Modular Semantically Integrated System Modelling Swaminathan Natarajan Tata Consultancy Services Research 54B Hadapsar I.E. Pune-411013 India +91 94030 94264 swami.n@tcs.com Subhrojyoti Roy Chaudhuri Tata Consultancy Services Research 54B Hadapsar I.E. Pune-411013 India +91 20 66086411 subhrojyoti.c@tcs.com Anand Kumar Tata Consultancy Services Research 54B Hadapsar I.E. Pune-411013 India +91 20 66086450 anand.ar@tcs.com Copyright © 2019 by Swaminathan Natarajan, Subhrojyoti Roy Chaudhuri and Anand Kumar. Permission granted to INCOSE to publish and use. Abstract. In current practice, we engineer systems using modular, composable components. However, when we model systems in languages such as SysML, we build a single model of the entire deliverable system (such as a radio telescope). Though the model is of course hierarchical, consisting of multiple levels of description, these are not quite independent models that can be composed modularly. For example, if we acquire a power supply off-the-shelf, ideally we would like the sup- plier to provide us a model of the power supply that can be integrated into our overall system model, reflecting the same concerns involved in integrating the power supply into our solutions. If we have distributed teams, can they create separate models that are composable? Can model integration be the early version of system integration? This paper discusses how modelling can support modular composability using roles: placeholders for other entities that the entity being modelled depends on in order to deliver its targeted behaviour. All the necessary assumptions relating to the role are captured as role profiles. Model composition then involves binding these context roles to models of actual concrete entities that fit the assumptions. Our paper presents an illustrative example of compositional modelling of a radio telescope, and discusses some of the issues involved in binding. It goes further to indicate how these system models can be linked to relevant domain knowledge frames, and also linked to observations of the system behaviour. The result is a SystemMap: modular, composable knowledge-centric models that enable capture of all the information and knowledge relating to a system, linked by a network of semantic relationships that enable internal consistency checking, as well as model and knowledge refinement based on consistency with observed real-world behaviours. Current Practice and Motivation Model-based systems engineering (MBSE) is practised widely, nevertheless there are gaps in current MBSE capabilities. These gaps become evident if we take a brief look at MBSE usage in a typical state-of-the-art systems engineering project. Recently we had the opportunity to participate in the design of the control system for the Square Kilometer Array (SKA), an international radio telescope project to design two radio telescopes, to be located in Australia and South Africa, involving over 500 engineers and scientists organized into multiple consortia that addressed various subsystems, including dish antennas, signal processing, production of science data products, networking, timing and synchronization, control and physical infrastructure. The project used SysML for functional analysis and to model the overall system de-