Automating the metamodeling process

Abstract

Model abstraction using metamodeling has demonstrated the capability to facilitate software reuse, large scale model integration, verification, and validation. Once restricted to static representations that represented the input-output behavior of models, research has developed the capability to build dynamic metamodels. This capability results from a new approach supported by a taxonomy of metamodeling problems, solution structures, and metamodeling methods. The development of the metamodel, however, still requires a thorough understanding of model abstraction, reduced order modeling, and system identification. In addition, even with the most robust procedures it is possible that the desired data generated by a simulation model will not meet the assumptions or numerical requirements of the procedure. Consequently, there is a requirement for a robust metamodeling support system that will support the subject matter expert. Automation of the metamodeling process will assist the analyst who is not familiar with model abstraction techniques but needs to reuse a piece of code, integrate different models, or verify a new version of a simulation. This paper describes the design of a Metamodeling Support System that provides this automation.