Enhancing Predictive Maintenance in Industrial IoT: A Machine Learning Approach for Real-time Anomaly Detection in Pressure Switches

This study investigates the real-time integration of machine learning techniques with Internet of Things (IoT) data to monitor and predict the behavior of pressure switches in an industrial environment. Predictive maintenance in industrial IoT systems is critical to increase operational efficiency and minimize unexpected failures. The increasing complexity of industrial processes has made integrating machine learning algorithms and IoT data a powerful solution for proactive maintenance. In this context, the study aims to perform anomaly detection and failure prediction in pressure switches by analyzing real-time sensor data and applying Random Forest, Isolation Forest, and Local Outlier Factor algorithms. The performance of the models is evaluated using the MetroPT-3 Train Dataset. Performance metrics like accuracy, precision, recall, and F1 score assess the models' effectiveness. The Random Forest Classifier showed the highest performance in anomaly detection with an accuracy rate of 99.92%. The findings emphasize the significant potential of machine learning and the Internet of Things in enhancing predictive maintenance, improving system reliability, and contributing to the broader field of industrial IoT.