Shaft Crack Identification using Artificial Neural Networks and Wavelet Transform data of a Transient Rotor

The dynamics and diagnostics of cracked rotors have been gaining importance in recent years. The present work deals with the detection of crack in rotor system and determination of crack parameters i.e. position of crack, depth of crack using Artificial Neural Networks (ANN). The detection of crack is based on the continuous wavelet transform (CWT) plots with 1/3 and 1⁄2 critical peaks as the crack indicators. The input to the CWT is the transient data which is obtained from the cracked run-up rotor with stepped acceleration, which is encountered in real time situations during the operation of steam turbines, aeroengines etc. The amplitudes of sub-critical and critical peaks of the CWT plots are fed as input to the neural network to obtain the crack parameters. The new concept in the present work is that both qualitative and quantitative treatment has been applied to diagnose the system and by using a simple data modification technique the number of sensors have been deduced from 21 to 3 for a 20 element FEM rotor model.