With traditional neutron activation analysis, the increase of sample quality leads to some problems in both irradiation process and measurement process. These problems include the neutron flux gradient in the range of the sample, the decrease of the thermal neutron flux rate around the sample and the influence of self-shielding in the sample in the irradiation process, In the process of measurement, the self-attenuation of γ-ray in the sample and the geometric effect of the sample lead to the effect of the detector on the measurement of characteristic γ-ray emissivity due to the difference in the detection efficiency of each point of the sample. So the neutron activation analysis of mass sample needs to make some additional modifications. By using the neutron activation analysis technique, the content of ²⁴Mg and ²⁸Si in a large amount of flour can be detected, and the content of talc powder in the flour can be given, so that the quality of flour can be monitored. The flour mainly contains C, H, O, N, Ca, and F element, but the main chemical constituent of talc powder is Mg₃[Si₄0₁₀](OH)₂. Therefore, the measured content of Si and Mg element can be used to judge whether the flour contains talcum powder and to determine its exact content. When the content of ²⁴Mg and ²⁸Si in flour are measured by the neutron activation analysis, the variation of neutron flux and energy with thickness in the measured sample and the effect of γ-ray self-absorption will have great influence on the measurement results. The relationship between the neutron flux and energy and the thickness of the sample is simulated by MCNP5 (Monte Carlo N-particle transport code system 5), and the neutron fluxes at different thickness of the sample are measured by a ³He proportional counter tube. The results show that the simulation results of MCNP5 are basically consistent with the experimental results. Using the simulation by MCNP5 and the measurements by a sodium iodide detector, the relationship between γ-ray self-absorption effect and sample thickness is studied, and the sample thickness is determined to be 6.6.cm that is adopted as an optimal experimental condition. Based on the simulated data, the function relationship between the counting of 1.779 MeV γ-ray and the thickness of the sample is obtained as follows: A = 1401 + 3815x – 720x² + 64x³ – 2.8x⁴ + 0.05x⁵. The curve trend of the experimental results is basically the same as that of the simulation results.