MATERIAL CLASSIFICATION THROUGH NEURAL NETWORKS

Ultrasonic back wall echoes received from copper and aluminium plates of varying thicknesses are classified through neural network analysis for in situ material identification. To reduce the effect of thickness variation on the time domain signals, and the dimensionality, the Kar hunen-Loeve transform was explored. Enormous data compression was achi eved; however, the dimensionality of the reduced space was not constan t and increased with the incorporation of the new ultrasonic signals f rom samples of different thicknesses. The power spectra in the frequen cy domain, on the other hand, was concentrated in the initial few disc rete frequency components independent of thickness. A multi-layered fe ed-forward artificial neural network was trained by the frequency doma in signals of the two classes. It was found that the performance of th e learned network was quite reliable on the test samples even in cases where the thickness of the test sample is different from the learned samples.