Experimental comparison of gearbox fault detection methods using acoustic and vibration signals

Over the past three decades, a variety of condition monitoring methods have been utilized for gearbox applications. The methods mainly rely on using vibration signals acquired from vibration sensors attached to a gearbox system. Although the use of vibration signals has been successful in providing accurate gearbox fault diagnosis for different various fault types, the requirement for contact vibration sensors may restrict the sensor placement, such as at gearbox locations that are directly exposed to moving components, high temperature or harsh chemicals. The work presented here experimentally investigates an alternative use of non-contact acoustic sensors for detection of gearbox faults by extracting useful dynamics information from acoustic signals. An experimental comparative study based on the utilization of either vibration or acoustic signals for gear fault diagnostics, is performed. For this study, a gearbox test rig with the broken tooth fault type is constructed, which is combined with attached accelerometers and a microphone for acquiring vibration and acoustic signals from the operating gearbox. The measurement noise that contaminates the raw acoustic and vibration signals is filtered by using the Time Synchronous Averaging (TSA) Method to remove the asynchronous signal components for improving the accuracy of fault diagnosis. Furthermore, the frequency domain analyses of both vibration and acoustic signals are performed to identify spectral characteristic differences of signals acquired from the healthy and faulty gearbox systems. A further investigation is undertaken by incorporating the Short Time Fourier Transform (STFT) analysis that provides a simultaneous representation of the faulty tooth location and the sidebands associated with the fault. The results demonstrate that acoustic signals can also provide a reliable fault diagnostic information compared to the commonly used vibration signals, hence providing an alternative sensing solution for gear fault diagnosis.