Document Type : Research Paper

Authors

• V. Samadi ¹
• M. Mostafaei ^{1, 2}
• A. N. Lorestani ¹

¹ Department of Mechanical Engineering of Biosystems, Razi University, Kermanshah, Iran

² Department of Biosystems Engineering, University of Tabriz, Tabriz, Iran

Abstract

This study presents a non-invasive method for detecting mechanical faults in a single-phase AC electromotor using processed acoustic signals. Sound data were collected via a USB-connected microphone installed in the motor's electrical casing under diverse operating conditions. Ten statistical features were extracted from the acoustic signals and used as input to three classification algorithms: Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), and Support Vector Machine (SVM). Model performance was evaluated using confusion matrix metrics, including specificity, accuracy, precision, and sensitivity. Among the classifiers, SVM outperformed others, achieving average values of 99.54, 99.16, 97.15, and 96.17, respectively, with 10-fold cross-validation confirming its superior consistency (99.88% specificity, 99.17% accuracy). The findings confirm that acoustic signal analysis is a reliable and cost-effective tool for real-time fault diagnosis in electromotors. Defects may be accurately found in the electromotor by using acoustic analysis to monitor its status. The proposed framework is adaptable to other rotating machinery through retraining, offering a valuable solution for predictive maintenance in industrial applications.

Graphical Abstract

Keywords

• Acoustic signals
• Classification
• Electromotor
• Fault detection
• Machine learning

Main Subjects

• Damage Mechanics