Abstract: In order to solve the fuzzy and uncertain problems in calculating the risk priority number in traditional failure modes and effect analysis, avoid the chaos of its rank, a ranking method of failure mode improvement priorities based on TOPSIS theory was developed.Because the attribute weights and values of the occurrence, severity and detection will directly affect the ranking result, the sensitivity analysis method based on the attribute weights under the TOPSIS theory was developed.The sensitivity analysis of the attribute values was performed based on the idea of the upper and lower deviation in the interchangeability theory for traditional FMEA and TOPSIS-based FMEA respectively.Corresponding algorithms were established and the stable intervals of the three attribute values under each decision method were obtained.A case of failure modes and effect analysis of a certain diesel variator gear shift shaft was taken as an example to illustrate the proposed method.Results show that the proposed analysis method can help to determine the improvement priorities of the component failure modes, and enhance the robustness of the ranking results.