Anomaly detection of induction motors using search coils

Thumbnail Image

Files

master_Khan_Dayyan_2025.pdf (9.25 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Electrical Engineering | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-06-11

Department

Major/Subject

Electrical Power and Energy Engineering

Mcode

Degree programme

Master's Programme in Automation and Electrical Engineering

Language

en

Pages

67

Series

Abstract

Induction motors are essential power sources in a wide range of industrial operations, converting electrical energy into mechanical energy to drive systems such as pumps, fans, compressors, and conveyors. As a critical component in manufacturing and processing plants, their continuous and fault-free operation is vital to prevent unplanned downtime, financial losses, and potential safety hazards. Eccentricity faults are one of the most frequently observed faults in industrial electric motors. These faults can often result in severe performance degradation and are particularly challenging to detect in their early stages. Conventional analytical techniques have traditionally been used to detect anomalies in induction motors. For example, signal analysis methods based on the Fast Fourier Transform (FFT) are widely used for analysing voltage and current waveforms. These techniques are computationally efficient and well-suited for deployment in embedded systems. In contrast, machine learning (ML) methods offer the potential to automate fault detection by learning patterns from historical data, enabling more robust and scalable diagnostic systems. In this thesis, a systematic comparative study was conducted on analytical and ML-based methods for detecting dynamic eccentricity faults using search coil voltage signals—an easily applied, non-invasive, and flux-sensitive diagnostic technique. The analytical method, based on frequency-domain analysis, was chosen for its simplicity and low memory footprint, making it suitable for product-level implementation. In contrast, machine learning approaches such as Principal Component Analysis (PCA) are capable of detecting subtle variations, but they require greater computational power and careful model calibration. A significant outcome of this research is the formulation of a threshold-driven framework for quantifying the severity of faults, which not only identifies the presence of dynamic eccentricity faults but also quantifies their severity based on deviations from a healthy condition. This enables prioritized maintenance decisions and enhances the practical value of condition monitoring systems.

Description

Supervisor

Belahcen, Anouar

Keywords

anomaly detection, induction motors, search coil voltage, eccentricity faults, FFT, machine learning

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202508196322

Collections

[dipl] Sähkötekniikan korkeakoulu / ELEC