Abstract: Lateral pressure on the silo wall is an important index for silo design. However, it is difficult to accurately obtain the lateral pressure due to various uncertainties. In this paper, the experimental lateral pressures were collected and then subjected to dimensionless treatment by calculating the ratio of the experimental value to the theoretical calculated value. Based on mathematical statistics method, an optimal processing model for the lateral pressure was developed. In this model, the collected data were sorted into good data and ordinary data, among which the data with large discreteness were removed. A random weighing method was used for processing the optimized small sample data. According to Bayesian statistics, inferential method for modeling the probability characteristics of lateral pressures on silo walls in food silo was established. The validity of this method was verified by the experimental and theoretical data from the No. 29 shallow silo in Henan State Grain Reserve. This method provides theory foundation for the reliability study of food silo and promotes the application of reliability design theory in granary structure.