Meta-modeling concepts and tools for model management: a systems approach

We present a new framework for model management based on system concepts and theory. Underlying the framework is a set of meta-modeling concepts that are useful in capturing the semantics of the modeling process in a modeling environment. These concepts include the notions of a general-model type, type specialization, atomic and composite model versions, model instances, and parameterized versions. We describe these concepts both conceptually and formally and then briefly present a Model Description Language (MDL) that embodies them. While other researchers have suggested some of these concepts primarily in different contexts, this paper makes at least four valuable contributions: (1) the identification of fundamental issues and principles related to model management; (2) the development and enhanced treatment of meta-modeling concepts specifically for model management; (3) the synthesis of those concepts into a coherent, unifying framework for model management; and (4) a demonstration of the practicality of those concepts through a prototype system implementation. Our framework proposes a graph-oriented, nonprocedural, and hierarchical approach for model composition. The framework also supports both model-solver independence and model-data independence. Moreover, it offers general solutions to two critical issues in model management: model-model linkage and model-data linkage. We argue that the system framework can serve as a guide for an effective design of a flexible and extensible model management system. An architecture of such a system and its prototype implementation---called SYMMS---are briefly described. Examples are presented to illustrate the features and advantages of our approach.