Abstract: ⁵⁶Fe is one of important nuclides in reactor shielding design. Quantifying the uncertainty introduced by the nuclear data of ⁵⁶Fe in shielding calculation is significant for determining the sources of uncertainty in reactor shielding design. The nuclear data library used for shielding calculation is generated based on the evaluated nuclear data file (ENDF). The generation of ENDF involves processes such as theoretical calculations, experimental measurements, and data evaluation testing. Inevitably, there is uncertainty in these processes, which leads to uncertainty in the calculation results. In this present work, the uncertainty of a couple of benchmarks caused by the nuclear data of ⁵⁶Fe was analyzed. The uncertainty quantification process was composed of the three steps. First, the NECP-SOUL code was used to generate random samples of ENDF based on the best estimates and covariance from JEFF-3.3 and ENDF/B-Ⅷ.0 evaluated nuclear data files, respectively. After constructing the covariance matrix from the ENDF, an improved Latin hypercube sampling technique was used to generate normally distributed nuclear data samples. Second, the ENDF samples were processed using nuclear data processing code NECP-Atlas to generate the multi-group cross section library. For the energy group structure of the multi-group library, the VITAMIN-B6 energy group structure was selected. Due to the strong resonance of the cross section of ⁵⁶Fe and other nuclides in structural materials, the resonance self-shielding effect makes the multi-group cross-sections dependent on the temperature, geometry, and material composition of the actual problem. Therefore, resonance self-shielding calculations were performed based on the actual problem to generate a problem-dependent multi-group cross section library. Third, the ASPIS-Fe benchmark and ALARM-CF-FE-SHIELD-001 benchmark were calculated using the discrete ordinate based neutron transport code NECP-hydra, and the uncertainty in the detector reaction rate calculation results was quantified based on the results obtained using all the library samples. The results show that the elastic and inelastic scattering cross section has significant impact on the uncertainty of fast neutron detector reaction rates. The uncertainty of the response of ³²S(n,p)³²P detector, which sensitive to neutron over 1 MeV, can reach over 20%. In addition, when both the covariance of cross section and the secondary angular distribution are considered, the uncertainty quantification results increase by approximately 2% compared to the case where only the cross section covariance is considered.