Development of an Intelligent Shaft Fault Diagnosis System for Machine Tools

Aiming at reducing cost and time of repair, condition-based shaft faults diagnosis is considered an efficient strategy for machine tool community. While the shaft with faults is operating, its vibration signals normally indicate nonlinear and non-stationary characteristics but Fourier-based approaches have shown limitations for handling this kind of signals. The methodology proposed in this research is to extract the features from shaft faults related vibration signals, from which the corresponding fault condition is then effectively identified. Besides Fourier Transform, two new algorithms are used to extract the feature of signals, empirical mode decomposition (EMD) and multi-scale entropy (MSE). With an incorporation of EMD method, the model applied in this research embraces some characteristics, like zero-crossing rate and energy, of intrinsic mode functions (IMFs) to represent the feature of the shaft condition. The other method called MSE is used to calculate the entropy of multi-scale of the signal. The curve of MSE can be used to identify some defect model of shafts clearly. The conventional approach for online monitoring of a machine's health based on linear time-frequency analysis has its limitation, as the mechanical vibration signal is nonlinear and non-stationary in nature. Multi-scale entropy (MSE) is a recently developed method for nonlinear time series analysis, which quantifies a system's complexity. EMD-based and MSE-based methods were implemented to develop a diagnosis system in this research. In the buildup stage a knowledge ware is created from the database compiled from the existing defect models. Finally, diagnosis system is a monitor system to be implemented in a machine tool manufacturing company to verify the ability.