Application of Statistical Pattern Recognizing Classifiers in Identifying Defects in FRP Composites

Ultrasonic inspection has been widely used to detect defects in composite materials such as internal discontinuity, cracks, delaminations, inclusions, lack of bond etc. However when the bonded specimen has small thickness, the echo from the defect region makes it difficult to identify the echo from the unflawed· region. To overcome these disadvantages, different statistical pattern recognizing classifiers can be implemented efficiently to make an effective classification of the scanned data. Effort is made to remove complications that arise due to unknown nature of interactions of these materials with the acoustic wave In the present work waveforms obtained from ultrasonic C-scans are classified for flawed and unflawed regions of the composite domain by different pattern recognition classifiers. The classifiers include Least Mean Square (LMS), Minimum Distance (MD) and the classification results are used to generate C-scan images. Both time frequency domain features of the ultrasonic signals are used.