Gear Fault Diagnosis in Time Domains by Using Bayesian Networks

Fault detection in gear train system is important in order to transmitting power effectively. The artificial intelligent such as neural network is widely used in fault diagnosis and substituted for traditional methods. In rotary machinery, the symptoms of vibration signals in frequency domain have been used as inputs to the neural network and diagnosis results are obtained by network computation. However, in gear or rolling bearing system, it is difficult to extract the symptoms from vibration signals in frequency domain which have shock vibration signals. The diagnosis results are not satisfied by using artificial neural network, if the training samples are not enough. The Bayesian networks (BN) is an effective method for uncertain knowledge and less information in faults diagnosis. In order to classify the instantaneous shock of vibration signals in gear train system, the statistical parameters of vibration signals in time-domain are used in this study. These statistical parameters include kurtosis, crest, skewness factors etc. There, based on the statistical parameters of vibration signals in time-domain, the fault diagnosis is implemented by using BN and compared with two methods back-propagation neural network (BPNN) and probabilistic neural network (PNN) in gear train system.