Artificial Neural Network Application for Damages Classification in Fibreglass Pre-impregnated Laminated Composites (FGLC) from Ultrasonic Signal
Ultrasonic testing (UT) is a major Non-Destructive Test (NDT) technique used in composite laminates inspection. The traveling ultrasonic waves in various mode display is used to detect any damage. A qualified NDT inspector who complies with ISO 9712 is required to interpret the damages form the ultrasonic signal. However, the inspection performance is subjected to human factors due to fatigue and lack of concentration. Therefore, a study of a damages detection system is carried out to detect and classify the damages. In this study, the damage detection of pre-impregnated laminated composites has been made using ultrasonic prototype machine namely ISI i-InspeX TWO and the classification from the extracted features of A-scan mode display has been performed using Back Proportional Network (BPN). The classification employs two classification stages which is CLASS-1 and CLASS-2 for the first and the second phase respectively. The results of the average performance of CLASS-1 concluded that the proposed approach attained reliable results with the accuracy of 99.99% while the performance result of CLASS-2 was 94.21%. Thus, these promising classification performances showed that the proposed system is applicable to assist NDT inspectors in their quality inspection process.
KeywordsNon-destructive testing Ultrasonic testing Fibre-glass pre-impregnated laminated composites Artificial neural network (ANN)
- 1.Federal Aviation Administration (FAA): Aviation Maintenance Technician Handbook, vol. 1. Aviation Supplies & Academics (ASA), Oklahoma (2012)Google Scholar
- 5.Sayer, M., Bektas, M.B., Demir, E., Callioğlu, F.: The effect of temperatures on hybrid composite laminates under impact loading. Compos. Part B Eng. 43, 2152–2160 (2012)Google Scholar
- 15.Dua, R., Watkins, S.E., Wunsch, D.C., Chandrashekhara, K., Akhavan, F.: Detection and classification of impact-induced damage in composite plates using neural networks. In: IEEE International Joint Conference on Neural Networks (2001)Google Scholar