Applications of machine learning to the analysis of engine in-cylinder flow and thermal process: A review and outlook. (5th February 2023)

• Record Type: Journal Article
• Authors: Zhao, Fengnian; Hung, David L.S.
• Is Part Of: Applied thermal engineering. Volume 220(2022)
• Publication Date: 2023-02-05
• Subjects: Internal combustion engine -- Engine flow features -- In-cylinder thermal processes -- Data-driven prediction -- Deep learning -- Physics-informed machine learning; Heat engineering -- Periodicals; Heating -- Equipment and supplies -- Periodicals; Periodicals; 621.40205
• DOI: 10.1016/j.applthermaleng.2022.119633
• Languages: English
• ISSNs: 1359-4311

Abstract

Highlights: Machine learning (ML) research on engine in-cylinder phenomena increases rapidly. ML is well suited for engine in-cylinder flow feature extraction and prediction. Image-based and deep ML emerge as effective approaches in recent engine studies. Future approach for ML-engine studies leans towards physics-informed ML methods. Abstract: To adequately elucidate the complex in-cylinder flow structures and its underlying effects on the thermal processes inside an internal combustion engine (ICE) has long been a daunting task since the flow behavior is primarily non-linear and transient. In recent years, the research related to engine in-cylinder phenomena is rapidly advancing, driven by the unprecedented volumes of engine data as well as the applications of data-driven machine learning (ML). Therefore, this paper contributes a timely review to this field by highlighting conventional methods of in-cylinder engine studies, summarizing existing ML applications with their strengths and limitations, and identifying future directions. First, traditional analysis approaches including laser diagnostics measurement and computational fluid dynamics (CFD) modeling are discussed briefly with their limitations, followed by an overview of promising ML methods to address engine in-cylinder research challenges. Then, this paper provides a detailed introduction of development and limitations of ML-based engine studies, which cover the areas of in-cylinder air flow, mixing and combustion. Finally, this review article highlights recent advances of deep learning and physics-informed ML applications in engine in-cylinder studies, and provides recommendations for future directions in this field.