adam aerodynamics

Upload: kyanboo

Post on 07-Apr-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/4/2019 Adam Aerodynamics

    1/5

    Aerodynamics of a Morphing Airfoil

    Adam Niksch

    Senior, Aerospace Engineering

    Texas A&M University

    The purpose of this research is to develop a MATLAB model which can calculate the aerodynamics, such

    as lift, drag, and pressure, on an airfoil as it changes shape. This model needs to be computationally

    efficient and should have reasonable accuracy. A constant strength doublet panel code, which takes

    thickness and camber distribution as inputs, is used to model the aerodynamics. This code has the ability

    for thickness and camber to change, which effectively allows the airfoil to morph.

    INTRODUCTION/BACKGROUND

    The purpose of this research is to develop a

    computational model with which to

    calculate the aerodynamics on a morphing

    airfoil. One of the major requirements forthis model is that it be computationally

    efficient. Some error is allowable, but themodel must be reasonably accurate and must

    produce the correct shapes for lift and drag

    plots.

    There are many possible ways to implement

    this model. One way is to use thin airfoil

    theory. The drawbacks to using thin airfoiltheory are the assumptions that must be

    made, namely that the airfoil must be thin.Another possible approach is to useconformal mapping, which is an exact

    method using complex variables. A third

    possibility is to use a panel method.

    However, high order panel methods can bevery computationally inefficient.

    During this research, both conformalmapping and a doublet panel code were

    attempted and each had its advantages and

    disadvantages. The doublet panel code waschosen in the end because it works for bothsymmetric and asymmetric airfoils without

    too much modification required. This panel

    code is able to accurately predict thepressure distribution on any airfoil. The

    panel code did have some problems, but it

    proved to be the best option in the end.

    EXPERIMENTAL SETUP

    A MATLAB code is developed to model the

    aerodynamic changes an airfoil would

    encounter as it changes its shape. This code

    accepts thickness and camber distributionfor the airfoil as inputs, not specified points

    as in some panel codes. This gives the userthe ability to easily change these parameters,

    which causes the airfoil to morph.

    This MATLAB code is a powerful tool in

    that it allows thickness, camber, chord, and

    angle-of-attack all to vary. If a simple for

    loop were to be placed around the functionwith all of these parameters varying, the

    code could simulate morphing the shape ofthe airfoil using 4 degrees of freedom. Thereason this code is so versatile is that it was

    designed to be a subprogram in a larger

    reinforcement learning program, which is abranch of artificial intelligence. With the

    combination of these two programs running,

    an aircraft will learn how to change theshape of its own airfoil for control purposes

    rather than use current control mechanisms.

    One of the assumptions made when writingthis code is that the flow is incompressible.

    This assumption is valid because currentinterests lie in the realm of micro air

    vehicles, which fly at speeds less than Mach

    0.3.

  • 8/4/2019 Adam Aerodynamics

    2/5

    Since the final model uses a panel method to

    calculate the aerodynamics, it is verysensitive to the grid, or location of the

    panels, and the number of panels created.

    The grid must be a sinusoidal spaced grid in

    the x direction, which puts more points atthe trailing edge of the airfoil. This is

    necessary because many aerodynamicchanges occur near the trailing edge. If the

    number of panels used were to decrease, the

    accuracy of the model would also decrease.The code also assumes inviscid flow, and

    thus it is only valid for the linear range of

    airfoils, or angles-of-attack prior to stall.

    RESEARCH PLAN

    Since thin airfoil theory required too many

    assumptions be placed on the airfoil, it was

    almost immediately decided not to use thatapproach. The first attempted model was a

    MATLAB code which used conformal

    mapping to calculate the aerodynamics. Asymmetric airfoil is modeled and only its

    thickness is allowed to change. The model

    produced very accurate results when

    compared to wind tunnel data in Ref. 1. Thebasic equations for conformal mapping are

    listed below in equations 1-3 [3].

    However, once camber was integrated into

    the model, the process became extremely

    complex and required complicatednumerical analysis. Conformal mapping was

    no longer a feasible approach given the

    scope of this research and was abandoned.

    A constant strength doublet panel method

    replaced conformal mapping as themodeling tool. This code required no

    changes to be made if the airfoil is

    asymmetric. Since it is a low order panelmethod, it is computationally efficient. The

    equations for the velocity on each panel are

    listed as equations 4 and 5 [3] where x and zare in the local panel coordinate system.

    Since these equations require panel

    coordinates, a transformation from theglobal coordinate system to the local panel

    coordinate system must be made. This

    transformation is listed as equation 6 [3].

    The panel code is based on the nopenetration condition, which states that the

    flow cannot cross the solid boundary of the

    airfoil, thus the velocity normal to the

    surface is 0 in the global coordinate system.Equation 7 is used to transform the

    velocities from equations 4 and 5 into the

    global coordinate system. [3].

    Now it is possible to solve for the doublet

    strengths using equations 4-7. These doubletstrengths can be used to find the tangential

    velocities at each point. Once the tangential

    velocities are calculated, the pressurecoefficient can be calculated using

    Bernoullis equation which, when modified,

    produces equation 8 [2].

    The pressure coefficient can be broken up

    into normal and axial forces using simpleintegration. These forces can also further be

    broken up into lift and drag using simpletrigonometry. These equations are listed asequations 9-12.

    RESULTS

    The first airfoil to be modeled using the

    constant strength doublet panel method wasa NACA 0012. This airfoil was chosen first

    because of its symmetric shape, which

    makes finding errors within the code easier.There is also proven experimental data for

    the NACA 0012 as can be seen in Ref. 1.

  • 8/4/2019 Adam Aerodynamics

    3/5

    Figure 1: Plot of Cp for a NACA 0012 at zero

    angle of attack

    After accurately calculating the pressure

    distribution on a symmetric airfoil, an

    asymmetric airfoil was chosen next,

    specifically a NACA 4412. As with theNACA 0012, there is proven experimental

    data with which to compare the results of the

    code. This allowed for quick and easyverification of the codes operation.

    Figure 2: Plot of Cp for a NACA 4412 at zero

    angle of attack

    Figures 1 and 2 show that the code

    accurately models both symmetric andasymmetric airfoil shapes. The pressure

    distributions predict no lift on the symmetricairfoil and positive lift on the cambered

    airfoil, which agrees with accepted theory

    and proven experiments on airfoils [1].

    Figure 3: Lift vs. Angle of Attack for a NACA

    0012

    Figure 3 checks model accuracy. The modelwas compared to proven experimental airfoil

    data taken from Ref. 1. Since the code is

    only able to model the linear range of theairfoil, the angles-of-attack tested ranged

    from -5 to 5 degrees. When compared to

    those from Ref. 1, there was some

    inaccuracy, but it was within tolerance forthis problem.

    An .avi file for Windows Media Player was

    also created in order to better understandhow the code works. In this video, the user

    is able to watch as the airfoil starts as aNACA 0006, then morphs into a NACA4418. After the airfoil has successfully

    morphed into a NACA 4418, it morphs back

    to its original NACA 0006 shape. The videoalso shows the airfoil sweeping through a 5

    degree angle-of-attack as it changes its

    shape. The user is also able to observe howthe pressure distribution changes across the

    airfoil as it changes shape and angle-of-attack.

    SUMMARY AND DISCUSSION

    The original objectives of this research wereto model the aerodynamics of an airfoil as it

    changes its shape. The model was required

  • 8/4/2019 Adam Aerodynamics

    4/5

    to be computationally efficient and have a

    reasonable amount of accuracy. Theconstant strength doublet panel method

    accomplishes these objectives. The

    MATLAB code takes a negligible amount of

    time to run using as many as 400 panelsalong the airfoil. The code produces

    reasonably accurate results.

    One unexpected problem which arose

    concerned the trailing edge of the airfoil. Inorder to accurately model the trailing edge,

    the Kutta Condition ought to be enforced by

    inserting a wake panel. The problem was

    discovered when rather high values for thepressure coefficient were seen. It was

    determined that the best way to deal withthis problem was to remove these unrealisticvalues. Once these values were removed, the

    results became much more accurate.

    CONCLUSIONS

    The doublet panel code is able to efficientlymodel the aerodynamic changes on an airfoil

    as it changes shape. It is also more accurate

    than expected for airfoils within the linearrange.

    One of the main things learned from thisresearch is conformal mapping is not a

    realistic option for morphing airfoils. The

    process becomes extremely complex and

    therefore becomes very inefficient.

    Also, if the final two or three values from Cp

    were removed, then the overall accuracy ofthe model increased.

    This code is developed primarily to be usedas a tool for future research. Structural

    effects will be added to the model later using

    basic Euler-Bernoulli beam theory so themodel can also predict how the morphing

    process will affect the physical object. This

    model will also be used as the basis for the

    morphing 3-D finite delta wing in the future.Also, as previously mentioned, a

    reinforcement learning program will use this

    model to learn the optimal airfoil shapes for

    a range of flight conditions.

    ACKNOWLEDGEMENTS

    The author wishes to thank Dr. John Valasek

    and Amanda Lampton of the Aerospace

    Engineering Department at Texas A&MUniversity for their guidance and assistance

    throughout this research. The author also

    acknowledges the Texas A&M FlightSimulation Laboratory for all of their help

    and assistance.

    This Research Experience for

    Undergraduates Site is sponsored by theNational Science Foundation Grant No.

    0453578, the Air Force Office of Scientific

    Research, U.S. Air Force, Department of

    Defense and NASA Cooperative AgreementNo. NCC1-02038.

    REFERENCES

    [1] Abbot, Ira and Von Doenhoff, Albert.

    Theory of Wing Sections. Dover

    Publications, 1959.

    [2] Anderson, John D. Fundamentals ofAerodynamics Third Edition. McGraw-Hill,

    2001.

    [3] Katz, Joseph, and Plotkin, Allen. Low-

    Speed Aerodynamics Second Edition.Cambridge University Press, 2001.

  • 8/4/2019 Adam Aerodynamics

    5/5

    APPENDIX A - LIST OF EQUATIONS

    1

    z x i y

    A B i C

    A zz

    = +

    = +

    = +

    2 2 2 2

    1 2

    1 2

    2 2 2 2

    1 2

    2 ( ) ( )

    2 ( ) ( )

    p

    p

    z zu

    x x z x x z

    x x x xw

    x x z x x z

    =

    + +

    =

    + +

    0

    0

    cos( ) sin( )

    sin( ) cos( )

    i i

    i ip

    x xx

    z z z

    =

    cos( ) sin( )

    sin( ) cos( )

    pi i

    i i p

    uu

    w w

    =

    2 2

    21p u wC

    V

    +=

    ( )0

    0

    1

    1

    c o s ( ) s in ( )

    s in ( ) c o s ( )

    lo w e r u p p e r

    u p p e r lo w er

    c

    n p p

    cu p p e r l o w e r

    a p p

    l n a

    d n a

    C C C d xc

    d y d yC C C d x

    c d x d x

    C C C

    C C C

    =

    =

    =

    = +

    (1)

    (2)

    (3)

    (4)

    (5)

    (6)

    (7)

    (8)

    (10)

    (9)

    (11)

    (12)