Development of Effective and Efficient Acoustic Emission Based Gear Fault Diagnosis Methods and Tools

2016-06-21T00:00:00Z (GMT) by Yongzhi Qu
Acoustic emission (AE) based sensing technology is considered an emerging technique for rotating machine fault diagnosis even though it has been successfully applied to non-destructive testing of static structures for many years. In comparison with the widely used vibration based techniques, it has numerous advantages. For example, it is capable of incipient fault detection. It is sensitive to the location of the faults and therefore could be used for fault location detection. However, there are a number of challenges in order to apply AE based technology for rotating machine fault diagnosis. In comparison with other sensors such as vibration sensors, AE sensors require a much higher sampling rate. The characteristic frequency of AE signals generally falls in the range of 100 kHz to several MHz, requiring a sampling rate of at least 2 MHz for AE data acquisition. AE data acquired with such a high sampling rate would add a tremendous burden on data storage and analysis for machine health monitoring, fault diagnosis and prognosis. In addition, there is a lack of well-developed signal processing methods that could effectively take advantage of the known structures of the machinery for fault diagnosis. In this dissertation, effective and efficient AE based methods and tools for gearbox fault diagnosis have been developed and validated with gearbox seeded fault tests on a notational split torque gearbox. Specifically, a frequency reduction method has been developed based on the heterodyne technique commonly used in telecommunication to reduce the AE data sampling rate to as low as 20 kHz. By heterodyning, the AE signal frequency can be reduced from several hundred kHz to below 50 kHz. Also through heterodyning, the AE signals can be demodulated to remove less useful high frequency components while keeping the fault characteristic frequency components in the demodulated AE signals. As a result, the demodulated AE signals can be sampled at a low rate comparable to that of vibration sensors. In order to extract useful features from AE signals sampled at a low rate, an effective AE signal processing method for gearbox fault diagnosis based on time synchronous average (TSA) has been developed. This is the first reported research effort in developing a physics based gearbox fault diagnosis method using AE sensors. The developed AE based gearbox fault diagnosis methods and tools have several significant advantages. First, the heterodyne based frequency reduction method could down shift the sampling rate to that comparable to the vibration signals. The original meshing frequencies of the gearbox can be retained in the AE signals sampled at a low rate. This enables well developed vibration analysis methods to be applied efficiently in practice to the AE signals for gearbox fault diagnosis, which is one of the major contributions of this work. Also, this could reduce the storage and computational burden for further signal processing so that great cost reduction can be achieved. By using TSA, the knowledge of the physical structure of the gearbox can be utilized effectively and efficiently for fault diagnosis. This is different from any of the previous data driven methods which completely rely on a black box type of reasoning for gearbox fault diagnosis. A comparative study between vibration analysis and AE analysis has also been performed. Different levels of tooth cut faults have been seeded and tested with both vibration and AE data collected. Results have shown that AE based approach has the potential to differentiate gear tooth cut levels in comparison with vibration based approach. While vibration signals are easily affected by mechanical resonance and background noises, the AE signals show more stable performance. The effectiveness of AE analysis under low sampling rate has also been investigated. The results have shown that AE sampling rate could be as low as 20 kHz without serious performance degradation.