Based on considerations such as structural compactness and safety, the monitoring points set for nuclear main pump mechanical seals are often insufficient to fully describe the operational condition of the seals, especially with regard to the specific status of each sealing stage. This paper utilizes a limited number of sensor points and establishes a probabilistic model to characterize the mechanical seal condition in the form of parameter distributions, which supports the rapid identification of locations with high failure probability while simultaneously displaying the degree of certainty in the results. The method is validated using real operational data from a nuclear power unit over approximately one year (including parameters such as differential pressure, flow rate, etc.). The results indicate that the fault diagnosis conclusions derived from the algorithm are highly consistent with the shutdown maintenance records and measured low-pressure leakage flow rate data. This approach not only effectively overcomes the limitations of insufficient monitoring information, but also provides reliable technical support for the health monitoring and fault tracing of nuclear main pump mechanical seals.

nuclear main pump,mechanical seal,health monitoring,probabilistic model