.png){kind=icon}
.png){kind=icon}
Return
A Robust Fault Diagnosis Method for Rotating Bearing of Inverter-Fed Machine Using PWM Switching Oscillations
.png){kind=icon}
PRE
AI
H
J
D
J
Y
DOI：10.1109/jestpe.2026.3687927.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Bearing fault diagnosis is very important for the safe and reliable operation of inverter-fed machine systems. Bearing fault diagnosis on inverter-fed machines has long been a research hotspot with fruitful results. In recent years, deep learning (DL) has been widely used in bearing fault diagnosis due to its strong feature extraction ability for big data. However, the bearing of an inverter-fed machine operates under variable working conditions, making it difficult to meet the requirements of DL methods in practical scenarios. Therefore, this article proposes a bearing fault diagnosis method based on high-frequency (HF) pulsewidth-modulation (PWM) switching oscillations. The approach utilizes MHz-level common-mode current single switching oscillation segments extracted via an endpoint detection (ED) algorithm for online mechanical bearing fault diagnosis. First, the mechanism of bearing faults and the bearing current in inverter-fed machines are analyzed. This current, as a component of the motor’s common-mode current, can be measured non-invasively. Second, to address the HF and weak nature of the signal, a dedicated ED algorithm is designed to reliably extract the maximum-energy oscillation events strongly correlated with the bearing condition. Finally, a lightweight 1-D convolutional neural network (1DCNN) is constructed. Experimental validation on a 3-kW induction motor (IM) drive system confirms the accuracy and operational adaptability of the proposed method.
Keywords:
Convolutional neural network (CNN)
fault diagnosis
inverter-fed machine
switching oscillation
Journal
I
IF:
4.9
Papers:
249
Citations:
0
Organization
Cited Papers