Extending the Knowledge Discovery Metamodel for architecture-driven simulation modernization

Abstract

With the rapid improvement of hardware, the constant evaluation of simulation development methodologies and environments has brought us a new challenge: simulations modernization. This paper presents a simulation knowledge discovery metamodel that extends the Object Management Group’s knowledge discovery metamodel by adding a simulation model package for enabling architecture-driven simulation modernization. While there are some endeavors that propose integration methods, modernization of legacy simulations has not been investigated. Architecture-driven modernization has been introduced as a process of comprehending and transforming existing software assets. It advocates a model-based approach to software modernization in which the knowledge extracted from software assets is captured in models that conform to a metamodel, namely the knowledge discovery metamodel. It specifies an ontology of software assets. Model transformations are then recommended as means of modernization of legacy assets. But diversity in methodologies and approaches to specify simulation modeling assets prohibits the Knowledge Discovery Metamodel from providing adequate meta definitions to capture knowledge in simulations. System Entity Structure has long been used for knowledge representation by the simulation community. It provides formalism for composition, taxonomy and coupling relations. Hence, System Entity Structure is proposed as an intermediate metamodel to capture the meta-constructs and their relations. It is then made available to define particular metamodels for particular simulation modeling methodologies or approaches. The promoted methodology is exercised with samples in order to collect evidence of its applicability. Metamodel segments are presented for knowledge discovery for discrete event systems specification and Modelica simulation programming language. A complete metamodel is then introduced for the real-time distributed simulation domain model of the German Aerospace Center (DLR) Institute of Flight Systems. Finally, this metamodel is used to introduce discovery artifacts. These samples provide valuable indications that the methodology works.