Multifidelity Data Fusion: Application to Blended-Wing-Body Multidisciplinary Analysis Under Uncertainty

Abstract

This paper presents a methodology to support improved decision making and identification of risks in early stage multidisciplinary design. The approach fuses multifidelity data from disparate information sources available to designers, such as disciplinary simulations, experiments, and operational data from previously deployed systems. The multifidelity data fusion is achieved using a fidelity-weighted combination of Gaussian process surrogate models. This weighting takes into account both the quality of the Gaussian process approximation and the designer’s confidence in the underlying information source being approximated. The resulting multifidelity surrogate provides a rapid analysis capability, which in turn enables the high number of evaluations needed to conduct multidisciplinary trade studies and to propagate uncertainty to the overall system level. The methodology is broadly applicable across engineering design problems where multiple information sources are available to support design decisions. In this paper, the approach is demonstrated on the stability and control analysis of a Blended-Wing-Body aircraft’s center-of-gravity limits, considering the longitudinal criteria of fly-to-stall, stall recovery, nose-wheel steering, and nose-wheel liftoff. The results show that low-fidelity models are enhanced by the presence of higher-fidelity data in key areas of the design space. The presence of even sparse high-fidelity data is key to reducing the variance in the overall analysis, thereby improving the quality of the predictions needed to inform early stage design decisions.