Assessment of Multi-Fidelity Surrogate Approaches for Expedient Loads Prediction in High-Speed Flows

Fast and accurate predictions about the flow field surrounding a hypersonic flight vehicle are necessary for both early design phases and autonomous vehicle control. The complex flow physics of the hypersonic flight regime like aerodynamic heating, real gas effects, and viscous-inviscid interaction can lead to catastrophic failure. Computational fluid dynamics (CFD) is widely used to study these phenomena but necessitates the use of High Performance Computing (HPC) infrastructure. Classical engineering methods provide quick and inexpensive solutions, but have uncertain applicability and accuracy. Data-driven model reduction presents an opportunity to harness the accuracy of high-fidelity techniques for expedient online application. Surrogate modeling is one approach that seeks to emulate the response of a system with evaluation times much faster than data source and without explicitly calling the governing equations online. Multi-fidelity surrogate modeling seeks to reduce the amount of high-fidelity data needed for a sufficiently accurate online model. The objective of this paper is to assesses the trade-offs between a Kriging-based multi-fidelity model and a neural network-based multi-fidelity model, applied to high-speed flow past a canonical flight geometry. All models studied in this work reproduce the results of the benchmark CFD (RANS) with less than 100 training cases. Both the single and multi-fidelity Kriging models marginally outperform the DNN-based models in accuracy, but requirements for training and storing the Kriging models scale exponentially as the training data gets larger. Kriging models offer built-in uncertainty quantification and require significantly less decision making about hyperparameters and architecture choices compared to DNNs. This study also indicates that the inclusion of data from classical engineering methods can improve the prediction accuracy of surrogate models at practically no cost, provided it is correlated with the high-fidelity system response.