Inequality-based reliability estimates for complex systems

Full-system testing for large-scale systems is often infeasible or very costly. Thus, when estimating system reliability, it is desirable to employ a method that uses subsystem tests, which are often less expensive and more feasible. This paper presents a method for bounding full-system reliabilities based on subsystem tests and, if available, full system tests. The method does not require that subsystems be independent. It accounts for dependencies through use of certain probability inequalities. The inequalities provide the basis for valid reliability calculations while not requiring independent subsystems or full-system tests. The inequalities allow for test information on pairwise subsystem failure modes to be incorporated, thereby improving the estimate of system reliability. We illustrate some of the properties of the estimate via an example application.