APPLICATION OF MULTI DIMENSIONAL CLUSTER ANALYSIS IN IDENTIFICATION OF DEFECTS IN FIBRE COMPOSITES

Abstract Composite materials provide a serious challenge to the investigators as these materials may develop several types of damages during service such as delamination, voids, porosity, matrix cracking, fibre break etc., either individually or in combinations. Evaluation of exact type of damage quantitatively in composite materials is far more complicated than in the conventional homogenous materials which contain voids and cracks as primary defects. In the present investigation, a non-destructive testing is performed on impacted and artificial inclusion implanted laminates by an automated ultrasonic C-scanning system. Amplitude and phase of frequency components of the received ultrasonic signal have been used for showing images instead of conventionally used peak amplitude and time of flight. Cluster analysis has been employed to group the data units, either individuaJly or in combinations of them in a systematic way. The C-scan images show that the analysis of certain harmonics individually or taken together bring out valuable information about the damage in graphite-epoxy and in glass-epoxy laminates. This investigation strongly shows harmonics as valuable parameters for identification of damages and multi dimensional clustering technique as a powerful tool for effective classification.