Development of electromagnetic models of induction motors with broken rotor bars

Abstract

This research presents the development of electromagnetic models of induction motors with broken rotor bars for extracting fault signatures. Despite being robust, induction motors are also susceptible to defects. A broken rotor bar is a severe defect that occurs due to mechanical and thermal stress. If neglected, it can lead to the total failure of the motor. Thus, condition monitoring, combined with machine learning methods, is emerging as a means to prevent the progression of defects, such as broken rotor bars. The drawback is that these methods require a training stage for fault identification, and they usually involve large databases containing faulty features that are costly to obtain from real motor tests. Therefore, this study presents an alternative and simulates six different scenarios of an induction motor using the finite element method in Ansys Maxwell software. Three different load scenarios were defined, and up to three bars were broken. The stator-current signal was chosen to evaluate the motor’s behavior, as the broken rotor bar creates sideband components due to the asymmetric current distribution. The post-processing of the current signal applied a fast Fourier transform-based method for fault signature extraction. First, the signal is transformed from the time domain to the frequency domain. Then, the identification of the sideband components was made using the first derivative information. The simulated models for one broken bar were compared to experimental data. They showed good convergence in the frequencies of the sideband components. The analysis was expanded to include even three broken bars, and ultimately, more than 70 fault features were obtained based on the amplitude of the sideband components. These fault features can be later used to build a synthetic dataset for training machine learning methods for fault diagnosis.