Intelligent Industrial Motor Fault Diagnosis and Predictive Maintenance using Multi-Sensor Fusion and Machine Learning

Abstract

Industrially used electric motors are crucial parts of industries and process industries, which failure might lead to significant financial losses due to the unplanned downtime. Current methods of motor maintenance based on regular inspections and reactive maintenance strategies are unable to detect the early symptoms of the motor condition deterioration. This paper describes the development of the Intelligent Industrial Motor Fault Diagnosis and Predictive Maintenance System using the principles of data fusion from multi-sensors and machine learning for monitoring of the motor condition and its further faults prediction. The proposed system includes ESP32 microcontroller along with multiple sensors measuring the important operating parameters, namely motor current, vibration, rotation speed (RPM), and gyroscope movement. The Wokwi simulation environment has been created to mimic various operating states of the motor, namely, normal, overloaded, vibrations and fault conditions. The collected sensor data are used as the training dataset for a machine learning model to automatically predict the state of the motor health. The proposed framework provides the ability to use simulation data generation along with machine learning algorithms for a cost-efficient intelligent monitoring of motor conditions without the need for expensive industrial testing setups. The proposed methodology is a solid base for future work and can be implemented in industry using IoT remote monitoring and cloud predictive maintenance.

Introduction

This paper presents a low-cost AI-based predictive maintenance system for industrial motor fault diagnosis using multi-sensor data fusion and machine learning. Industrial motors are essential components of manufacturing plants, power stations, and automated production systems, where unexpected failures can cause production downtime, increased maintenance costs, equipment damage, and safety risks. Traditional maintenance approaches, including reactive and preventive maintenance, are often inefficient because they either respond only after failures occur or schedule maintenance regardless of the motor’s actual condition. Predictive maintenance addresses these limitations by continuously monitoring equipment health and predicting faults before they become critical.

The proposed system integrates advances in the Industrial Internet of Things (IIoT), embedded systems, and machine learning to monitor multiple motor parameters simultaneously. Unlike conventional systems that rely on a single measurement, the system combines data from current, vibration, rotational speed (RPM), and gyroscope sensors to provide a comprehensive assessment of motor health. Current measurements detect overloads and electrical abnormalities, vibration analysis identifies bearing wear and shaft misalignment, RPM indicates performance degradation, and gyroscope readings reveal abnormal shaft motion. Multi-sensor data fusion improves fault detection accuracy compared with single-sensor approaches.

The literature review highlights that traditional motor monitoring systems typically monitor only one parameter and depend on fixed alarm thresholds, limiting their ability to detect complex faults. Artificial intelligence and machine learning techniques, including Decision Trees, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Random Forests, Gradient Boosting, Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) networks, have demonstrated strong performance in motor fault diagnosis. However, many existing AI systems require expensive industrial sensors, large datasets, and powerful computing resources, making them unsuitable for educational institutions and small industries.

The proposed system addresses these limitations by using an ESP32 microcontroller with low-cost sensors and generating training data through the Wokwi simulation platform. Simulated operating conditions—including healthy operation, overload, vibration faults, and critical failures—are used to create a labelled dataset containing current, vibration, RPM, gyroscope measurements, health scores, and fault labels. The dataset is preprocessed and used to train machine learning models capable of classifying motor conditions, estimating motor health, predicting future failures, and generating maintenance recommendations.

A comparison with previous studies shows that most existing systems focus primarily on fault detection using limited sensor inputs and rarely provide additional information such as motor health scores, future risk prediction, or maintenance recommendations. The proposed system distinguishes itself by integrating four sensor types, implementing real-time monitoring on an affordable ESP32 platform, and providing comprehensive decision support for maintenance planning.

The research objectives include developing a low-cost monitoring system, collecting multi-sensor data, creating a labelled dataset, training and evaluating machine learning models, estimating a Motor Health Score (0–100%), predicting future fault probabilities, generating maintenance recommendations, and validating the approach using simulation.

The system architecture consists of four main stages: multi-sensor data acquisition, embedded processing, dataset generation, and machine learning-based decision making. The ESP32 continuously acquires sensor readings, filters noise, calculates RPM, generates feature vectors, and provides real-time feedback through an OLED display, LEDs, and a buzzer. The simulated dataset exported from Wokwi serves as the foundation for training machine learning algorithms to identify motor conditions and support predictive maintenance.

Conclusion

The paper has introduced an industrial motor fault detection and predictive maintenance system based on Artificial Intelligence with multi-sensors data acquisition and machine learning techniques. In the proposed system, sensors like current, vibration, RPM, and gyroscope data are collected via an embedded device consisting of an ESP32 module. With the help of the collected data, a machine learning model was created that is used to detect various motor operational states, such as healthy, overloaded, vibration warning, and critical state. With the help of the predicted state, the proposed model detects the motor health score and gives intelligent recommendations for preventing any possible malfunctioning. From the results of the simulation, it is clear that the proposed method is cost-effective and efficient. Further, the future work will focus on implementing the system in actual industry and collection of real-time datasets for cloud-based monitoring.

References

[1] B. A. Tama, M. Vania, S. Lee, and S. Lim, \"Recent Advances in the Application of Deep Learning for Fault Diagnosis of Rotating Machinery Using Vibration Signals,\". https://link.springer.com/article/10.1007/s10462-022-10293-3 [2] S. Ayankoso, A. Dutta, Y. He, F. Gu, A. Ball, and S. K. Pal, \"Performance of Vibration and Current Signals in the Fault Diagnosis of Induction Motors Using Deep Learning and Machine Learning Techniques,\". https://journals.sagepub.com/doi/10.1177/14759217241289874 [3] Advancements in Induction Motor Fault Diagnosis and Condition Monitoring: A Comprehensive Review, Sensors, vol. 25, no. 19, 2025. https://www.mdpi.com/1424-8220/25/19/5942 [4] S. Barik et al., \"Predictive Maintenance of Induction Motors Using Machine Learning: A Random Forest Based Fault Prediction Approach,\" World Journal of Advanced Research and Reviews, 2026. https://wjarr.com/content/predictive-maintenance-induction-motors-using-machine-learning-random-forest-based-fault

Copyright

Copyright © 2026 Piyush Chinde. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.