V2I6P58

Advanced Machine Learning Algorithms for Predictive Maintenance in Industrial Manufacturing Systems

Vennila P.^1*, Maniraj V.²

Abstract

Predictive maintenance, which is used in industrial manufacturing systems, is the process of identifying equipment that is likely to fail early on in order to increase productivity and reduce downtime. This study’s primary goal is to examine the potential for improving predictive maintenance techniques through the application of contemporary machine learning techniques. We evaluate the efficacy of several contemporary deep learning, ensemble, and anomaly detection techniques in forecasting maintenance needs using sensor and operational data. The results of each algorithm’s work are evaluated for predictive accuracy, feature ranking, and anomaly detection with the aid of a substantial amount of data. The results show differences in the algorithms’ accuracy, precision, and recall as well as the deep learning models’ capacity to recognize complex and unusual patterns. Performance metrics and feature importance visualizations are used to show how well the maintenance predictions performed. It also outlines the shortcomings of the current models, including the issues with generalization and data quality. The study highlights the potential for using advanced machine-learning techniques to raise PM’s efficacy in industrial settings. Future research could focus on improving the developed models’ capacity for generalization and broadening the application of contemporary machine learning trends to improve maintenance tactics.

Keywords:

Predictive Maintenance, Deep Learning, Sensor Data, Algorithms, Sensors, Convolutional neural networks (CNNs).