IOT-BASED PREDICTIVE MAINTENANCE SYSTEM USING MACHINE LEARNING FOR FAULT DETECTION IN HOUSEHOLD APPLIANCES

Abstract

The spread of Internet of Things (IoT) devices in smart homes has provided revolutionary possibilities of real-time monitoring and predictive maintenance of household appliances. The paper introduces a thorough IoT-based predictive maintenance system combining machine learning (ML) algorithms to predict defects early in typical household appliances, such as refrigerators, washing machines, air conditioners, and water heaters. The suggested system implements low-cost sensor nodes with vibration sensors, current transducers, temperature sensors and acoustic emission sensors connected to microcontroller units (MCUs) transmitting time-series data to a cloud-based analytics system. A set of 42780 labeled samples of both fault and normal operation were used to train four ML classifiers: Random Forest (RF), Support Vector Machine (SVM), Long Short-Term Memory (LSTM) networks, and Gradient Boosting (GB). The experimental results indicate that the LSTM-based model is the most accurate in detecting faults with the highest fault detection accuracy of 97.3% and a precision of 96.8, recall of 97.1, and F1-score of 96.9, which is 23.4 percentage points higher than the traditional threshold-based methods. The system saves 61% of unnecessary maintenance interventions and increases average lifespan of appliances by an estimated 2.3 years. The average household saving of 18.7% in monthly utility spending is achieved with energy consumption anomaly detection.