Development and Evaluation of a Multi-Sensor IoT System for Real-Time Condition Monitoring and Fault Classification of an Industrial Electric Motor

Abstract

Industrial electric motors require continuous condition monitoring because developing electrical, thermal, and mechanical abnormalities can lead to unplanned shutdowns and equipment damage. This study developed and evaluated a multi-sensor Internet of Things system for real-time condition monitoring and fault classification of an industrial electric motor driving a fire-hydrant pump under field operating conditions. An ESP32 edge device was integrated with an MLX90614 infrared temperature sensor, an MPU6050 three-axis accelerometer, PZEM-004T electrical measurement modules, and an optical rotational-speed sensor. The system acquired temperature, vibration, phase voltage, current, frequency, power, power factor, and rotational-speed measurements. The acquired data were transmitted through authenticated Message Queuing Telemetry Transport communication to Node-RED for processing and routing, stored in InfluxDB Cloud, and presented through Grafana dashboards for real-time monitoring, trend analysis, and alert generation. Field measurements were combined with systematically labelled electrical, thermal, and mechanical fault-condition records to develop the classification dataset. Four recurrent deep-learning architectures—Long Short-Term Memory, Gated Recurrent Unit, Bidirectional Gated Recurrent Unit, and attention-enhanced Gated Recurrent Unit—were trained and comparatively evaluated using multimodal time-series sequences. The LSTM model achieved a test accuracy of 99.95%, while the GRU, Bi-GRU, and GRU-Attention models each achieved 99.97%. A hybrid decision mechanism further compared model predictions with predefined operational thresholds to identify inconsistent classifications and provide an independent condition check. The findings demonstrated that the developed system supported continuous field monitoring, reliable cloud-based data handling, and accurate classification of motor operating conditions. The system provides a practical basis for data-driven fault detection and predictive maintenance decision support for industrial electric motors.