Bayesian network (BN) method using in error separation is the hot issue of the current research. BN is a powerful tool for uncertainty reasoning and knowledge representation. When we use it combining with the statistical method, it shows a lot the advantage of data processing. This paper considers the many factors in measuring dynamic errors and of the complicated relationship, and brings the small sample BN into the measuring dynamic error separation. In view of the measurement error, we build the basic dependent relationships among variables and the basic construction among nodes and so on. Finally, we use simulation for experiment and analysis through the study of small sample BN.