Diagnostic algorithm for analysis of induction motors current based on a hodograph estimation of the generalized current vector and frequency spectra

Abstract

The article considers a diagnostic algorithm for analyzing the technical condition of an asynchronous electric motor based on the signals of the consumed linear current. The purpose of the development of this algorithm is the timely detection of defects of asynchronous electric motors in their operation mode. A literature review on the statistics of the most common malfunctions and methods of their diagnosis is carried out. These include phase breakage, inter-turn closure, phase closure to the core, rotor defect and rotor eccentricity. The disadvantages of existing approaches based on bench tests can lead to untimely diagnosis and, consequently, the development of a malfunction to a critical level and damage to the nodes adjacent to the electric motor. In this regard, diagnostics in operation mode is required, which will prevent aforehand the development of defects in the asynchronous electric motor. Diagnostic methods can be divided into four types: time domain current analysis, neural network current analysis, frequency domain current analysis, and generalized current vector hodograph analysis. The last two methods were chosen for the developed algorithm. They allow you to reliably classify specific types of faults, and the parallel use of the both increases the robustness of the system. To test the operation of the algorithm, a test bench was made, as well as the following malfunctions were artificially recreated: inter-turn locking and a rotor defect. After testing on the stand and setting the response sensitivity, the algorithm is tested on three electric motors of a real electric train. This algorithm is adapted for the diagnosis of asynchronous motors with a squirrel-cage rotor and the «star» type connection. However, in the future it is planned to scale it on motors with a phase rotor and motors with a «triangle» connection.