AIRFOILS CLASSIFICATION USING PRINCIPAL COMPONENTS ANALYSIS (PCA)

Similar documents
Principal Components Analysis (PCA)

Karhunen-Loève Transform KLT. JanKees van der Poel D.Sc. Student, Mechanical Engineering

Introduction to Atmospheric Flight. Dr. Guven Aerospace Engineer (P.hD)

Multivariate Statistics Fundamentals Part 1: Rotation-based Techniques

Applied Fluid Mechanics

Computational Fluid Dynamics Study Of Fluid Flow And Aerodynamic Forces On An Airfoil S.Kandwal 1, Dr. S. Singh 2

7. Variable extraction and dimensionality reduction

Data Preprocessing Tasks

Lecture-4. Flow Past Immersed Bodies

Introduction to Flight Dynamics

Mechanics of Flight. Warren F. Phillips. John Wiley & Sons, Inc. Professor Mechanical and Aerospace Engineering Utah State University WILEY

ME 425: Aerodynamics

Computational paradigms for the measurement signals processing. Metodologies for the development of classification algorithms.

PCA FACE RECOGNITION

APPLICATIONS The eigenvalues are λ = 5, 5. An orthonormal basis of eigenvectors consists of

1. Fluid Dynamics Around Airfoils

Copyright 2007 N. Komerath. Other rights may be specified with individual items. All rights reserved.

Introduction to Aeronautics

Math 1553, Introduction to Linear Algebra

Definitions. Temperature: Property of the atmosphere (τ). Function of altitude. Pressure: Property of the atmosphere (p). Function of altitude.

Eigenimaging for Facial Recognition

Applied Aerodynamics - I

Eigenface-based facial recognition

Drag Analysis of a Supermarine. Spitfire Mk V at Cruise Conditions

CHAPTER 3 ANALYSIS OF NACA 4 SERIES AIRFOILS

Fundamentals of Airplane Flight Mechanics

Masters in Mechanical Engineering Aerodynamics 1 st Semester 2015/16

Chapter 5 Wing design - selection of wing parameters 2 Lecture 20 Topics

ISSN: (Online) Volume 3, Issue 5, May 2015 International Journal of Advance Research in Computer Science and Management Studies

Lecture 7 Boundary Layer

SPC Aerodynamics Course Assignment Due Date Monday 28 May 2018 at 11:30

AA 242B/ ME 242B: Mechanical Vibrations (Spring 2016)

PRINCIPAL COMPONENT ANALYSIS

THE EFFECT OF WING GEOMETRY ON LIFT AT SUPERSONIC SPEEDS

Engineering Mechanics I. Phongsaen PITAKWATCHARA

Flight Vehicle Terminology

AE Stability and Control of Aerospace Vehicles

INSTITUTE OF AERONAUTICAL ENGINEERING (Autonomous) Dundigal, Hyderabad

Given the water behaves as shown above, which direction will the cylinder rotate?

A tutorial on Principal Components Analysis

Robot Image Credit: Viktoriya Sukhanova 123RF.com. Dimensionality Reduction

AEROSPACE ENGINEERING

Dimensionality Reduction: PCA. Nicholas Ruozzi University of Texas at Dallas

Flight Dynamics and Control

Study of Preliminary Configuration Design of F-35 using simple CFD

PENN STATE UNIVERSITY MATH 220: LINEAR ALGEBRA

Thin airfoil theory. Chapter Compressible potential flow The full potential equation

Experimental Evaluation of Aerodynamics Characteristics of a Baseline Airfoil

GEOG 4110/5100 Advanced Remote Sensing Lecture 15

Econ Slides from Lecture 7

ACD2503 Aircraft Aerodynamics

Performance. 5. More Aerodynamic Considerations

AN ENGINEERING LEVEL PREDICTION METHOD FOR NORMAL-FORCE INCREASE DUE TO WEDGE SECTIONS

ν δ - 1 -

Analysis Of Naca 2412 For Automobile Rear Spoiler Using Composite Material *

Syllabus for AE3610, Aerodynamics I

Linear Algebra Primer

Rotor reference axis

Stability and Control Some Characteristics of Lifting Surfaces, and Pitch-Moments

Example Linear Algebra Competency Test

FLIGHT DYNAMICS. Robert F. Stengel. Princeton University Press Princeton and Oxford

Machine Learning 2nd Edition

Notes on Linear Algebra and Matrix Theory

Airfoil Lift Measurement by Surface Pressure Distribution Lab 2 MAE 424

Mestrado Integrado em Engenharia Mecânica Aerodynamics 1 st Semester 2012/13

ACTIVE SEPARATION CONTROL ON A SLATLESS 2D HIGH-LIFT WING SECTION

COMPUTATIONAL SIMULATION OF THE FLOW PAST AN AIRFOIL FOR AN UNMANNED AERIAL VEHICLE

Why Should You Consider a Freezing Point Depressant Ice Protection System? Icing Certification Present and Future. CAV Aerospace Limited

Principal Component Analysis

Announcements (repeat) Principal Components Analysis

Maximum variance formulation

APPENDIX C DRAG POLAR, STABILITY DERIVATIVES AND CHARACTERISTIC ROOTS OF A JET AIRPLANE (Lectures 37 to 40)

Brenda M. Kulfan, John E. Bussoletti, and Craig L. Hilmes Boeing Commercial Airplane Group, Seattle, Washington, 98124

Math 302 Outcome Statements Winter 2013

Compressible Potential Flow: The Full Potential Equation. Copyright 2009 Narayanan Komerath

Introduction to Aerospace Engineering

Given a stream function for a cylinder in a uniform flow with circulation: a) Sketch the flow pattern in terms of streamlines.

Principal Component Analysis (PCA) Theory, Practice, and Examples

Stability and Control

HOSTOS COMMUNITY COLLEGE DEPARTMENT OF MATHEMATICS

Properties of Linear Transformations from R n to R m

The Mathematics of Facial Recognition

Airfoils and Wings. Eugene M. Cliff

Linear Algebra & Geometry why is linear algebra useful in computer vision?

UNIT-05 VECTORS. 3. Utilize the characteristics of two or more vectors that are concurrent, or collinear, or coplanar.

Multiplying matrices by diagonal matrices is faster than usual matrix multiplication.

Eigenvalues, Eigenvectors, and an Intro to PCA

The E80 Wind Tunnel Experiment the experience will blow you away. by Professor Duron Spring 2012

Aircraft Pitch Control Design Using Observer-State Feedback Control

Dynamics and Control Preliminary Examination Topics

Unsupervised Learning: Dimensionality Reduction

Introduction to Flight

Introduction to Mechanical Engineering

Repeated Eigenvalues and Symmetric Matrices

CS4495/6495 Introduction to Computer Vision. 8B-L2 Principle Component Analysis (and its use in Computer Vision)

MODIFICATION OF AERODYNAMIC WING LOADS BY FLUIDIC DEVICES

Model Rocketry. The Science Behind the Fun

Experimental Study on Flow Control Characteristics of Synthetic Jets over a Blended Wing Body Configuration

ME 6139: High Speed Aerodynamics

Outline Week 1 PCA Challenge. Introduction. Multivariate Statistical Analysis. Hung Chen

Transcription:

AIRFOILS CLASSIFICATION USING PRINCIPAL COMPONENTS ANALYSIS (PCA) Camila Becker, camila_becker_87@hotmail.com Post Graduation in Industrial Systems and Processes, University of Santa Cruz do Sul 96815-900, Santa Cruz do Sul, RS, Brazil Rubén Edgardo Panta Pazos, rpazos@unisc.br and rpp@impa.br Department of Mathematics and Post Graduation in Industrial Systems and Processes, University of Santa Cruz do Sul 96815-900, Santa Cruz do Sul, RS, Brazil Abstract. The importance of air transport has grown considerably in recent decades. Therefore, much research on the aircraft is made. Research on airfoils, or aerodynamical profiles, is an example of a focus of study. Basically, the airfoils consist of a two-dimensional section used in order to fly with changes of velocities of a flux around the airfoil. In aircrafts, the airfoils are present in the wings and empennage, being the former generally asymmetric airfoils (generating lifting and greater moment, so the drag is lower), and for the second, as symmetrical airfoils. The objective of this study is to classify airfoils using principal components analysis (PCA), which is a statistical technique that aims to find patterns to represent changes in many variables, using a smaller number of factors. Its operation is to build a new system of principal components for the representation of the samples, so less dimensions can be considered. Thus, this method presents a lower computational complexity and also the benchmark to obtain results is reduced. The methodology to be developed is as follows: Initially, the data are digitalized. In the next stage, pre-processing is carried out. Subsequently, the correlation matrix is estimated. In the fourth stage, the eigenvalues and eigenvectors of correlation matrix are determined, so that the eigenvectors are indexed by increasing order of eigenvalues. Then, the eigenvalues and the more representative associated eigenvectors are chosen, in order to form the characteristic vector. Subsequently, the sample is projected into a new sub-vector space. Finally, the image is classified with the database formed. The results were favorable, in order to classify airfoils using principal components analysis; this was achieved with a computer algebraic system. Keywords: Airfoils, Principal Components Analysis, computer algebraic system, correlation matrix. 1. INTRODUCTION Currently, much research about the aircraft components is accomplished. This is in reason of the importance of air transport. One focus of the study are the airfoils because of the remarkable role in the study of aerodynamics not only aircraft but also in cars. For the airplanes the airfoils are employed as sections of the wing. For racing cars, the airfoils have great importance because it allows greater stability of the vehicle in order to provide greater adherence on rear wheels. In this work, then, are objectively classified the aerodynamic profiles, or airfoils, using the Principal Component Analysis (PCA) and considering some key components. This paper is organized as follows. In the following section, some considerations about airfoils are presented. In section 3, some ideas on the Principal Component Analysis (PCA) are depicted. Then, there are included some results. Finally, conclusions and possible extensions of this work are given. 2. AIRFOILS The aerodynamics began to have industrial importance with the advent of airplanes and automobiles, because they need to move with the least possible friction with the air for faster and spend less fuel. The study of the airfoils meant an important great step in the aerodynamics; the airfoils represent a section with the capacity for generate lifting (which allows the aircraft up in the air and remain there during the flight) producing so the lowest drag (aerodynamic force against the movement of an object). Figure 1. The balancing forces on an airplane.

Basically, the airfoils are classified as symmetrical and asymmetrical. Both have advantages: the first exhibit simple construction and have easy adaptation to the purposes of the flight. The second have greater aerodynamic efficiency. Figure 2 shows the components of an airfoil: The frontal point of the airfoil is called the leading edge, while the point farthest from the rear edge is called the trailing edge. The segment connecting these two points is called chord. The top half of the airfoil is defined by a curve called upper camber line. The curve that defines the bottom half is called lower camber line. The curve in the middle between these two curves is called the mean line and refers to the arithmetic mean of the coordinates of both camber lines. The greatest distance between the chord and the mean line is called the curvature. The angle of attack is the angle between the chord and the direction of movement of air on the airfoil. Figure 2. Components of the airfoil (example) In this work, in order to classify the airfoils, it was used the principal component analysis (PCA). For this, three parameters are employed: the aspect ratio (i.e. is the ratio between the length and height of airfoil), the curvature of the nose and curvature of the back. 3. PRINCIPAL COMPONENTS ANALYSIS (PCA) The principal components technique was first described by Karl Pearson (1901). He believed that was the correct solution for some problems of interest in biometrics, although his proposal was a practical method of calculation for two or three variables only. A description of practical computational methods came later, but even then the calculations were daunting for some variables because they had all by hand. Only after computers become widely available is that the principal components technique reached widespread use. (Manly, 2008). The principal components analysis is the transformation of a matrix of data into a smaller number of factors, which have more information as possible, in order to represent these variations. Thus, in order to reduce the dimensionality of the original set of data through by means of mutually orthogonal new variables, called principal components. The principal components analysis is a statistical approach that can be used to analyze inter-relationships between a large number of variables and explain these variables in terms of their inherent common dimensions (factors). The goal is to find a way to condense information from a number of original variables into a smaller set of statistical variables (factors) with a minimum loss of information. (Hair et al, 2005) Thus, the principal components analysis is a way of identifying patterns in data, and expressing the data in such a way as to highlight their similarities and differences. Since patterns in data can be hard to find in data of high dimension, where the luxury of graphical representation is not available, PCA is a powerful tool for data analysis. The other main advantage of PCA is that once you have found these patterns in the data, and you compress the data, i. e., by reducing the number of dimensions, without much loss of information. (Smith, 2002). The steps to be chosen to apply the PCA, Smith (2002), are the following: 3.1. Acquire data Initially it is acquired the data to which it this wanted to apply the principal components analysis (PCA). 3.2. Substract the mean For PCA to work properly, the mean is subtracted from each of the data dimensions. The mean subtracted is the average across each dimension. So, all the x values have x (the mean of the x values of all the data points) subtracted, and all the y values have y subtracted from them. The new data set has a null mean.

3.3. Calculate the covariance matrix The covariance is a measure of the strength of the correlation between two or more random variables, defined by: cov ( X,Y ) = n ( X i X )( Yi Y ) i = 1 ( n 1) A useful way to get all the possible covariance values between all the different dimensions is to calculate them all and put them in a matrix. The definition for the covariance matrix for a set of data with n dimensions is: C n n = ~ ~ ( c c cov( X, X ) i, j, i, j = i j (1) (2) Where n n C is a matrix with n rows and n columns, and X ~ is the new variable. 3.4. Calculate the eigenvectores and eigenvalues of the covariance matrix The eigenvalues and eigenvectors are a special set of scalars and vectors, respectively, associated with a linear system of equations (i.e., a matrix equation). Definition: Let be A an n x n matrix. A non-zero vector x in R n is called a eigenvector of A if A x is a scalar multiple of x, i.e., Ax = λx for some scalar λ, which is called the eigenvalue of A and we say that x is associated eigenvector with λ. (Anton and Rorres, 2006) 3.5. Choosing eigenvalues more representatives and forming a feature vector: In fact, it turns out that the eigenvector with the highest eigenvalue is the principle component of the data set. In our example, the eigenvector with the larges eigenvalue was the one that pointed down the middle of the data. It is the most significant relationship between the data dimensions. In general, once eigenvectors are found from the covariance matrix, the next step is to order them by eigenvalue, highest to lowest. This gives you the components in order of significance. What needs to be done now is you need to form a feature vector ( V ), which is just a fancy name for a matrix of vectors. This is constructed considering the eigenvectors chosen from the list of eigenvectors (for dominant eigenvalues), and forming a matrix with these eigenvectors in the columns. V = vet vet vet vet ) (3) ( 1 2 3 n 3.6. Deriving the new data set Once we have chosen the components (eigenvectors) that we wish to keep in our data and formed a feature vector, we simply take the transpose of the vector and multiply it on the left of the original data set, transposed. FD = V C T D A T (4) Where V C is the matrix with the eigenvectors in the columns transposed so that the eigenvectors are now in the rows, with the most significant eigenvector at the top, and D A is the mean-adjusted data transposed, i.e. the data items are in each column, with each row holding a separate dimension. 3.7. The PCA is represented in a graphical way Finally, the PCA is plotted, allowing for greater understanding of the data, since the samples that have greater similarity are grouped. 3.8. To go back when original data O D t t ( V C D ) + mean data = F, (5) where OD represents the original data.

4. RESULTS There will be presented the results for the classification of thirty kinds of airfoils using the Principal Components Analysis (PCA). For this, there are considered three variables associated of the airfoils: aspect ratio, curvature of the nose and curvature of the upper surface. A remarkable point is that the calculations of the curves were accomplished in an approximate way. For this work they ware used the following models of airfoils: Althaus 93k132, Archer 18sm, Boeing 103, Boeing 707e, Boeing 737b, Clark ys, Drela ag45c03, Drela ag13, Eppler 379, Eppler 857, Fage&collins 1, Fage&collins 3, Goettingen 394, Goettingen 492, John Yost eh1070, John Yost eh2070, Martin Hepperle 23, Martin Hepperle 91, NACA 0006, NACA 0008, NACA 0010, NACA 0012, Raf 27, Raf 34, Selig s1046, Selig s2060, Quabeck hq209, Quabeck hq1511, Wortmann 05188 and Wortmann fx84w097. A table with these chosen parameters is presented as an appendix, showing only some airfoils (Tab.1). Figure 3 shows the original data set of the airfoils. Figure 3. Original data. Initially, it was calculated the mean values for each variable and focused on the data in the associated mean (each value was subtracted from the corresponding average). Later, it is founded the covariance matrix of the data, whose representation is outlined in the figure below, see Fig. 4. Figure 4. Covariance matrix of the data used. The covariance matrix plot shows that the curvature of the nose is dominant, not allowing the finer analysis of the other components. Therefore, the data were calculated in a dimensionless way and again the covariance is represented in Fig. 5:

Figure 5. The dimensionless covariance matrix. In this representation is possible to check the covariance of each principal component, showing that the aspect ratio and curvature of the upper have a significant covariance. The next step is calculate of the eigenvalues and associated eigenvectors of the covariance matrix, and it is chosen the most representative eigenvalues to form the characteristic vector. In this case, there will be used eigenvectors associated with the main eigenvalues of the covariance matrix for the construction of the characteristic vector. Thereafter, it is determined the product between the transposed matrix of eigenvectors and the transposed matrix of the dimensionless adjusted data. The next step is the graphical representation of the principal components analysis, which is presented in Fig. 6, comparing to the original data. Figure 6. Comparison chart of the original data with the results obtained by Principal Components Analysis, where X is a new aspect ratio and Y is a new curvature of the upper surface It can be observed the generation of two groups, in reason of the similarity of the samples. It is checked also that the obtained data after the PCA have been lined up. 5. CONCLUSION The main conclusion of this work is that the statistical methods of classification and data analysis are satisfactory. The employment of PCA allows the airfoil classification with only three parameters. The data are grouped following the similarity between them. This method can be used for diverse applications. 6. ACKNOWLEDGEMENTS The authors thank University of Santa Cruz do Sul (UNISC), specially the Post Graduate Program in Industrial Systems Processes for financial support. Furthermore, the first author is grateful to the Brazilian Commission for Personal Improvement of Higher Education - CAPES for the outstanding support.

7. REFERENCES Anton, H.; Rorres, C., 2006. Elementary Linear Algebra with Applications, Wiley 9 th edition. Hair, J. F. (Et al.), 1998. Multivariate data analysis with readings. Prentice-Hall, New Jersey, 4.ed. Manly, B. F. J., 1994. Multivariate statistical methods: a primer. Chapman and Hall, London, 215p. Smith, L. I., 2002. A tutorial on principal component analysis, www.cs.otago.ac.nz/cosc453/student_tutorials/principal_components.pdf Table 1. Associated Parameters for the Airfoils (excerpt) Airfoil Aspect Ratio Nose Curvature Back Curvature Drela_Ag45c03 13,60040805 27,94827534 0,1484127995 Boeing 707 11,08592143 51,94352895 0,4443226132 Eppler 379 11,44323644 19,28333669 0,3163567058 Naca 12 8,330972891 26,57038122 0,2693973025

8. RESPONSIBILITY NOTICE The authors are the only responsible for the printed material included in this paper.