Trajectory-Informed Sampling for Efficient Construction of Multi-Fidelity Surrogate Models for Hypersonic Vehicles. Needels, J. T. & Alonso, J. J. In AIAA SCITECH 2024 Forum, 2024. AIAA Paper 2024-1013.
doi  abstract   bibtex   
This paper studies the use of multi-fidelity surrogate modeling for analysis of hypersonic vehicles with a focus on the computation of trajectory sensitivities to inform sampling policy. A multi-fidelity aerodynamic database is generated for a hypersonic glide vehicle and surrogate accuracy is assessed for different numbers of high-fidelity samples. A methodology for computing sensitivities of trajectory quantities of interest to surrogate-modeled aerodynamic parameters is applied to hypersonic glide vehicle dynamics. The necessary adjoint equations are derived to compute sensitivities to aerodynamic parameters, which are then used to inform high-fidelity sampling. Of several strategies assessed, sensitivity-based sampling is found to give the best performance for fixed computational cost. Extensions of this methodology are discussed in the context of future work, including the use of sensitivities to allocate simulations among several disciplines in multidisciplinary design problems.
@inproceedings{needels2024,
	title = {Trajectory-{Informed} {Sampling} for {Efficient} {Construction} of {Multi}-{Fidelity} {Surrogate} {Models} for {Hypersonic} {Vehicles}},
	doi = {10.2514/6.2024-1013},
	abstract = {This paper studies the use of multi-fidelity surrogate modeling for analysis of hypersonic vehicles with a focus on the computation of trajectory sensitivities to inform sampling policy. A multi-fidelity aerodynamic database is generated for a hypersonic glide vehicle and surrogate accuracy is assessed for different numbers of high-fidelity samples. A methodology for computing sensitivities of trajectory quantities of interest to surrogate-modeled aerodynamic parameters is applied to hypersonic glide vehicle dynamics. The necessary adjoint equations are derived to compute sensitivities to aerodynamic parameters, which are then used to inform high-fidelity sampling. Of several strategies assessed, sensitivity-based sampling is found to give the best performance for fixed computational cost. Extensions of this methodology are discussed in the context of future work, including the use of sensitivities to allocate simulations among several disciplines in multidisciplinary design problems.},
	urldate = {2026-09-21},
	booktitle = {{AIAA} {SCITECH} 2024 {Forum}},
	publisher = {AIAA Paper 2024-1013},
	author = {Needels, Jacob T. and Alonso, Juan J.},
	year = {2024},
	doi = {10.2514/6.2024-1013},
}

Downloads: 0