| Summary: | 碩士 === 中原大學 === 電機工程研究所 === 100 === Electric vehicles (EV) have become one of the major issues of industry in recent year. The capacity of battery and motor manufacture are promoted dramatically. This thesis studies step motors and wheel motors, and the current signals and vibration signals in these motors are analyzed to detect the bearing fault. Since bearing faults can be diagnosed quickly, the EV’s operation could be more stable.
This thesis proposes a system to detect the bearing fault. First, various damage types of motor bearings derived from electrical discharge machining (EDM) are tailor-made. Then envelope analysis (EA), wavelet transform (WT) and Hilbert Huang transform (HHT) respectively analyzed the measured current and vibration signals to extract the feature of motors. Back propagation neural network (BPNN) are applied the features to recognize the bearing faults. The feasibilities of these manners are also demonstrated by cross-validation in the thesis. Secondly, this research proposes a manner which uses greedy algorithm based on empirical mode decomposition selector (GEMDS) to select the optimal intrinsic mode function (IMF). The selected optimal intrinsic mode functions (IMFs) can best respresent the characteristic of motors since the manner effectively eliminate nosie signals measured from the motors. The result indicates that this approch can improve the definition of the HHT spectrum, and more efficiently diagnose motors.
Finally, the motor signals are putted nosie under 30dB, 20dB and 10dB signal-to-noise ratio (SNR) to validate the robustness of methods. The GEMDS based HHT has better noise-resistance than traditional HHT has in step motors and wheel motors. The experiment results indicate that GEMDS decreases the number of IMFs’ layers to detect motor fault effectively.
|
|---|
