Improving aircraft performance using machine learning a review Soledad Le Clainche1 Esteban Ferrer1 Sam Gibson2 Elisabeth Cross2 Alessandro Parente3 Ricardo Vinuesa4

2025-05-08 0 0 1.92MB 56 页 10玖币

侵权投诉

Improving aircraft performance using machine learning: a review

Soledad Le Clainche1, Esteban Ferrer1, Sam Gibson2,

Elisabeth Cross2, Alessandro Parente3, Ricardo Vinuesa4

1Universidad Polit´

ecnica de Madrid, Spain

2University of Shefﬁeld, United Kingdom

3Universit´

e Libre de Bruxelles, Belgium

4KTH Royal Institute of Technology, Sweden

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.

Contents

1 Introduction 2

2 Machine learning methodology: a general overview 5

2.1 Neuralnetworks ...................................... 5

2.2 Regression and Classiﬁcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 Semi-supervisedlearning ................................. 9

2.4 Unsupervised learning: clustering and dimensionality reduction . . . . . . . . . . 11

2.4.1 Clustering...................................... 11

2.4.2 Dimensionality reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4.3 Principal Component Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4.4 Local dimensionality reduction . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.4.5 KernelPCA ..................................... 14

2.4.6 Autoencoders.................................... 14

3 Fluid mechanics 14

3.1 Computational ﬂuid dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2 Reduced-ordermodels................................... 16

3.3 Experiments......................................... 18

4 Aerodynamics 19

4.1 Aerodynamic coefﬁcients estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

4.2 Aeroelasticity ........................................ 21

4.3 Designoptimization .................................... 22

∗Corresponding author: XX

arXiv:2210.11481v1 [cs.LG] 20 Oct 2022

5 Aeroacoustics 23

6 Combustion 25

6.1 Data analysis and feature extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

6.2 Dimensionality reduction, classiﬁcation and adaptive chemistry . . . . . . . . . . . 26

6.3 Combustionclosures .................................... 27

6.4 Reduced-order models for realistic combustion systems . . . . . . . . . . . . . . . . 28

7 Structural assessment 29

8 Conclusions 31

1 Introduction

Climate change and increasing resource scarcity are challenges that Europe needs to face in the

coming decades. All this has a direct impact on air transport, which is struggling to maintain its

performance and competitiveness while ensuring a development focused on sustainable mobil-

ity. Research and innovation are essential to maintain the capabilities of the aviation industry,

driven by the rise of new markets and new competitors as a result of globalization. A new long-

term vision for the aeronautics sector is essential to ensure its successful advancement. In this

line, new requirements for the future aviation industry have been deﬁned by the ACARE Flight-

path 2050, a Group of Recognized Personalities in the aeronautic sector, including stakeholders

from the aeronautics industry, air trafﬁc management, airports, airlines, energy providers and

the research community. Aeronautics and air transport comprises both: air vehicle and system

technology. The future of aviation should focus on improving design, reducing manufactur-

ing time and cost (including certiﬁcation and upgrade processes), and also improving the parts

forming the overall air travel system (general aviation, aircraft, airlines, airports, air trafﬁc man-

agement and maintenance, repair and overhaul).

The ACARE Flightpath 2050 has deﬁned 5 goals that should be achieved by 2050 to guarantee

the path through sustainable mobility:

• Compared to the capabilities of a typical aircraft in the year 2000, by 2050 new technolo-

gies should allow 90% reduction in NOx emissions, 75% reduction in CO2 emissions per

passenger/kilometre and 65% reduction in noise emission of ﬂying aircraft.

• When taxiing, aircraft movements should be emission-free.

• Novel strategies to design and manufacture aerial vehicles should be developed to make

them recyclable.

• Sustainable alternative fuels should be developed to position Europe as the centre of excel-

lence in the ﬁeld, and sustained by a strong European energy policy.

• By 2025, Europe should take the lead to establish global environmental standards, formu-

lating and prioritizing an environmental action plan, and being at the forefront of atmo-

spheric research.

At the same time, ensuring safety and security is also a major priority, with the aim at reducing

by 80% the number of accidents by 2050 compared to 2000, taking into account the rising trafﬁc.

To achieve these goals, it is extremely important to ﬁnd newer eco-friendly alternatives suit-

able for the industry to reduce the aviation net carbon emissions and noise. To this aim, the

aerospace industry is gathering efforts towards developing new aerodynamic designs, more ef-

ﬁcient, reducing the oil consumption whilst maintaining the safety in the ﬂight performance.

Moreover, ﬁnding new alternatives to fossil fuels, improving the energy efﬁciency in combus-

tion systems, or ﬁnding optimal routes for air trafﬁc management (ATM) are also some of the key

points where the aerospace industry should advance to minimize the environmental footprint.

However, to achieve these objectives it is not enough to improve the ‘standard’ conﬁgurations.

The aerospace engineering industry is aware that to go beyond the state-of-the-art it is necessary

to develop novel ground-breaking disruptive technologies.

Fluid and solid mechanics need to be advanced with applications in aerodynamics, acoustics,

and combustion, to develop new technologies, resulting in novel aircraft designs with reduced

environmental impact (see Fig. 1). Researchers in collaboration with the aeronautical industry

should explore: (i) new aircraft conﬁgurations able to reduce noise and pollution emissions, (ii)

cruise drag reduction by manipulating turbulent ﬂow structures close to the aircraft surface (i.e.,

delaying the boundary transition from laminar to turbulent ﬂow), using novel friendly low-risk

practices, (iii) novel strategies for ﬂow control (rising the beneﬁts achieved by only changing the

external shape) to enhance the aerodynamic performance reducing drag, noise or ﬂow transition,

rising lift or controlling unsteadiness or ﬂow separation, and (iv) reduce the system complexity

with novel aircraft materials and lighter designs (which directly results in less fuel consump-

tion), and reduced aircraft maintenance and life cost cycle [3].

Figure 1: Towards sustainable aviation using machine learning.

High-ﬁdelity numerical simulations and advanced experimental techniques (i.e., wind tunnel

experiments or open-air experiments as in the case of ﬂight test, et cetera) allow collecting a

large variety of data, containing relevant information about physical principles connected to

the aerodynamic performance of the aircraft, the efﬁciency of the combustion system or the

main instabilities driving the ﬂow dynamics and the possibility of attenuating or boosting such

instabilities using active or passive ﬂow control techniques [2]. Additionally, experiments (i.e.,

ultrasounds, non-intrusive testing, et cetera) and simulations provide information connected

to the fundamentals of solid mechanics, the presence of noise and the structural health of the

aircraft, allowing for noise control and early failure detection.

However, the economic and computational cost, related to the performance of experiments

and simulations, encourages researchers to look for new alternatives, which allow to advance

in the ﬁeld i.e., developing relevant technologies for the aerospace industry, while avoiding de-

lays in the manufacturing and time-to-market process. Aircraft development, manufacturing,

maintenance and support are four critical levels that must be accurate and reliable to ensure the

success of the aerospace industry.

Artiﬁcial intelligence (AI) and machine learning (ML) have been introduced in the aerospace

industry for various applications connected to the reduction of aircraft’s environmental impact,

including data interpretation [174], system management, customer service or aircraft modelling

and to generate new high-ﬁdelity databases at a reduced (economic and CPU) cost [191], solving

problems of optimization, ﬂow control, or even providing optimal sensors distributions for solid

mechanics or aeroelasticity applications. In the recent review article by Brunton et al. [40], the

authors summarize the new trends and perspective of ML in the aerospace industry, including

its application for smart manufacturing, and in the development (and aircraft design), produc-

tion and product support phases (aircraft design, manufacturing, veriﬁcation and validation).

The authors reveals the possibilities of ML to process data in light computations increasing the

production rate, based on the idea of the use of ML techniques that are measurable, interpretable

and certiﬁable.

ML is generally understood as a branch of AI, although there are nuances in the deﬁnition:

ML aims at improving systems performance using self-learning algorithms, while AI tries to

mimic natural intelligence solving complex problems and enabling decision making (although

not maximizing the system efﬁciency) [242]. Both AI and ML are connected to Big Data, a term

that linked to the enormous volume of data that ﬂoods the aeronautical sector every day or that

is generated from CFD simulations or experimental measurements and connected to aircraft

aerodynamic performance [62, 216]. Combining Big Data with ML techniques, it is possible to

develop reduced dimensional systems, such as reduced order models (ROMs) [171, 318], capable

to accurately predict the evolution of the ﬂow dynamics [2, 110, 172] (i.e., ﬂow control, reduce

cruise drag [295], boundary layer transition, etc.), or surrogate models, capable to predict the

aerodynamic forces and moments acting on the aircraft as a function of some parameters (i.e.,

Reynolds number, Mach number, geometry shape, etc.).

This review provides a state-of-the-art of AI and ML applications in the aerospace engineer-

ing ﬁeld. The basic concepts and most relevant strategies for ML and AI are brought together to

explain the similarities found in the nomenclature of similar techniques used in different ﬁelds,

also shedding light on new applications of these algorithms, quite extended in other ﬁelds but

not known to the aerospace industry. For example, the review details the use of machine learning

for reduced order modelling, which can be used to accelerate numerical simulations, or for tem-

poral and spatial forecast (including non-intrusive sensing). Additionally, we include relevant

applications of the ﬁeld, including ﬂow control, acoustics, combustion, ﬂight test and structural

health monitoring.

This article intends to explore the possibilities of ML, an emerging ﬁeld for the aerospace in-

dustry, identifying new research lines of potential interest and bringing new ideas to developing

the technologies of the future, and founded on a primary goal: to ﬁght climate change. This ar-

ticle reviews the main disciplines connected to the aerodynamic performance of the aircraft (see

Fig. 1): fundamental ﬂuid dynamics, aerodynamics, acoustics, combustion, and general solid

mechanics; and based on the idea of ﬁnding novel efﬁcient designs, capable to reduce noise and

pollutant emissions, while at the same time, ensuring safety and security.

The article is organized as follows. Section 2 introduces the machine learning methodologies,

and the literature review of the main applications in aerospace engineering is presented in Sec-

tion 3 for fundamental ﬂuid dynamics, Section 4 for aerodynamics, Section 5 for aeroacoustics,

Section 6 for combustion and Section 7 for solid mechanics. The main conclusions are presented

in Section 8.

2 Machine learning methodology: a general overview

Current advances in computer science are strongly related to the increasing amount of data gen-

erated and stored in the different disciplines conforming aerospace engineering. The valuable

information contained in these databases encourages researchers to develop and test sophisti-

cated algorithms to exploit such information, to gain insight and knowledge from the data and

to subsequently propose and develop new commercial strategies aligned with the ideas behind

the concept of sustainable aviation: developing new cleaner and safer aircraft designs. ML is a

fast-growing science in the ﬁeld of aerospace engineering due to its good capabilities to extract

information from complex databases, which it later used to develop models, such as ROMs or

surrogate models. Based on available data and the type of training carried out within the analy-

sis, ML algorithms can be classiﬁed into unsupervised, semi-supervised or supervised learning,

as presented in Fig. 2. This section brieﬂy introduces the basic idea behind some of these ML al-

gorithms, which have been used for different applications in the ﬁeld of aerospace engineering.

Figure 2: Machine learning methods: a general overview. Classiﬁcation extracted from Ref. [39].

In bold, the most popular techniques in the ﬁeld of aerospace engineering.

2.1 Neural networks

ML uses artiﬁcial neural networks (ANNs), also called as neural networks (NNs), to process

and extract information from databases. The name of this computing system is inspired by the

biological neural networks of the human brain. ML uses NNs to solve an optimization prob-

lem. More speciﬁcally, using back propagation and stochastic gradient descent algorithms, ML

optimizes the following compound function

arg min

(fP(AP,··· , f2(A2, f1(A1,X)) ···) + λg(Aj)),(1)

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Improvingaircraftperformanceusingmachinelearning:areviewSoledadLeClainche1,EstebanFerrer1,SamGibson2,ElisabethCross2,AlessandroParente3,RicardoVinuesa41UniversidadPolit´ecnicadeMadrid,Spain2UniversityofShefeld,UnitedKingdom3Universit´eLibredeBruxelles,Belgium4KTHRoyalInstituteofTechnology,SwedenAbs...

展开>> 收起<<

Improving aircraft performance using machine learning a review Soledad Le Clainche1 Esteban Ferrer1 Sam Gibson2 Elisabeth Cross2 Alessandro Parente3 Ricardo Vinuesa4.pdf

共56页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Improving aircraft performance using machine learning a review Soledad Le Clainche1 Esteban Ferrer1 Sam Gibson2 Elisabeth Cross2 Alessandro Parente3 Ricardo Vinuesa4

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: