Failure prediction of induction motors: a case study using CSLGH900/6-214, 5.8 MW, 11 KV/3PH/50HZ Sag Mill Motor at Goldfields Damang Mine.

Abstract

This research proposes a generalised feed-forward artificial neural network model that fulfils the failure prediction of CSLGH900/6-214, 5.8 MW, 11 kV/3 ph/50 Hz slip ring SAG Mill induction motor at Goldfields Ghana Limited, Damang Mine. It provides a general understanding of three phase induction motors, faults associated with induction motors and also emphasizes the use of intelligent system, particularly artificial neural network, in modern failure prediction technology of induction motors. Site analysis and motor data (Current, Power and winding temperatures) collection were conducted at Goldfields, Damang Mine. Simulation results are presented using MATLAB software (2017a) package to develop the fault prediction model. The proposed feed-forward neural network used the Levenberg-Marquardt and Bayesian Regularisation in training. The research also employed the use of Log sigmoid and Tan sigmoid as the activation functions of the hidden layer, with hidden layer size being kept at 10 neurons. The simulation and calculation were done with real-time on-load measurement from the SAG Mill motor. Analysis of the model’s output performance were done using correlation of coefficient of network performance, R and Mean Squared Error, MSE. When the proposed model is implemented, common faults could be prevented from escalating into major breakdowns, thereby reducing the downtime and hence increasing the availability of the motor for general operation. In the absence of major breakdowns, safety of employees and equipment could be assured. It is therefore worthwhile to invest in deploying this model to augment the conditional monitoring needs like temperature, current and vibration of the SAG Mill motor and other such equipment like the Ball Mill motor in the plant.