Abstract: A systematical research on the steam turbine pump in naval lubricating oil systems is conducted. Particularly，the structure and functional relationship between different components of the system，as well as its typical failure modes are investigated. In order to test the reliability of the system，a Bayesian Network is established based on the FTA model of three typical failure modes. Then，the conditional probability table is corrected by considering the inherent uncertainties within the system. This modification improves the accuracy of the quantitative reliability analysis. Also，the reliability of the system and the importance of root nodes are derived with the precise inference algorithm. The results reveal several critical components that need regular backups and focus monitoring. In brief，the proposed method can be employed to improve the general reliability of the system.