QPRC 2016

Integration of Multiple Data Sources for System Reliability Assessment


C. Shane Reese

Brigham Young University, Provo, UT


Often in the development of complex defense systems significant procurement and strategic decisions must be made in the presence of sparse system level testing. A poignant example is the production decision to commit large resources to low rate initial production concurrent with the completion of developmental and operational testing. The decision to proceed with resource commitment introduces significant risk into the program and technical management of these type systems. Often these type systems are not designed from scratch, but utilize components and subsystems from previous programs that have extensive usage data in similar or identical environments. We explore Bayesian hierarchical models to combine different types of data and different levels of data into a mathematically coherent model used to evaluate system reliability for systems with many components and subsystems. The model presented utilizes a parametric framework that is estimated using computational methods based in Markov Chain Monte Carlo techniques.  The approach allows combination of expert opinion, previous system data, component and subsystem level testing with a limited amount of system level testing to develop a more comprehensive reliability case early in the system level test phase, but at point when significant program decisions must be made.