In this paper we suggest a framework for integrating multiple model types in model based diagnosis. The framework is based on two concepts. Firstly, a model typology where model types are classified and characterized in order to guide the diagnostic reasoning in focusing on relevant models. Secondly, a model ontology which is represented in a semantic network. This network constitutes the communication medium between the different model types, and facilitates switching between them. We illustrate how the framework enforces the robustness as well as the performance of the diagnostic system.