CASPIAN JOURNAL

MANAGEMENT AND HIGH TECHNOLOGIES

FAULT ANALYSIS OF SYNCHRONOUS MOTORS WITH PERMANENT MAGNETS BASED ON VIBRATION MONITORING USING NEURAL NETWORKS

Read Saksonov Evgeniy A., Simonov Sergey E., Gorodnichev Mikhail G., Moseva Marina S. FAULT ANALYSIS OF SYNCHRONOUS MOTORS WITH PERMANENT MAGNETS BASED ON VIBRATION MONITORING USING NEURAL NETWORKS // Caspian journal : management and high technologies. — 2022. — №2. — pp. 141-153.

Saksonov Evgeniy A. - Moscow Technical University of Communication and Informatics, Moscow, Russian Federation

Simonov Sergey E. - Moscow Technical University of Communication and Informatics, Moscow, Russian Federation

Gorodnichev Mikhail G. - Moscow Technical University of Communication and Informatics, Moscow, Russian Federation

Moseva Marina S. - Moscow Technical University of Communication and Informatics, Moscow, Russian Federation

Permanent magnet synchronous motors are becoming increasingly popular both in industry and in electric and hybrid vehicle drives. Unfortunately, these engines, like other types of engines, are prone to wear. In them, as in drive systems with asynchronous motors, rolling bearings often fail. This article discusses the possibilities of detecting this type of mechanical damage by analyzing mechanical vibrations using neural networks. The fast Fourier transform and the Hilbert transform (TG) were used to highlight diagnostic features. To automate the fault detection process, three types of neural networks were tested: Multilayer Perceptron (MLP), Radial Basis Function (RBF) Neural Network, and Kohonen Map (Self Organizing Map, SOM). The input signals of these networks were the amplitudes of the harmonic components characteristic of damage to the supporting elements, obtained as a result of the fast Fourier transform or PG analysis of the vibration acceleration signal. The efficiency of the analyzed structures of neural networks was compared in terms of the influence of the network architecture and various parameters of the learning process on the detection efficiency.

Key words: permanent magnet synchronous motors, vibration diagnostics, neural networks, predictive analytics