Fault Diagnosis of Automotive Gearbox Based on Signal Processing Using Machine Learning Techniques

Abstract

Dynamic mechanical systems such as gearboxes have wide automobile and industrial applications. The unexpected failures of gears and bearings are not only extremely damaging to the system but also lead to economic loss and serious safety issues. Due to continuous operations and various cyclic loading conditions, faults occur in the components of the gearbox which lead to the risk of catastrophic failure. Condition monitoring is a technique that has been deployed over the years for the predictive maintenance of a gearbox. It has been used to find the state of the gearbox and predict the presence of faults at the initial stage to prevent breakdown. Thus, it helps in reducing the maintenance costs and the downtime caused due to faults. This thesis work relies on data driven approach for fault diagnosis of gearbox. The main objective of this work is to develop a systematic approach to fault diagnosis of an automotive gearbox by acquiring and fusing the vibration signals, acoustic emission signals, and sound signals and then classifying these signals using efficient machine learning algorithms. Different signal processing methods have been employed by engineers and scientists to gather information about the condition of a gear box to test or schedule the maintenance activities. These methods include vibration analysis, acoustic emission, sound signals, oil debris analysis, visual inspection and various non-destructive techniques. Vibration signal analysis is one of the widely used method for detecting the condition of the gearbox. Very few works have been carried out using sound signals and acoustic emission signal analysis compared to vibration signal analysis for the fault diagnosis of a gearbox. The reason behind this approach is that the gearbox produces lot of sound which contains low signal to noise ratio in its running environment. Microphone is used to detect the unusual level of sound produced due to the presence of fault where the sound level exceeds the threshold range. The application of acoustic emission..