A framework for probabilistic model-based engineering and data synthesis

Modern computing resources provide scientists, engineers, and system design teams the ability to study phenomena, such as system behavior, in a virtual setting. Computational modeling and simulation (M&S) enables engineers to avoid many of the challenges encountered in traditional design engineering, including the design, manufacture, and testing of expensive prototypes prior to having an optimized design. However, the use of M&S carries its own challenges, such as the computational time and resources required to execute effective studies, and uncertainties arising from simplifying assumptions inherent to computer models, which are intended to be an approximate representation of reality. In recent year advances have been made in a number of areas related to the efficient and reliable use of M&S for system evaluations, including design & analysis of computer experiments, uncertainty quantification, probabilistic analysis, response optimization, and data synthesis techniques. In this review paper, a general framework for systematically executing efficient M&S studies at the component-level, product-level, system-level, and system-of-systems-level is described. A case study is used to demonstrate how statistical and probabilistic techniques can be integrated with M&S to address those challenges inherent to model-based engineering, and how this aligns with the proposed workflow. The example is a gun-launch dynamics model of an artillery projectile developed by US Army engineers, and illustrates the application of this workflow in the study of subsystem system reliability, performance, and end-to-end system-level characterization.