A Complexity Measure for System Safety Assurance

Abstract

AbstractThis paper describes a two‐year research effort to define complexity measures for avionics systems in order to help the FAA identify when systems are too complex to be able to assure their safety. The research project developed a measure of complexity related to the number of ways that an avionics system error (fault) could propagate from one element to another. Since each potential propagation requires another sub‐argument in the safety case, the number of such arguments should be proportional to certification effort. Thus, the ability to show that the system is safe, through the certification process, should depend on whether a system has low enough complexity (number of ways for errors to propagate).Our results include a formula for estimating the “error‐propagation complexity” from a system design, the results of using that formula on small and medium systems, and steps for using the formula. The test of the formula was performed by repeating that calculation on a second design for the first system and briefly on a larger design identified by the FAA, from a NASA technical report.