During the collaborative development of production systems, models are used to represent different views on the systems. These models are heterogeneous in forms but dependent in contents. Inconsistencies among them should be properly handled in an early time. An automated process is favored due to high efficiency and low error proneness when compared to the conventional manual fixing. In this study, an approach for knowledge-based automatically resolving inconsistencies (KARI) is proposed to resolve inconsistencies across development models. The approach can resolve several different inconsistencies simultaneously without causing new inconsistencies. At the same time, changes required to resolve inconsistencies can be minimized. The feasibility of this approach is proved by selected industrial cases in the simulation.