Load-Related and Eccentricity Fault Diagnosis in Induction Motors Using Decision Tree and Genetic Algorithms

This paper addresses the classification of common load-related mechanical problems in induction motors: imbalance, misalignment, and eccentricity using stray magnetic flux signals, as a single processing magnitude. The proposed methodology employs a decision tree optimized through a genetic algorithm, enhancing diagnostic accuracy while mitigating overfitting. This approach enables the analysis of fault characteristic frequencies based on rotor rotational frequencies, thereby improving fault detection precision. Experimental results on a 0.75 kW induction motor demonstrate a satisfactory fault differentiation, achieving up to 100% accuracy under varying load conditions. This method not only detects the presence of faults but also identifies their specific type, facilitating targeted maintenance and reducing downtime in industrial applications.