SMART MAINTENANCE STRATEGIES FOR ELECTRICAL SYSTEMS USING MACHINE LEARNING

Abstract

Predictive maintenance (PdM) is a proactive strategy aimed at anticipating equipment failures before they occur, thereby ensuring the reliable operation of electrical systems. By leveraging advanced data analytics and machine learning, PdM analyzes machinery conditions and performance metrics to enable timely and efficient maintenance. This approach is critical for minimizing downtime and reducing maintenance costs in industrial operations. Traditional maintenance strategies are typically either reactive or preventive. Reactive maintenance, or "run-to-failure," involves repairing equipment only after a breakdown, resulting in significant downtime and higher costs. Preventive maintenance, performed at regular intervals regardless of equipment condition, often leads to unnecessary servicing and does not account for actual wear and tear. Both methods fail to fully utilize the vast amount of operational data generated by modern machinery, missing opportunities to optimize maintenance schedules based on real-time equipment conditions. The primary limitation of traditional maintenance strategies lies in their inefficiency and cost. These approaches do not incorporate real-time data to accurately predict maintenance needs, leading to either excessive servicing or unexpected equipment failures. This highlights the need for a more intelligent system that can predict maintenance requirements based on actual equipment conditions. The motivation for this project is to enhance the efficiency and reliability of electrical equipment maintenance. Machine learning-based predictive maintenance can significantly reduce downtime, lower operational costs, and extend equipment lifespan. By using data from various sensors that monitor critical operational parameters, this project aims to predict potential failures and ensure maintenance is conducted only when necessary. The proposed system employs a Decision Tree Classifier to predict maintenance requirements using a comprehensive dataset of operational parameters. The dataset includes features such as RPM, motor power, torque, pressure, airflow, noise levels, temperature, and acceleration, along with categorical labels for components like bearings, water pumps, radiators, exhaust valves, and AC motors.