Multi-objective optimization of high Mach waverider based on small-sample surrogate model DOI
Yue Ma, Anlin Jiang, Mingming Guo

и другие.

Physics of Fluids, Год журнала: 2024, Номер 36(9)

Опубликована: Сен. 1, 2024

Advancements have been achieved in the optimization of waverider designs with aid machine learning to expedite design process. However, these approaches are hampered by need for extensive sample sizes and susceptibility becoming ensnared local optima. This study undertakes a parametric based on wedge-derived, power-law-shaped waverider, increasing configuration diversity creating dataset limited samples calculating geometry aerodynamic parameters. At Mach number 10, multi-objective is implemented using Young's double-slit experiment-least squares support vector regression (YDSE-LSSVR) surrogate model conjunction improved congestion distance particle swarm algorithm, focusing maximizing lift-to-drag ratio volumetric efficiency as much possible. The results indicated that, under conditions samples, YDSE-LSSVR outperforms standard models such regression, LSSVR, Kriging, Polynomial Chaos Expansions-Kriging regarding prediction accuracy. Pareto solutions both concave convex waveriders, obtained through optimization, improve 17.36% 21.70%, respectively, increase 88.89% 105.56%, comparison baseline configurations. In addition, research examines impact various parameters solutions. Finally, applies K-means method conduct cluster analysis solutions, generating three-dimensional configurations distinguished from different clusters.

Язык: Английский

Multi-objective optimization of high Mach waverider based on small-sample surrogate model DOI
Yue Ma, Anlin Jiang, Mingming Guo

и другие.

Physics of Fluids, Год журнала: 2024, Номер 36(9)

Опубликована: Сен. 1, 2024

Advancements have been achieved in the optimization of waverider designs with aid machine learning to expedite design process. However, these approaches are hampered by need for extensive sample sizes and susceptibility becoming ensnared local optima. This study undertakes a parametric based on wedge-derived, power-law-shaped waverider, increasing configuration diversity creating dataset limited samples calculating geometry aerodynamic parameters. At Mach number 10, multi-objective is implemented using Young's double-slit experiment-least squares support vector regression (YDSE-LSSVR) surrogate model conjunction improved congestion distance particle swarm algorithm, focusing maximizing lift-to-drag ratio volumetric efficiency as much possible. The results indicated that, under conditions samples, YDSE-LSSVR outperforms standard models such regression, LSSVR, Kriging, Polynomial Chaos Expansions-Kriging regarding prediction accuracy. Pareto solutions both concave convex waveriders, obtained through optimization, improve 17.36% 21.70%, respectively, increase 88.89% 105.56%, comparison baseline configurations. In addition, research examines impact various parameters solutions. Finally, applies K-means method conduct cluster analysis solutions, generating three-dimensional configurations distinguished from different clusters.

Язык: Английский

Процитировано

1