On-line implementation of a fault diagnosis system for three-phase induction motors

Abstract

Condition monitoring of electric machinery has received increased attention due to the advantages it offers in terms of productivity. Automation of the fault detection and diagnosis processes would not only allow for extensive monitoring but also provide advanced warnings regarding the health of machinery. The implementation of an efficient fault detection system into a form that allows for continuous on-line condition monitoring in industry could reduce the uncertainties related to machinery failures. The objective of this research is to implement the fault diagnosis system described in Kim (2001) so that continuous, on-line condition monitoring and assessment of 3-ø induction motors becomes possible. The use of only voltage and current sensors, already present in the motor control centers, for assessing the overall health of the motor, provides for a very inexpensive and non-intrusive method for condition monitoring. The use of mechanical sensors is eliminated by integrating the fault diagnosis system with a sensorless speed estimation technique. The algorithm developed in Kim (2001) is based on artificial neural networks. Introduction of artificial intelligence in the analysis isolates the need for a human expert to make a decision regarding the health of the motor and automates the monitoring process. Negative sequence components and the root mean square of the harmonics of the current residuals are used as indicators for identifying electrical and mechanical faults in the motor. The framework required for on-line implementation is developed using the LabVIEW software. The framework is composed of a neural network based current predictor and different signal processing modules. The framework is tested on-line using a 3-ø, 3 hp motor under steady operating conditions. On-line testing, using data collected off-line, is done on motors with higher power ratings to test the applicability of the framework to different motors. The current research demonstrates the feasibility of an automated fault diagnosis system for induction motors through the use of electrical measurements. The use of a speed filter instead of a speed sensor reduces the overall cost of the monitoring process. The framework is effective in identifying both mechanical and electrical faults with an effectiveness of over 91.6%.