Abstract
AN OPTIMAL MULTILAYER PERCEPTRON-BASED NEURAL NETWORK FOR DETECTION AND CLASSIFICATION OF FAULTS IN A THREE-PHASE PERMANENT MAGNET SYNCHRONOUS MOTOR
P. G. Asutkar* and Dr. Z. J. Khan
ABSTRACT
Permanent Magnet Synchronous Motor (PMSM) is subjected to various operating, environmental, and other conditions; due to which incipient faults occur. If these faults are undetected, lead to catastrophic failure. For reliable and safe operation of PMSM. Online condition monitoring, fault detection, and diagnosis were required. Many researchers have proposed various techniques for fault detection and diagnosis, which requires good domain knowledge and costly pieces of equipment. This paper presents an optimal multilayer perceptron (MLP) neural network for fault detection and classification, which is simple, reliable, and cost-effective. Two faults are created on a three-phase permanent magnet synchronous motor stator inter-turn and eccentricity with varying load conditions. The experimental data is generated on 1 hp, 3 phase, 4 poles, 1500 rpm permanent magnet synchronous motor during healthy and faulty conditions. Various sensors are inbuilt internally and externally into the PMSM motor for the measurement of different parameters. 12 different measurable parameters include three-phase motor intake current, three-phase motor applied voltage, power factor, winding temperatures, and bearing sound. The proposed classifier with 12 input parameters is designed and verified for optimal performance for fault identification and classification, Nearly 100% classification accuracy is achieved.
[Full Text Article]