UML Diagram Refinement (Focusing on Class-and Use Case Diagrams)

Large and complicated UML models are not useful, because they are difficult to understand. This problem can be solved by using several diagrams of the same system at different levels of abstraction. Unfortunately, UML does not define an explicit set of rules for ensuring that diagrams at different levels of abstraction are consistent. We define such a set of rules, that we call diagram refinement. Diagram refinement is intuitive, and applicable to several kinds of UML diagrams (mostly to structural diagrams but also to use case diagrams), yet it rests on a solid mathematical basis-the theory of graph homomorphisms. We illustrate its usefulness with a series of examples.