Fault diagnosis of bevel gearbox using artificial intelligence technique

Abstract

The project consists of the diagnosis of bevel gear faults as fault diagnosis is the important parameter for the rotating machinery. Hence the early detection of the fault will result in avoiding the serious damage to the machinery or any harm to the operator working on it. As gearbox is a crucial component of any machinery the regular monitoring of its condition is required. The study in this thesis is focused on prediction of faults using artificial intelligence technique. The vibrational responses are obtained and analyzed for various defaults possible to occur in bevel gear inside the gear box. The bevel gear with defects like one chipped tooth and one missing tooth are used for the experiment. The various techniques used for fault assessment considering vibrational data explications includes artificial neural network (ANN), deep neural network (DNN), support vector machine (SVM) and random forest (RF). For the reduction of dimensionality in the original vibration signal, the classifiers are clubbed with first attribute evaluators alone and then with wavelet transform. The results are compared for having the combination giving us the best results. This study is the combination of two stages. In the former stage, the vibrational signals are optimized using three attribute evaluators one by one and further the data is given to classifier as input and finally the best accuracy is evaluated. In the second stage, the wavelet transform is used to decompose the signals and again other attribute evaluator is used with classifier and to achieve the best accuracy. The result also shows that this algorithm can be effectively used for other components subjected to vibrations during its functionality, mainly for rotating components.