An on-line diagnostic system for a power generator is developed and implemented in a PC interfaced with the existing VAX-based monitorign /control system. The system is capable of detecting deterioration of bearings, local overheatings, bending of the shaft, and weakening of magnetic field at early stages of their development. Diagnostic rules are formulated in terms of specially defined primary and secondary generator variables and performance models providing reliable failure detection/identification. Diagnostic procedures, utilizing readily available information, detect various trends in the generator data and interpret these trends as changes in the generator performance caused by incipient failures. Results of trend analyses, subjected to statistical validation, facilitate failure prediction and identification thus providing the justification for "service when needed". The diagnostic system is validated using simulated generator data that contains intentionally introduced trends of variables corresponding to particular incipient failures.
展开▼