Abstract: The benchmarking critical experimental data can be used as a reference, providing basic data support for the macroscopic test of nuclear data, the design method verification of new nuclear energy system and the nuclear critical safety analysis, etc. As a crucial link in the process of benchmarking critical experimental data, the uncertainty analysis and evaluation of critical experimental data have been widely concerned and studied at home and abroad. Traditional GUM (Guide to the Uncertainty in Measurement) department based on the uncertainty analysis of the process “one change strategy at a time” uncertainty analysis method in calculation efficiency is low, cannot consider interval probability distribution characteristics of the input parameter error, defects of inaccurate results. In this paper, an uncertainty analysis method for critical experimental data was proposed based on perturbation theory and combination sampling method, and the probability distribution characteristics of input parameters in its error interval were restored by efficient Latin hypercube sampling method, so as to improve the accuracy and rationality of the uncertainty analysis of critical experimental data. By means of conjugate flux perturbation theory, the efficiency of critical experimental data uncertainty analysis was improved without significantly increasing the calculation time. Moreover, the efficient Latin hypercube sampling method and the conjugate flux perturbation method were verified and the feasibility of their application to the uncertainty analysis of critical experimental data was proved. Then, the Latin hypercube sampling program and complex modeling program were compiled, and the compiled program was verified and analyzed on the international critical safety benchmark question, and the calculation results obtained by the traditional critical experimental data uncertainty analysis method and the results given on the international critical safety benchmark question were compared. The conjugate flux perturbation uncertainty analysis method based on MCM (Monte Carlo method) analysis process improves the computational efficiency, makes the analysis results more reasonable and accurate, and proves that the uncertainty program is accurate and feasible for the uncertainty analysis of critical experimental data. Finally, the program was applied to the uncertainty analysis of critical experimental data of neutron toxicological effects in uranium solution obtained by China Institute of Atomic Energy, and a good result was obtained. Compared with the traditional critical experimental data uncertainty analysis method, the standard uncertainty of core characteristic value keff is improved, and the efficiency of uncertainty analysis is greatly improved. The uncertainty analysis method of critical experimental data studied in this paper and the program of uncertainty analysis can provide a more efficient and reasonable analysis method for benchmarking uncertainty analysis and multi-parameter sensitivity and uncertainty analysis of subsequent batches of critical experimental data.