All Categories

Ph.D. Research Proposal of John Tsinetakes on Incipient Soft-Fault Detection in DC/DC Converters through Switching Observers

 

Advisor

Dr. Leonid Hrebien

 

Abstract

 Luenberger observers are used to estimate the states of a linear system. The difference between an estimated state and a measured state can be used for fault detection in a linear system. A switched linear system requires a switched observer for estimating the switched system’s states. For a switched linear system the standard theory developed for a Luenberger observer does not apply and non-linear techniques must be used. In this dissertation, these non-linear techniques are examined for a special case of a switched observer as it applies to a switched DC/DC converter. The switched observer can be used for the detection of faults in DC/DC converters. The goal of the fault detection system is to detect an incipient soft-fault within the DC/DC converter either in the switching semiconductors or the output filter components. Hard semiconductor faults and filter capacitor failures have been widely documented, but these faults usually indicate a failure after a power system has shutdown. The focus of a fault detection system using a switching observer is to detect incipient soft-faults. A digitally controlled power supply using an embedded controller provides a useful platform to implement a switching observer. An experimental power supply will be constructed to demonstrate the fault system with methods to inject faults and the ability to store and retrieve faults. This experimental power supply testbed will be used to determine the design constraints in making the system stand-alone on a single embedded microprocessor. A complete design methodology for the fault system will be developed and the limitations of the system based on the embedded microprocessor’s capabilities will be determined.