A model based approach to system of systems risk management

Abstract

This paper discusses the approaches required for risk management of `traditional' (single) Systems and System of Systems (SoS) and identifies key differences between them. When engineering systems, the Risk Management methods applied tend to use qualitative techniques, which provide subjective probabilities and it is argued that, due to the inherent complexity of SoS, more quantitative methods must be adopted. The management of SoS risk must be holistic and should not assume that if risks are managed at the system level then SoS risk will be managed implicitly. A model-based approach is outlined, utilizing a central Bayesian Belief Network (BBN) to represent risks and contributing factors. Supporting models are run using a Monte Carlo approach, thereby generating results, which may be `learnt' by the BBN, reducing the reliance on subjective data.