Fault Detection in Bevel Gears Using Existing and Proposed Condition Indicators through Vibration Signals

Abstract

Fault detection at an incipient stage is of utmost importance to avoid catastrophic failures in machine elements such as the gearbox. Various techniques have been presented in the literature to find out the condition of gears. The evaluation of Condition Indicators (CI) using vibration signals is simple and a promising technique that is used to detect the fault condition in gearboxes. The raw vibration signals of gears or bearing faults are often categorized as well-affected and non-stationary by vibrations from other components in the transmission path and equipment. Therefore, the useful information of the signals is often restricted by extra noises. Therefore, an efficient denoising method is necessary before analyzing the signals. In this work, a comparative analysis among various existing CIs (RMS, peak, peak to peak, impulse kurtosis, standard deviation, crest factor, clearance factor indicator, shape factor and Shannon entropy) and three proposed CIs (PI-I, PI-II, and PI-III) have been presented before and after denoising for missing tooth chipped tooth, and, healthy tooth condition of bevel gear for constant input speed. Denoising of the raw vibration signal was done through Discrete Wavelet Transform (DWT) by using block JS as the denoising method. The experimental investigation was performed on a machine fault simulator (MFS) and vibration data as time domain signals were acquired from healthy as well as faulty gears using an accelerometer. The results indicate the response of various CIs before and after denoising the presence of a fault in bevel gears.