Application of neural networks to nondestructive evaluation

This paper proposes the use of massively parallel learning networks for classifying signals from electromagnetic transducers used in nondestructive evaluation. A brief description of the application problem is given. A distributed processing network is developed for preprocessing the transducer signal. Preprocessing is required for achieving invariance under rotation, translation and scaling of the signal. Issues relating to implementation of the preprocessing network are presented. The complete architecture and the learning algorithm are described. Finally, results of implementing the network and a brief discussion on its performance are presented. >