Abstract: Online model generation method was used to solve the problem of the undiagnosable system caused by incomplete modeling. The method determined whether there was the incomplete behavior online by establishing the extensible model offline and determining the maximum length of observations when diagnosablity was available. The incomplete behavior was determined and classified according to the historical data and observation time of online operation, and the diagnostic model was redefined and the diagnoses was generated. The completeness hypothesis was cancelled, when undefined behavior existed in the system, the diagnosis results could be obtained instead of blindly extending. Therefore, the device dependence was further reduced in the modeling process.

Key words: modelbased diagnosis, discrete event system, incomplete model