Condition monitoring of industrial exhaust motor-fan assembly using statistical feature optimization with nature-inspired optimizers

Exhausters are very crucial components in sinter plant of steel manufacturing industries. Problems in fan or motor side of the exhausters lead to machine stoppages or reduced speed, resulting in stoppage of the system and decreased sinter production. Common issues include imbalance, misalignments, looseness, eccentricity, and motor bearing failure. The proposed IoT-based system aims to continuously analyze vibration signals, predict issues, and prevent sudden stoppages and exhauster tripping, thereby enhancing overall plant efficiency. This study focuses on online monitoring and health prediction of industrial exhausters by selection of statistical features extracted from triaxial time-series vibration data by using multiple optimizers, followed by implementation of machine learning classifiers. By incorporating a novel feature selection process based on nature-inspired optimizers, this methodology aims to enhance the overall generalization performance of classification. The retrained models with cross-validation were evaluated with real-time industrial data, which shows 99.5% cross-validation accuracy and 95.83% real-time prediction accuracy. The proposed methodology holds great potential for detecting machine faults in industrial applications.