Fault Diagnosis of Rotor Broken Bar in Induction Motor Based on Variable Mode Decomposition

When the induction motor is operated under the steady-state operating condition with a light load, the fault characteristic frequency of the broken bar fault is close to the fundamental frequency, making it difficult to extract the fault current component from the stator current effectively. Moreover, the fault characteristic signal energy shares a fairly minor fraction of the total stator current energy in the early stages of incomplete broken bars, which increases the difficulty of fault current component extraction. In order to improve the diagnosis effect of the broken bar fault at the early stage when the motor is running at light load, according to the characteristics of distinguishing the characteristic signal of the broken bar fault from the fundamental signal during the starting period, this paper proposes the variational mode decomposition (VMD) method to process the stator start-up current, and the broken bar fault diagnosis is carried out by utilised the energy value of the fault characteristic current obtained by decomposition. The proposed broken bar fault diagnosis scheme is mainly divided into three stages: stator current pre-processing, fault characteristic current extraction, fault characteristic current energy time-frequency distribution calculation and fault severity quantification. In the first stage, the Clarke transformation is applied to obtain the current space vector, and the amplitude variation of the current space vector during the starting process is utilised to adaptively extract the stator current for analysis, and the influence of high frequency interference on low frequency signal is removed by an anti-aliasing filter. In the second stage, a VMD algorithm based on the optimisation of the mean instantaneous frequency is proposed to extract the left frequency component of the fault characteristic current. In order to optimise the extraction effect, it is necessary to determine the appropriate parameters for the penalty factor and the number of modes in the VMD algorithm. The penalty factor should take into account the narrow-band nature of the fundamental signal and the broad-band nature of the fault characteristic current, and have low sensitivity to the change of the energy value of the fault characteristic current to accommodate to the diagnosis under different fault degree. Concerning the number of modes, in this paper we propose to determine the optimal number of modes to ensure the energy concentration of the fault characteristic current by exploiting the variation law of the centre frequency and the mean instantaneous frequency of the sub signals obtained from the VMD decomposition. In the third stage, Wigner-Ville Distribution (WVD) is utilised to extract the energy time-frequency distribution of the fault characteristic current, and Otsu’s method (OTSU) is applied to perform anti-noise processing on the time-frequency distribution to highlight the fault characteristic information. In order to further improve the disturbance rejection performance of the diagnosis scheme, we take the energy value of the characteristic curve of the right part of the left frequency component as the fault quantization factor, and set the threshold based on the mean and standard deviation of multiple groups of experimental data under different fault states, so as to realise the purpose of automatic fault warning. In this paper, the experimental verification is carried out on a 5.5 kW induction motor. The results show the proposed fault diagnosis scheme performs well in terms of interference rejection and is able to extract the fault characteristic current component in the condition of incomplete broken bars fault at high noise level. Moreover, the energy value of the right part of the fault characteristic signal has a strong positive correlation with the fault severity, which can perform the rotor broken bar fault detection well and can realise the discrimination of different rotor broken bar fault severity including incomplete broken bars. © 2023 Chinese Machine Press. All rights reserved.