FATIGUE CRACK SIZE ESTIMATION FROM VIBRATION INDUCED ACOUSTIC SIGNALS USING CHIRP Z-TRANSFORM AND NEURAL NET

摘要

The objective of this research is to develop a practical method to estimate the size of a crack on a rotating beam in a laboratory setting. The effort includes selecting a sensor and a measurement variable, devising a method for crack size estimation and carrying out experimental validations. The study employed a microphone to measure the pressure wave, i.e., acoustic signal, excited by the vibration of the rotating beam, utilized chirp-z transform to extract the 1st and 2nd mode frequencies, and established a diagnostic neural network to map modal frequencies to crack size. Four fatigue tests were conducted to initiate and then propagate a crack. Microphone outputs and crack size were periodically recorded during the four tests. Data from test No. 1 was used to calibrate the neural net and data from the other three tests were used for testing. The experimental results show that the proposed approach can provide reasonably good estimate of the crack size using only the indirect acoustic signal. The root mean square errors ranged from 0.037mm to 0.124mm over the four tests.