Abstract: The assembly deviation algorithm based on model-based definition (MBD) system is suitable for the assembly scenes of ideal digital models in cyberspace, but not for the assembly scenes of physical space under non-ideal conditions, resulting in a lot of repairs. Aiming at this problem, simulating real assembly scenes, the transmission and accumulation process of assembly deviation is studied based on the non-ideal state of parts, an assembly deviation prediction algorithm is proposed. Also, this algorithm takes into account the topological relationship between mating features. Firstly, for the hole-axis fit and slot-axis fit, the assembly constraints can be transformed into point location constraints. Based on the maximum material principle, the minimum bounding circle of the cylindrical point cloud is constructed to determine the contact state, and a serial assembly deviation algorithm is provided. Moreover, the above two kinds of matches are defined as two basic elements, the corresponding assembly deviation algorithms are defined as two independent primitive algorithms. Secondly, the parallel assembly deviation algorithm is given based on planar match, hole-shaft match and slot-shaft match. And it is defined as a combined primitive algorithm. Finally, an example is verified based on non-ideal simulated data to accurately predict assembly deviation, which provides a theoretical basis for subsequent assembly deviation calculation based on measured data in cyberspace to promote high-precision assembly.