Classification of Ultrasonic Shaft Inspection Data using Discrete Wavelet Transform

K. Lee and V. Estivill-Castro (Australia)


Neural Networks, Learning Algorithms and Training,Wavelets, Signal Processing Applciations in Engineering


Artificial Neural Networks (ANN) have been used to pro cess ultrasonic signals for many non-destructive scenarios. However, these scenarios normally involve very shallow surfaces. When testing shafts, the signals are long and the new problem of mode-converted reflections emerges. They are echoes that do not correspond to cracks in the mate rial, neither to characteristics of the shaft. Also, the length of the signals demands the application of feature extraction mechanism to reduce the dimension of the pattern vectors and make classifier training feasible. The results here es tablish experimentally that DWT provides faster and more reliable feature extraction for ANN in these long signals in shafts. This results match the recent studies for shallow signals where comparisons between FFT and DWT indi cate DWT as the preferred feature extraction policy.

Important Links:

Go Back