The locating datum system is an important content to uniform positioning throughout the design, manufacture, assembly, and inspection in dimensional management of sheet metal product. This paper develops a more accurate variation propagation model in multi-station assembly processes using a discrete time nonlinear state-space model to provide a mathematical representation for process-oriented locating datum system design. The proposed model which includes detail design parameters of locating datum system establishes the quantitative relation between Key Control Characteristics (KCCs) and Key Product Characteristics (KPCs) using the motion vector of the component-oriented reference system as a state variable. In addition, the nonlinear accumulated deviation of locating datum (holes and slots) on parts caused by the fixture deviation is considered in the process of modeling for the first time. And the effectiveness of the model is verified through comparison with widely used software, Variation System Analysis (VSA) in a four-station assembly process of a body side inner panel.