The more-electric aircraft (MEA) concept is widely viewed as the next evolutionary step towards enabling the industry goal of developing optimised, fuel efficient aircraft. MEA have an increased dependency on electrical energy for distribution to secondary systems and, in order to service this increased dependence, the electrical power systems (EPS) are more complex with increased voltage distribution levels, power conversion stages and safety critical components compared with their conventional counterparts. These complexities will only increase in future platforms as they further embrace the MEA concept - the migration to increasingly novel, critical and complex EPS will incur several development and integration challenges.This thesis considers the fundamental challenge of maintaining high reliability standards within future aircraft EPS through the development of accurate and discriminative real-time protection systems which will react during fault conditions. Specifically, the thesis researches novel methods that improve real-time aircraft EPS protection and health management systems by 1) accurately diagnosing degraded faults before their progression to critical failure and 2) diagnosing faults that are difficult to detect using only conventional protection methods â in particular, series arc faults are considered.Within future aircraft EPS, the volume of operational data is expected to significantly increase beyond that of the conventional systems; consequently, the thesis focuses on the use of data-driven, machine learning based methods, to enable these extended functionalities of the EPS protection and health management systems. The types of machine learning modelling techniques that were chosen are explained and justified. Conventional protection methods are described, including a discussion on the difficulties in using them to detect both degraded fault modes and arcing conditions. The necessity to detect these types of faults in an accurate and timely manner is also discussed.One of the main contributions of the thesis is the proposal of the EPSmart method that can autonomously diagnose and isolate a multitude of degraded faults within an aircraft representative EPS. These degraded faults include intermittent and incipient conditions, which, in comparison to overcurrent faults, often lack the energy to be detected by conventional means. Early, and accurate, detection of these conditions will improve overall system health management and reliability and ensure safe operation of the aircraft.Further contribution is the design of the IntelArc method that can detect series arc faults within direct current supplied systems. Accurate detection of series arc faults is extremely challenging as, despite their presence being a serious fire hazard, they result in a decrease of load current. Although methods do exist for diagnosis of series arcing, there remain challenges with regards to accurate detection across different system configurations and operating conditions. The thesis shows the potential for IntelArc to provide accurate detection across a variety of configurations and operating conditions.While the thesis only describes the initial development of these novel methods, the significant conclusions are that application testing has shown the potential for them to enhance real-time network protection, fault tolerance and health management of aircraft EPS through detection of degraded fault and arcing conditions.
|Date of Award||25 Apr 2017|
- University Of Strathclyde
|Sponsors||EPSRC (Engineering and Physical Sciences Research Council)|
|Supervisor||Stuart Galloway (Supervisor) & Graeme Burt (Supervisor)|