PhD Oral Exam - Ashutosh Patel, 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

Most of the electrified propulsion systems are equipped with traction machines that are driven with voltage source inverters (VSI), which allow efficient control over a wide speed and torque envelope. Despite significant efforts to design the system in the best way possible, the reliability and safety of such systems get compromised due to various faults. Short circuit faults or failures of insulation are prevalent and account for approximately 30% of motor failures. Given the uncertainties associated with the degradation of insulation, monitoring the insulation condition becomes necessary to ensure the safe and reliable operation of the VSI-fed electric machines, especially in applications like EVs and aircraft where the safety of human life is very crucial. Detecting and identifying types of degradation of insulation in an early stage can help prevent major failures such as short circuits in the machine. Such information can also be used for predictive maintenance and fault-preventive control strategies to ensure the safe operation of the machine. Therefore, this PhD research focuses on online monitoring of electrical machine winding insulation degradation.

A comprehensive review of existing literature revealed certain research gaps. The first one is on how to select the most effective insulation indicator for online condition monitoring. In addition, the integration of online condition monitoring should not increase the motor drive cost. To this end, only the signals available in the existing EV motor drives, such as the line current measurements, are considered for the proposed online condition monitoring, to avoid any additional hardware cost. However, there is limited information on how insulation degradation can impact the line currents from the existing literature. There is a need to address this knowledge gap through comprehensive investigations and analysis before developing an appropriate condition monitoring methodology. In response to this challenge, this PhD work first focuses on the development of a high frequency (HF) stator winding model which can be used for both time-domain and frequency domain investigations. This model enables holistic investigations into the influence of various insulation degradations on both time-domain currents and impedance spectrums. Findings from such studies set the stage for improved condition monitoring methodologies.

Existing literature on condition monitoring methods also presents notable limitations. Firstly, while these techniques can accurately determine the degradation or State of Health (SOH) of either groundwall (GW) or turn-to-turn (TT) insulations, they fall short in classifying types of insulation degradation. There is a need for a new approach that can be utilized for simultaneous condition monitoring TT and GW insulations. This is crucial because different types of insulation are exposed to different temperatures, potentially leading to a varied degradation rate. Secondly, it is crucial to address the variability of noise in measured current signals. Present methodologies, often neglect this variability. In real-world applications, such as in EVs, the noise level in measured signals fluctuates due to various factors, leading to the need for a condition monitoring method that can handle noise variability while determining the SOH of insulation. Additionally, the existing approaches rely primarily on predefined thresholds and manual analyses, which necessitates expert knowledge for data interpretation and feature analysis. Such dependency limits their adoption across different machine types and operational conditions. Such complexities associated with different degradation patterns and noise levels lead to a need for a more universal method. Hence, this PhD work focuses on addressing these limitations and proposed multiple novel condition monitoring methodologies. Firstly, this thesis introduces and utilizes a HF stator winding model for time-domain and frequency-domain investigations, crucial for understanding the effects of insulation degradation on line currents and impedance spectrums. Based on such investigations, this thesis proposes the utilization of prominent oscillations in line current for condition monitoring. Thereafter, to address the identified limitations, a novel insulation condition monitoring technique for simultaneous condition monitoring of TT and GW insulation is proposed. To address the limitation posed by noise variability and the reliance on manual analyses in existing methodologies, a novel data-driven methodology for robust insulation condition monitoring has been proposed.