Optimum Multi-Fault Classification of Gears with Integration of Evolutionary and SVM Algorithms
No Thumbnail Available
Of late, the subject of machine condition monitoring as a component of maintenance system became popular worldwide due to the prospective advantages to be gained from reduced maintenance costs, improved productivity, increased machine life time and safer work places. Vibration signals extracted from rotating parts of machineries carry a lot of information within them about the condition of the operating machine. Further processing of these raw vibration signatures, measured at convenient locations of the machine, unravels the condition of components or assemblies under study. Gears in rotating machines are major components of interest for the vibration based condition monitoring. Their failure causes increase in the downtime and the maintenance cost. Artificial intelligent techniques, like artificial neural networks (ANN), the fuzzy logic system (FLS), the fuzzy neural network and the support vector machine (SVM), have been commonly used in the machine fault diagnosis. However, in the application of SVM for the machine condition monitoring and fault diagnosis still a great deal of work to be explored. The SVM has excellent performance in the generalization, so it can produce high accuracy in the fault classification of the machine condition monitoring and diagnosis. The use of SVM in the expertise and problem orientated domain in machine condition monitoring is growing rapidly. The main focus of the present thesis is to examine the performance of the multiclass ability of the SVM technique in a gear box by optimizing SVM parameters using the grid-search method (GSM), genetic algorithm (GA) and artificial bee colony algorithm (ABCA). Four fault conditions, i.e. the chipped tooth (CT), the missing tooth (MT) and the worn tooth (WT) along with the normal gear (or no defect, i.e. ND) have been considered.
Supervisor: Rajiv Tiwari