Physics-informed neural networks in clean combustion: a pathway to sustainable aerospace propulsion

Abstract

Achieving clean combustion systems is crucial for addressing environmental impacts, decarbonization needs, and sustainability challenges. Traditional combustion modeling techniques via computational fluid dynamics with accurate chemical kinetics face obstacles in computational cost and accurate representation of turbulence-chemistry interactions. Physics-Informed Neural Networks (PINNs), as a novel framework merging physical laws with data-driven learning, demonstrate great potential as an alternative methodology. By directly integrating conservation equations into their training process, PINNs achieve accurate mesh-free modeling of complex combustion phenomena despite having limited datasets.
This review examines state-of-the-art PINNs applications in clean combustion systems, focusing on their impact in aerospace propulsion. We systematically analyze implementations across flame dynamics and propagation (achieving computational speedups of 2.3-4.9 times), turbulent combustion modeling including thermoacoustic instabilities, emissions prediction for NOx, soot, and CO, stiff reaction systems (with speedups of 6.0-14.6 times for chemical source terms), and optimization and control strategies. The review provides detailed comparisons with traditional CFD methods, highlighting PINNs advantages in computational efficiency, mesh-free operation, and native inverse problem capability, while acknowledging challenges in training stability, uncertainty quantification, and industrial validation.
We present a research roadmap spanning short-term priorities (2025–2027) for algorithm development and uncertainty quantification, medium-term goals (2027–2030) for industrial deployment and multi-physics integration, and long-term vision (2030+) encompassing quantum-enhanced PINNs and self-learning systems. Cross-cutting themes include evolution toward physics-discovering frameworks, integrated experimental-computational workflows, and transferable knowledge across scales. Critical analysis reveals that while PINNs have progressed rapidly from fundamental demonstrations to industrial applications within four years, significant challenges remain in real-time control, safety certification, and industrial deployment.
Next-generation aerospace engines rely on PINNs to reduce computational costs while increasing predictive performance and enabling real-time control methods. This review demonstrates how PINNs can revolutionize sustainable and efficient combustion technologies in aerospace propulsion systems, contributing to climate change mitigation while maintaining performance requirements of modern propulsion systems.

More information

Identifiers

ORCID for Nader Karimi: orcid.org/0000-0002-4559-6245

Catalogue record

Export record

Altmetrics