Articulated objects like doors, drawers, valves, and tools are pervasive in our everyday unstructured dynamic environments. Articulation models describe the joint nature between the different parts of an articulated object. As most of these objects are passive, a robot has to interact with them to infer all the articulation models to understand the object topology. We present a general algorithm to estimate the inherent articulation models by exploiting the momentum of the articulated system along with the interaction wrench while manipulating the object. We validate our approach with experiments in a simulation environment.