Article Article
Automated Defect Classification Using Artificial Neural Networks

This work discusses on the application of Hopfield Neural network (HNN) for the automated defect classification in Carbon Fiber Reinforced Plastics (CFRP) laminates with simulated defects. Three of the most common and critical defects found in composite laminates, i.e. foreign object inclusion, porosity, and delamination due to impact damage, were considered. In order to test the classification strategy, one hundred (100) time-domain 5 MHz ultrasonic echo signals were acquired for each type of defect and from defect-free regions in CFRP panels. The recorded signals were start-point aligned with the means by using cross-correlation prior to feature extraction process. The feature extraction methodology developed consisted of five features for each signal: mean, standard deviation, peak-to-peak front-wall echo, peak-to-peak back-wall echo, and time of flight values. In addition, two other conventional feature extraction techniques: Fast Fourier transform (FFT) and Discrete Wavelet transform (DWT) were also considered for the comparison purpose. After extracting the features, the signals for each category were randomly partitioned into equal-size training and test sets. The partitioning process was repeated so that at least 1000 signals could be tested for each category and the results could be averaged to obtain a robust estimate of the performance. The first set of experiments consisted of the dichotomous classification between each type of defect and defect-free signals. The results show that classification accuracies exceeding 99% are possible for each of the three cases by using HNN. The next experiment focused on a 4-class problem: three defect types and defect-free. For this experiment too, it is shown that classification accuracies exceeding 96% are possible by using HNN. Based on the results obtained, it can be concluded that the Hopfield Neural network can classify defects in composite laminates quite accurately. With few modifications, this methodology can be used to classify other types of defects in CFRP and/or other materials.

References
1. Poudel, A., R. Kanneganti, L. Gupta and T.P. Chu, “Nearest Mean Classifier for Defect Classification in CFRP Panels,” Proceedings of ASNT 22nd Research Symposium and Spring Conference, Memphis, TN. 2. Poudel, A. and T.P. Chu, 2012, “Intelligent Nondestructive Testing Expert System for Aircraft Carbon/Carbon Composite Brakes Using Infrared Thermography and Air-coupled Ultrasound,” Materials Evaluation, Vol. 70 (10), pp. 1219-1229. 3. Carvalho, A.A., J.M.A. Rebello, L.V.S. Sagrilo, C.S. Camerini and I.V.J. Miranda, 2006, “MFL Signals and Artificial Neural Networks Applied to Detection and Classification of Pipe Weld Defects,” NDT & E International, Vol. 39(8), pp. 661-667. 4. Moura, E.P., R.R. Silva, M.H.S. Siqueira and J.M.A. Rebello, 2004, “Pattern Recognition of Weld Defects in Preprocessed TOFD Signals Using Linear Classifiers,” Journal of Nondestructive Evaluation, 23(4), pp. 163-172. 5. Acciani, G., G. Brunetti, E. Chiarantoni and G. Fornarelli, “Automatic Evaluation of Flaws in Pipes by Means of Ultrasonic Waveforms and Neural Networks,” Proceedings of IEEE International Joint Conference on Neural Network, pp. 892-898. 6. Ramuhalli, P., J. Kim, L. Udpa and S.S. Udpa, 2002, “Multichannel Signal Processing Methods for Ultrasonic Nondestructive Evaluation,” Proceedings of IEEE Sensor Array and Multichannel Signal Processing Workshop, pp. 229-233. 7. Lee, K. and V. Estivill-Castro, 2004, “A Hybrid Classification Approach to Ultrasonic Shaft Signals,” Proceeding of Advances in Artificial Intelligence, Vol. 3339, pp. 284-295. 8. Yegnanarayana, B., 2001, Artificial Neural Networks, Prentice-Hall of India, New Delhi, India. 9. Haykin, S., 1999, Neural Networks: A Comprehensive Foundation, Prentice Hall, Upper Saddle River, N.J.
Metrics
Usage Shares
Total Views
154 Page Views
Total Shares
0 Tweets
154
0 PDF Downloads
0
0 Facebook Shares
Total Usage
154