Observer-based fault detection and isolation for motor drive inverters

Abstract Observer-based methods can be used for fault detection and isolation for switched linear systems. A state observer is constructed based on the physical system. Changes in the observer error are monitored to detect specific types of faults. This thesis uses the mathematical basis to offer a generic format that can be applied to any linear or switched-linear system. These methods are then applied to a motor drive system, particularly the DC-AC converter, in order to quickly detect parameter changes, sensor faults, and switch faults. Simulations demonstrate that these methods could be used to detect several different fault types. However, the number of distinct fault types that can be distinguished is limited by the number of states of the system