PREDICTIVE MAINTENANCE FOR CEMENT FACTORY PLANT EQUIPMENT USING MACHINE LEARNING METHODS

Abstract

Maintenance is essential in ensuring smooth and reliable operation of equipment in the cement plant. Predictive maintenance stands to be cost effective, ensure quality product and plant safety compared to corrective and preventive maintenance. Induction motor plays a crucial role in operation of kiln in the setup of the cement factory. This studyused machine learning models to predict the maintenance conditions of induction motor main drive based on three historical datasets of some of its components.The dataset consist of motor current signature analysis that is made up of rotor current measurements as its variables. The study tested five machine learning models, namely, decision tree, k-nearest neighbours (kNN), support vector machine (SVM), gradient boost tree (GBT), and random forest (RF) to ensure outstanding outcome.A 25:75 ratio holdout validation was used in the study. It has been found that four of the models could accurately predict condition of the induction motor main drive. However, the kNN model performed the best due to its ability to handle nonlinear relationships.It has accuracy of 89.47%, precision of 87.82 %, recall of 87.82% and f-score of 87.82% for the rotor cable dataset 1, while GBT has the least performance among the prediction models with accuracy of 68.42%, precision of 68.42%, recall of 50% and f-score of 57.78%. The performance for the other datasets shows similar trendto the one obtained in the rotor cable dataset with kNN having the best performance and GBT has the least performance among the prediction models. Therefore, GBT model...