Using HiGraph to define a formal integrated system modeling framework that ensures complete system consistency

The evolution of the design of complex systems leads to increasing complexity and requires the joint analysis and refinement of different views of the same system which generally consist of: (1) A functional view that describes the main features of the system; (2) An implementation view that allocates functions on system constituents; (3) A non-functional view ensuring that properties such as quality of services, real-time constraints... are satisfied by the system; (4) As well as a dysfunctional view that defines the reliability requirements. Despite the complexity of systems, the consistency of views when exploring the solution space must be ensured. For example: (1) A decision on the required availability may induce new functions or involve redundancy of function/constituent; (2) Another difficulty comes from the fact that the functions are being described using different formalisms, therefore the system engineer must always be able to handle all the following aspects: The availability or reliability models that are mainly based on probabilistic models, the functional view that can be expressed using finite state machines or by event models; the quality of the services that can be expressed either by using a probabilistic approach or an approach based on a bounded set... The work described in this paper focuses on the implementation of a unified industrial modeling process using the graphical language of Hi-Graphs, a specific class of hyper graphs, in support to SysML. This process brings in addition functional views, taking into account, at all stages of the life cycle, non-functional and dysfunctional views of the system in order to make the right choices/compromises in terms of both software engineering and formal verification. It provides end-to-end assurance that the system meets the requirements and contracts associated with service quality during the process of exploring and refining the solution among the different views of the system. It also offers multiple semantics so that existing modeling languages and tools are taken into account.