Impeller geometry optimization using a machine learning-based algorithm with physics embedded dynamic sampling method for a high-speed miniature pump

This paper proposes an optimization approach leveraging machine learning, integrated with physics-informed dynamic sampling, to improve the hydraulic efficiency of highspeed pumps utilized in aerospace. These pumps are difficult to optimize due to their sensitivity to various impeller geometrical pa...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of mechanical science and technology Jg. 39; H. 7; S. 4043 - 4065
Hauptverfasser: Song, Xueyi, Zheng, Kexin, Luo, Xianwu
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Seoul Korean Society of Mechanical Engineers 01.07.2025
Springer Nature B.V
대한기계학회
Schlagworte:
Algorithms
Control
Dynamical Systems
Efficiency
Engineering
Flow separation
High speed
Hydraulics
Impellers
Industrial and Production Engineering
Machine learning
Mechanical Engineering
Optimization
Original Article
Parameter sensitivity
Physics
Pumps
Sampling methods
Suction
Uniform flow
Vibration
Vorticity
기계공학
Machine learning-based algorithm
Physics embedded dynamic sampling method
High-speed miniature pump
Optimization
ISSN:1738-494X, 1976-3824
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:This paper proposes an optimization approach leveraging machine learning, integrated with physics-informed dynamic sampling, to improve the hydraulic efficiency of highspeed pumps utilized in aerospace. These pumps are difficult to optimize due to their sensitivity to various impeller geometrical parameters. To address the large design space and reduce computational cost, the study introduces fundamental design physics equations and integrates a distance criterion within the optimization algorithm. The optimization process focuses on 13 key design variables and utilizes CFD simulations to predict hydraulic performance, with the goal of maximizing efficiency while ensuring the pump head within specified limits. The results show a 4.35 % increase in hydraulic efficiency and improved flow uniformity. Analysis of entropy generation rates and boundary vorticity flux reveals more uniform flow along the blades’ suction side and reduced vorticity near the trailing edge, indicating reduced flow separation and turbulence. This study offers an effective tool for optimizing high-speed miniature pumps, providing insights for future pump designs.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1738-494X
1976-3824
DOI:10.1007/s12206-025-0628-0