The involvement of multiple stakeholders in the design of complex engineered system presents many challenges. One of the challenge is that the overlapping stakeholder concerns leads to semantic relationships appearing between models. From an inconsistency management perspective, it is critical to investigate how these relationships appear and what are their types. Based on a decision-theoretic foundation, this paper investigates the types of semantic relationships between multiformalism models. We argue that the semantic relationships can be formally captured for two cases -between model versions, and between models from a given instant. Further, we argue that semantic overlap for these two cases can be described through three relationships: equivalence, refinement and abstraction. These relationships can lead to the three types of inconsistencies: inconsistent constraints, inconsistent predictions and inconsistencies between specification and analysis. The paper presents a set of generic rules which can be used to detect these inconsistencies.
I. INTRODUCTIONComplex Engineered Systems (CES) cannot be designed by an individual but involve multiple stakeholders from product design to manufacturing and supply chain. In Model-Based Systems Engineering (MBSE), the predominant way to manage the CES design complexity is to divide the overall problem into sub-problems and address the multiple stakeholder concerns by studying the system from different viewpoints. In this way, different aspects of the system are captured through models developed at different levels of abstractions and using different formalisms. While a viewpoint provides an advantage of dealing with only a limited set of concerns, overlaps between viewpoints cannot be avoided since partitioning does not guarantee complete separation of concerns. As a consequence of these overlaps, relationships appear between models that support different viewpoints. In current MBSE practice, these relationships are managed on adhoc basis and efforts to capture them formally have been minimal. Inadequate management of such relationships can result in inconsistencies, where information in one model logically contradicts the other. Making decisions on contradictory information can prove to be fatal and/or very costly, e.g., the case with Mars Climate Orbiter [1], where the models had inconsistent units. We argue that investigating the nature of overlaps between heterogeneous models is vital to build a support for capturing the relationships between them formally and explicitly. In addition, modeling these relationships is essential for inconsistency management -i.e., to detect and resolve inconsistencies.