Improving aircraft performance using machine learning: a review

TL;DR

The paper surveys the state of machine learning across aerospace disciplines (fluid dynamics, aerodynamics, aeroacoustics, combustion, and structural health monitoring) to understand how data-driven methods can improve aircraft performance and sustainability. It synthesizes methodologies (neural networks, regression, clustering, ROMs, and physics-informed approaches) and surveys their applications, benefits, and challenges, including data fusion, interpretability, and uncertainty quantification. Key contributions include mapping ML techniques to aerospace problems, outlining successful data-driven ROMs and surrogate models, and identifying opportunities for integrating ML with physics-based models to accelerate design, optimization, and real-time decision making. The work emphasizes that ML should augment CFD and experiments—enhancing efficiency and insight while maintaining physical fidelity and certifiability—driving toward more sustainable, safer, and cost-effective aerospace systems.

Abstract

This review covers the new developments in machine learning (ML) that are impacting the multi-disciplinary area of aerospace engineering, including fundamental fluid dynamics (experimental and numerical), aerodynamics, acoustics, combustion and structural health monitoring. We review the state of the art, gathering the advantages and challenges of ML methods across different aerospace disciplines and provide our view on future opportunities. The basic concepts and the most relevant strategies for ML are presented together with the most relevant applications in aerospace engineering, revealing that ML is improving aircraft performance and that these techniques will have a large impact in the near future.

Figures (8)

• Figure 1: Towards sustainable aviation using machine learning.
• Figure 2: Machine learning methods: a general overview. Classification extracted from Ref. BruntonetalAARR2019. In bold, the most popular techniques in the field of aerospace engineering.
• Figure 3: Sketch representing a NN architecture with three layers. Structure extracted from Ref. LeClaincheetalJFM2022.
• Figure 4: Summary of methods for regression and classification in machine learning. From left to right: linear regression, logistic regression, K-nearest neighbours and support vector machines.
• Figure 5: Schematic representation of a GANS architecture used to increase the resolution of the quantities measured at the wall in a turbulent channel flow. The color coding for each layer is 2D convolution (beige), parametric-ReLU activation (dark green), batch normalization (blue), sub-pix convolution (pink) and ReLU activation (light green). Note that the kernel size and the number of filters are shown at the bottom of each of the layers. Figure adapted from Ref. guemes2 with permission of the publisher.