Internal Fault Detection/Classification System for Permanent Magnet Machines

Due to recent advances in high-performance rare earth magnet materials, permanent magnet (PM) machines are one of the fastest growing electrical machine markets. PM machines provide the excitation field without the significant excitation losses of field-excited machines, nearly eliminate rotor cooling requirements, simplify stator cooling, and reduce the size and weight of electric machines (i.e., generators and motors). Although short circuits in well-designed drive motors are extremely rare, they have the potential to damage the machine due to I2R heating and arcing. This makes reliable and robust short circuit fault detection and management crucial. The work outlined herein proposes to develop algorithms for insulation integrity monitoring and incipient (i.e., developing) fault diagnoses based on an understanding of PM machine fault mechanisms and identification of relevant machine parameters. Trend analysis, statistics, and pattern recognition techniques, implemented via polynomial neural networks, will be used to detect and classify incipient faults. The work will lead to an all-digital inverter controller/insulation integrity monitoring system that is consistent with the Navy Integrated Power System (IPS) vision of commonality and reduced cost. Such a software programmable controller could be hardware standardized for all of the inverter-related functions aboard ships.