A neural network architecture for ultrasonic nondestructive testing

The application of neural networks for detecting defects in materials consisting of dense microstructures using ultrasonic pulse-echo systems is discussed. The motivation for this work is the desire to detect defects modeled by small complex scattering centers. To preserve signal features useful for defect detection, minimal preprocessing is performed on the data presented to the neural network. Modifications to the typical feedforward multilayer perceptron architecture for direct application to sampled RF ultrasonic A-scans are described