System of systems hazard analysis using simulation and machine learning

In the operation of safety-critical systems, the sequences by which failures can lead to accidents can be many and complex. This is particularly true for the emerging class of systems known as systems of systems, as they are composed of many distributed, heterogenous and autonomous components. Performing hazard analysis on such systems is challenging, in part because it is difficult to know in advance which of the many observable or measurable features of the system are important for maintaining system safety. Hence there is a need for effective techniques to find causal relationships within these systems. This paper explores the use of machine learning techniques to extract potential causal relationships from simulation models. This is illustrated with a case study of a military system of systems.