Of a variety of NDE techniques, ultrasonic methods are versatile and relatively inexpensive for such tasks. In recent years, with the advent of high speed Analog to Digital (A-D) converter cards, attention has been focused on the automated C-Scan technique through interfacing and control by computers.

In the present investigation, an automated ultrasonic NDE procedure has been implemented for identification of defects in composite materials. The procedure involves scanning of specimens by an automated C-scan system, followed by generation of C-scan images for any selective ultrasonic feature(s) by multi dimensional cluster analysis and subsequently, quantitative evaluation of these images by fractal analysis.

The automated C-scan system, developed in-house, is computer controlled and is capable of scanning in two dimensions. The ultrasonic signal (analog) is digitized over a user selected time gate by a high speed A-D converter card at a sampling rate of 100 MHz. Features, both in time (peak amplitude) and in frequency domain (harmonics, obtained from FFT on the digitized time domain data) are extracted from the digitized data for evaluation purpose.

Cluster analysis is an effective tool for classifying data set pertaining to any single or multi variables. In the present work, hierarchical cluster analysis with agglomerative procedure has been implemented. Here, the category structure of the data set is unknown and the desired number of groups is the only user defined parameter. The data set (i. e., feature values of each scanned location) is classified into the desired number of groups for any selective feature(s) and the respective C-scan images are readily drawn by assigning appropriate grey shades to each groups. Thus image generation by this analysis is automatic and eliminates the necessity of careful selection of multi- thresholds as required in the conventional practice.

In place of the C-Scan images being evaluated by visual inspection, fractal theory has been suggested as a viable supplement (if not an alternative) and a simple algorithm has been developed in this regard. In principle, the fractal analysis on any C-scan image generates the fractal graph from which the fractal dimension (FD) of the image is evaluated. The FD is a measure of the image roughness and thus can be correlated to the image quality.

In the present investigation, the automated ultrasonic C-scan has been performed on impact damaged graphite epoxy composite specimen. C-scan images have been obtained by multi dimensional clustering technique for different feature(s). This clustered data is processed by the fractal algorithm for quantitative evaluation of the images.

Encouraging results have been obtained from this analysis leading to the conclusion that the automated C-scan set up, combined with clustering technique and fractal analysis may constitute a fully automated flaw detection procedure for composite materials.