PhD Oral Exam - Solihah Sharief Shiekh, Electrical and Computer Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

Wound rotor induction machines (WRIMs) power heavy-duty industrial drives and variable-speed wind turbines, offering smooth startups and controllable acceleration. However, ensuring their reliability and efficiency is challenging, as unexpected faults can raise maintenance costs and reduce availability. Condition-based maintenance and early fault detection are therefore crucial for minimizing downtime. Mechanical faults are responsible for most failures in electrical machines, particularly those caused by eccentricity faults. Detecting and mitigating these faults is therefore essential for reliable WRIM operation in industrial applications. Eccentricity occurs when the rotor deviates from its center due to manufacturing defects, vibrations, or misalignment, creating a non-uniform air gap between the rotor and stator. A common type is axially non-uniform (inclined) eccentricity, where the air-gap length varies both axially and circumferentially. Although more prevalent in practice than uniform eccentricity, it has received less research attention. If undetected, it can cause overheating, insulation degradation, and rotor-stator contact, potentially leading to machine failure. These effects are often more severe in WRIMs than in squirrel-cage induction machines. Accurate modeling of faulty induction machines can significantly improve testing efficiency and reduce costs by eliminating the need for expensive experimental setups. Introducing faults in physical machines is challenging and may cause severe damage, particularly when testing under full-voltage or high-load conditions. In contrast, virtual machines enable safe and flexible emulation of various fault conditions, allowing comprehensive testing without risking physical equipment. Power-hardware-in-the-loop (PHIL) emulation has recently emerged as a cost-effective approach for testing drive systems. This technique enables real-time emulation of electrical machines in laboratory environments while maintaining realistic power exchange between hardware and simulated components. Although PHIL has been extensively investigated for healthy machines and electrical faults, its application to the emulation of mechanical faults remains unexplored.

This PhD research develops both finite element and analytical models to investigate static inclined and static axially uniform eccentricity faults in wound rotor induction machines. The developed models are validated using an actual scaled-down 2.5 MW WRIM prototype. The validated models are then implemented in PHIL-based emulator setups to reproduce fault conditions in real time. Both switched-converter-based and linear-amplifier-based emulator configurations are utilized to accurately replicate the behavior of the faulted machine. Various modeling approaches, including voltage-behind-reactance models, are analyzed and compared to achieve accurate emulation of eccentricity faults. Furthermore, deep learning techniques are employed to estimate the severity of eccentricity faults and to distinguish between axially uniform and axially non-uniform eccentricity. Several data-driven models are developed to improve detection accuracy while reducing dependence on large training datasets. The study is further extended to other eccentricity types, including dynamic and mixed eccentricities. In addition, a comparative analysis of stator and rotor current frequency spectra is conducted to identify effective diagnostic indicators for eccentricity fault detection in WRIMs.