Prediction of induction motor faults using machine learning.

Area of Science:

• Industrial Engineering
• Computer Science
• Electrical Engineering

Background:

• Unplanned industrial downtime significantly impacts production efficiency and profitability.
• Predictive maintenance, using data analytics, machine learning, and IoT, offers real-time equipment monitoring to mitigate disruptions.
• Optimizing operations and reducing disruptions are key goals for industries to meet demands and financial targets.

Purpose of the Study:

• To develop a versatile machine learning model for predicting induction motor faults in industrial settings.
• To enable proactive maintenance strategies, thereby decreasing industrial operational downtime.
• To evaluate the performance of various machine learning algorithms for fault prediction in induction motors.

Main Methods:

• Acquired a dataset of healthy and faulty conditions for four 3-phase induction motors.
• Trained multiple machine learning algorithms including Random Forest (RF), Artificial Neural Network (ANN), k-nearest Neighbors (k-NN), and Decision Tree (DT).
• Evaluated model performance using accuracy metrics and confusion matrices for detailed class-specific analysis.

Main Results:

• The Random Forest model achieved the highest prediction accuracy at 0.91.
• Artificial Neural Network and k-nearest Neighbors models demonstrated strong performance with 0.9 accuracy.
• Decision Tree model showed the lowest accuracy at 0.89, with confusion matrices confirming effective classification of motor conditions.

Conclusions:

• Machine learning models show significant promise for predicting induction motor faults, enabling proactive maintenance.
• The Random Forest model is particularly effective for this predictive task, offering high accuracy.
• Future work can enhance performance through model refinement, ensemble methods, and diverse dataset integration for improved generalization.