From flow physics to machine learning: A comprehensive review and future perspectives on axial flow and centrifugal compressors

Axial flow and centrifugal compressors (AF & CCs) have ushered in a technological revolution in the fields of thrust generation, power storage, and aeronautical engineering, steering them towards environmental sustainability, low-carbon practices, energy efficiency, and reduced emissions. This p...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Energy (Oxford) Jg. 338; S. 138921
Hauptverfasser: Li, Quan, An, Guangyao, Ooi, Kim Tiow, Zhang, Qian, Lang, Jinhua, Yuan, Wei, Yan, Junhao, Zhang, Lei
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 30.11.2025
Schlagworte:
(Un-) supervised algorithm
Artificial intelligence
Axial flow & centrifugal compressors
Machine learning
Uncertainty quantification
Uncertainty quantification
Artificial intelligence
(Un-) supervised algorithm
Axial flow & centrifugal compressors
Machine learning
ISSN:0360-5442
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Axial flow and centrifugal compressors (AF & CCs) have ushered in a technological revolution in the fields of thrust generation, power storage, and aeronautical engineering, steering them towards environmental sustainability, low-carbon practices, energy efficiency, and reduced emissions. This paper reviews the flow control mechanisms and aerodynamic optimization strategies in AF & CCs, and focuses on the potential application of machine learning (ML) in compressors design and flow modeling. By combining the integration paths of traditional physics modeling methods and modern data-driven technologies, the impact of key control methods (e.g., vortex generator, boundary layer control, and tandem impeller configuration, etc.) on the enhancement of AF & CCs stability and performance is systematically summarized. Meanwhile, it deeply analyzes the application results and future trends of ML in performance prediction, flow field reconstruction and uncertainty quantification. It has been found that the hybrid modeling method, which integrates physics knowledge and data-driven intelligence, is gradually changing the compressor design paradigm, but challenges such as high sample dependence and poor consistency with physical laws still exist. It not only summarizes the current state, but also proposes future research directions in multi-scale modeling, low-data scenario learning, and interpretable modeling to provide theoretical support and research references for efficient, intelligent, and robust compressors design. •Summarizing recent advancements in ML for fluid dynamics, covering theories, methodologies and applications comprehensively.•Reviewing traditional approaches to improving AF & CCs performance while emphasizing the efficiency gains achievable through ML.•Providing an integrated perspective to encourage further exploration of ML in engineering applications and frontier research.•Examining ML trends, its potential for cross-domain applications and its connections to interdisciplinary fields such as neuroscience.
ISSN:0360-5442
DOI:10.1016/j.energy.2025.138921