A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

72
A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY LIFTING POTENTIAL FLOW PROBLEMS by BONG-JIN IM, B.E. A THESIS IN MECHANICAL ENGINEERING Submitted to the Graduate Faculty of Texas Tech University in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE IN MECHANICAL ENGINEERING Approved Accepted August, 198 2

Transcript of A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

Page 1: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

A NUMERICAL METHOD FOR THE CALCULATION OF

UNSTEADY LIFTING POTENTIAL

FLOW PROBLEMS

by

BONG-JIN IM, B.E.

A THESIS

IN

MECHANICAL ENGINEERING

Submitted to the Graduate Faculty of Texas Tech University in

Partial Fulfillment of the Requirements for

the Degree of

MASTER OF SCIENCE IN

MECHANICAL ENGINEERING

Approved

Accepted

August, 198 2

Page 2: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

/I'C^ IP

^ '-^'^:&^'

//Of he-'

ACKNOWLEDGMENTS

I sincerely thank Dr. J. W. Oler for his direction of

this project and for much of my training in fluid dynamics.

I also express my appreciation to Dr. J. H. Strickland and

Dr. Raouf A. Ibrahim for their many helpful suggestions.

I further express my gratitude to the other graduate

students of Dr. Oler and Dr. Strickland's group, especially

to Tony G. Smith and Gary Graham, for without their unself­

ish assistance on many aspects of this work, this project

would have been impossible.

11

Page 3: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

TABLE OF CONTENTS

ACKNOWLEDGMENTS ii

TABLE OF CONTENTS iii

ABSTRACT iv

LIST OF FIGURES v

NOMENCLATURE vii

CHAPTER page

I. INTRODUCTION 1

1 . 1 Purpose and Scope of the Research 2

1.2 Review of Previous Research 2

II. FORMULATION OF THE POTENTIAL FLOW MODEL 5

2.1 The Mathematical Representation 5 2.2 Solution Method 8 2.3 Numerical solution by the Collocation

Method 11 2.4 Evaluation of the Influence Coefficients . 18 2.5 Calculation of Airloads 22 2.6 Separated Flow Model 24 2.7 Outline of Computer Program 34

III. RESULTS 37

3.1 Steady Flows 37 3.2 Unsteady Flows 45 3.3 Separated Flow 52

IV. CONCLUSIONS AND RECOMMENDATIONS 60

REFERENCES H

111

Page 4: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

ABSTRACT

A potential flow model for two-dimensional airfoils in

unsteady motion with boundary layer separation is described.

The airfoil and wake surfaces are represented by a finite

set of uniform strength doublet panels. The doublet

strengths on the airfoil surface are determined by applying

a kinematic surface tangency condition to a Green's function

representation of the potential field, while simultaneously

enforcing the Kutta condition. Wake shedding is governed by

a dynamic free surface condition and the characteristics of

the flow near any boundary layer separation points. Wake

deformation is predicted by applying a geometric free sur­

face condition.

Calculation results are presented for steady motion,

impulsively started rectilinear motion and sinusoidal pitch

oscillations .

IV

Page 5: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

Figure

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

3.1

3.2

3.3

3.4

LIST OF FIGURES

Page

Finite Element Repersentation of Airfoil

and Wake Surfaces 13

The Kutta Condition at the Trailing Edge . . . . 14

Evaluation of Element Influence Coefficients . . 21

Separated Flow Model 25

The Bound Vorticity on an Airfoil 29

The Net Rate of Vorticity Shedding at the

Trailing Edge 31

The rate of Change of Potential Jump Across

the Trailing Edge 32

The Flow Chart 36

The Steady Pressure Distribution on a Circular

Cylinder 39

The Steady Pressure Distribution on a Joukowski

Airfoil 40

The Steady Pressure Distribution on a NACA0015

at 0 Angle of Attack 41

The Steady Pressure Distribution on a NACA0015

at 3 Angle of Attack il2

V

Page 6: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

3.5

3.6

3.7

3.8

3.9

3.10

3. 11

3.12

3.13

3. 14

3.15

3. 16

The Steady Pressure Distribution on a NACA0015

at 6° Angle of Attack 43

The Steady Pressure Distribution on a NACA0015

at 11.3° Angle of Attack 44

Potential Jump Distribution on an Impulsively

Started Flat Plate 47

Indicial Lift and Circulation on an Impulsively

Started Flat Plate 48

The Trailing Wake Shape behind the Impulsively

Started Flat Plate 49

The Trailing Wake Shape behind the Periodically

Plunging Flat Plate 50

The Unsteady Pressure Distribution Development

on a NACA0015 at 0.1 rad. Angle of Attack . . . 51

Calculated Separated Wake Geometry(I) 53

Calculated Separated Wake Geometry(II) 54

The Calculation of The Airloads 56

Assumed Separation Point by Katz's Data . . . . 57

Velocity Profiles of Single Vortex 58

VI

Page 7: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

NOMENCLATURE

A = Normal induced velocity coefficient

C = Potential influence coefficient

c = Chord length of the airfoil

CI = Lift coefficient

Cd = Drag coefficient

D = Normal downwash

P = Pressure

R = Distance between control point and field point,

r = Position vector of control point

S = Airfoil surface

TE = Trailing edge

t = time

U = Local velocity

Uco = Free stream velocity

W = Wake surface

D = Velocity potential

a = Doublet strength

= Unit normal vector on the airfoil surface

E, - Position vector of field point

CO = Angular velocity of the airfoil

vii

V

Page 8: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

r = Circulation

Y = Vorticity

Vlll

Page 9: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

CHAPTER I

INTRODUCTION

From airplane wings to automobiles, the prediction of

aerodynamic characteristics has been an important subject.

Aerodynamic optimization techniques have been developed for

military, aerospace, and commercial needs. Recently, the

effective utilization of wind energy has been the focus of

many aerodynamic studies.

In the design of airfoils, the maximum lift and stall

behaviour are among the most important characteristics.

They strongly influence the efficiency and maximum power

output of wind turbine power generation systems, and the

take-off and landing distances for airplanes.

Potential flow analysis has been the principle tool for

the prediction of aerodynamic characteristics of two-dimen­

sional airfoils where viscous effects are limited to thin

boundary layers. In cases where the boundary layer becomes

thick or separates, classical potential airfoil theory must

be modified to account for the added vorticity in the flow.

Page 10: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

The presently reported research has been directed at devel­

oping a model capable of handling cases such as these.

J.J Purpose and Scope of the Research

The research described herein has been directed at de­

veloping a mathematical model which can predict the aerody­

namic loads for airfoils of arbitrary geometry in unsteady

and possibly, separated flows. The following is a list of

efforts which were conducted during the course of this work.

(a) Develop an unsteady aerodynamic airfoil model to be

used for both flat plates and airfoils with thick­

ness in nonseparated flows.

(b) Adapt that mathematical model for airfoils at high

angles of attack with boundary layer separation.

(c) Verify the model with existing airfoil data.

1.2 Review of Previous Research

A historical review by Kraus(1978) of panel methods in

potential flow analysis reveals that one of the first uses

of this method was by A. M. S. Smith(1962) for an airfoil

Page 11: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

with zero lift. Since that time, most applications of panel

methods have been for steady flows, although the general

method is suited to unsteady flows as is evidenced by the

works of AshleyC1966), Djojodihardjo and Widnall(1969),

Summa(1976), and 01er(1976). All of these investigations

have delt with nonseparating flows where the wake vorticity

is shed smoothly from a well defined trailing edge.

The potential flow analysis of separated flows has been

limited to bluff bodies for which conformal mapping

techniques may be applied. Sarpkaya's(1979) investigation

of the flow behind a circular cylinder is a typical example.

Finite element methods have not previously been applied to

bluff body problems involving boundary layer separation.

Katz(1979) has used a discrete vortex method to

represent the separated non-steady flow about a cambered

airfoil. His flow modeling is based on thin airfoil

theory. The separation point was assumed to be known from

experiment or from a separate boundary layer calculation.

The calculated results are in good agreement with the

corresponding experimental data.

The preceeding may be generalized by noting that panel

methods for nonseparated flows and discrete vortex methods

for separated bluff body flows have been advanced to the

Page 12: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

point that acceptable engineering predictions may be made.

The present research is directed at the formulation of a

numerical model which incorporates both panel and discrete

vortex methods for the prediction of separated flows from

airfoil shapes.

Page 13: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

CHAPTER II

FORMULATION OF THE POTENTIAL FLOW MODEL

2,J_ The Mathematical Representation

*

Consider the motion of a two-dimensional airfoil

through a homogeneous, incompressible, and inviscid fluid.

The airfoil surface is represented with respect to a sta­

tionary coordinate system by,

S(r,t) = 0 . (2.1)

The wake following the airfoil may be defined by a surface

of potential discontinuity given by,

W(r,t) = 0 . (2.2)

The possibility of separated flow is accounted for by allow­

ing the wake to include surfaces of potential discontinuity

emanating from a boundary layer separation point as well as

from the trailing edge.

Page 14: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

Since the airfoil plus wake comprise a complete lifting

system and assuming that the ideal fluid was started from a

state of rest or uniform motion, it follows that the motion

is irrotational for all tim.es. This requirement of irrota-

tionality is a necessary and sufficient condition to guaran­

tee the existence of a velocity potential, i.e..

V X u = 0 (2.3)

therefore,

u = u + ve (2.4)

Conservation of mass for an incompressible fluid re­

quires that the vector velocity field not diverge. This re­

quirement may be expressed as

U = 0. (2.5)

Substituing £q.(2.4) into Eq.(2.5) yields Laplace's equation

which is the governing equation for this flow, i.e..

V = 0. (2.6)

The solution to Eq.(2.6) may be obtained through appli­

cation of the appropriate boundary conditions:

(1) The Infinity Condition

Page 15: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

The disturbance potential resulting from the

presence of the airfoil must vanish at infinity.

(2) The Kinematic Surface Tangency Condition

On the airfoil surface, the normal relative fluid

velocity must be zero.

(3) The Kutta Condition

At all times, the flow of fluid from the trailing

edge must be smooth and continuous.

(4) The Boundary Layer Separation Condition

The sheet of potential discontinuity emanating from

a boundary layer separation point must reflect the

injection of the boundary layer vorticity.

(5) The Dynamic Free Surface Condition

The pressure must be continuous through the wake

surfaces, since they cannot sustain a load.

(6) The Geometric Free Surface Condition

The wake particles are convected downstream at the

local convection velocities.

Once the potential field has been determined, the

pressure distribution on the body may be found from

conservation of momentum which takes the familiar form of

Bernoulli's equation:

p = p^ . p | | A , 1 ,V„2} (2.")

Page 16: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

8

2.2 Solution Method

As stated in the previous section, the governing equa­

tion is the linear Laplace equation. By use of Green's

theorem, it may be shown (see Lamb, 1932, pp. 57 - 59) that

the velocity potential at any point in the flow may be given

by

o W ± dS (2.8)

a

W

= the potential doublet strength on S(^,t)=0

A(|)" r the potential doublet strength on W(I,t)=0 -r ->

V = the surface normal on S( i ,t)30 or W(Z,t)=0

R = the vector distance between the "field" point r,

and "source" point, i .

It should be noted that the infinity condition is inherently

satisfied by Eq.(2.8).

The kinematical surface tangency condition on the sur­

face of the airfoil may be expressed (see Karamcheti, 1966,

pp. 190 - 192) as

1^ + li + 3 3 t 3 n

n = 0 (2.9)

on S(r,t) = ' n

Page 17: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

->-n is a local surface normal and — represents the downwash

3t

on the airfoil due to the airfoil motion. It may be

expressed for a body fixed coordinate system as

3n at - (SR B ^ ^ n (2.10)

-> where U^ = airfoil translational velocity vector

CO = airfoil angular rotation vector. a Substituting into Eq.(2.9) yields

ad) / -> -^ ->•, n + U n = 0 (2.11)

which is valid for a body fixed reference frame.

Substituing Eq.(2.8) into Eq.(2.9) yields

^=rrr- f f a -—,-- — dS = 2n dndv \R /

n + . ^ /; A: 4-(i)ds! (2.12)

on S(r,t) = 0 .

This provides a governing integral equation for the unknown

doublet strength distributions on the airfoil and wake

surfaces. Once Eq.(2.12) has been solved subject to tr.e

remaining boundary conditions, the potential at any point ir,

the flow may be determined through Eq.(2.8).

Page 18: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

10

The solution of equation Eq.(2.12) is made difficult by

the nonlinearity which arises from the fact that the doublet

distribution on the wake as well as the wake location is

dependent on the doublet distribution on the airfoil

surface. That is, the location of the wake at any instant

is a function of the previous velocity potential fields

which are also functions of the previous wake geometries.

A complete solution may be obtained by employing the

following step-by-step procedure.

(1) At t = 0, let the airfoil be started impulsively

and the freestream velocity brought instantaneously

to U^ with respect to the stationary coordinate

system. For this instant, there is no wake surface

and no contribution to the downwash on the airfoil

by the wake. A unique solution for the potential

field may be found through a simultaneous

application of the surface tangency (Eq. 2.9) and

Kutta conditions.

(2) Over the next infinitesimal time increment, assume

that the corresponding potential and velocity

fields are unchanged. As a result, the wake

surface generated during that time increment may be

predicted through application of the Kutta and

boundary layer separation conditions.

Page 19: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

11

(3) For the next time step, the integral over the wake

surface in Eq.(2.12) is known and the equation may

once again be solved with the Kutta condition for

a .

(4) Again assuming the velocity field to remain

constant over the time increment, the new geometry

of the existing wake surface may be calculated

through application of the geometric free surface

condition. In addition, new wake elements are shed

as before.

(5) Steps (3) and (4) are repeated so that the solution

proceeds in a step-by-step manner towards the

steady state or periodic final result.

2.3 Numerical solution by the Collocation Method

In its exact integral form, Eq.(2.12) does not lend it­

self to efficient solution by a digital computer. The situ­

ation may be improved by applying a collocation or finite

element solution technique. For this purpose, the airfoil

and wake surfaces are discretized into M and N(t) elements,

respectively, as shown in Fig.2.1. Over each surface ele­

ment, the unknown doublet strength distribution is

approximated with a uniform distribution of unknown

Page 20: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

12

magnitude. In addition, the centroids of the surface

elements on the airfoil are idenfified as control points at

which the surface tangency condition is satisfied exactly.

In this way, the integral form of the surface tangency

condition given by Eq.(2.12), may be reformulated as M

simultaneous equations. Each equation represents the

application of the surface tangency condition to an

individual control point.

In addition to the surface tangency condition, a Kutta

condition must be applied at the trailing edge to uniquely

specify the net circulation about the airfoil. Although

there are many ways in which this condition may be applied,

the essential requirement of all forms of the Kutta

condition is that the flow proceeds smoothly from the

trailing edge of the airfoil. Actual enforcement of the

condition may be accomplished by specifying the direction of

wake shedding or by matching the upper and lower surface

trailing edge pressures (or velocities if a steady flow).

Whatever the method of application, the Kutta condition

provides an additional boundary condition which serves to

represent the essential consequence of viscous boundary

layers in a real fluid flow.

Consider the case of an isolated airfoil discretized

into M elements as shown in Fig.2.1. There are M surface

Page 21: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

13

0 <¥• L.

-w—

CO

«•-0

c o

f-

+> a

V c 0 vi V L.

M 0) 0 (0

»•-i .

Q- 3 o m

Q:

+* c

0) ^ (0

O 2 £ 03 '

.—

u 0)

•»-»

•"• c

• w -

u.

•D c a

OJ

Page 22: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

14

tc

CVJ I z

CVJ

D

u

c

Id

1} X

+> 10

c o

TO c o

(J

•p •p

3

X OJ

OJ

Lt.

Page 23: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

15

tangency conditions plus the Kutta condition or M+1

equations. To satisfy the M+1 simultaneous relations, there

are only M unknown doublet panel strengths, a , so that the

problem, as stated, is overspecifled. Either an additional

singularity of unknown strength must be added to the flow or

the number of boundary conditions must be reduced. The

latter approach has been followed in the present

investigation .

Rather than applying surface tangency conditions on

both the upper and lower surface elements at the trailing

edge, the flow is required to be tangent to the trailing

edge bisector as shown in Fig.2.2. In this way, the Kutta

condition as well as approximate forms of the surface

tangency conditions at the trailing edge elements are

satisfied. Therefore, the three boundary conditions at the

trailing edge are replaced by a single one and the total

number of boundary conditions becomes M-1. A final

condition on the unknown doublet strengths is formed by

assuming that the potential jump across the trailing edge

has equal contributions from the upper and lower elements,

i.e. A4)_ = 0-0 and a - -a^ TE TE

/2.

With these approximations, the surface tangency

condition of Eq.(2.12) may be rewritten in matrix for-na as

[A(i,j)]ja.} = [Dj - [A^(I,D)] |A:''J1 (2.13)

Page 24: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

16

where A(i,j) = normal induced velocity coefficient at

ith control point on airfoil surface due

to jth source element on airfoil surface.

1 ;2 2n ^J an.Bv. S. 1 J ^ ) -

A^(i,j) = normal induced velocity coefficient at

ith control point on airfoil surface due

to the jth source element on the wake. •

1 // 5'

3 -^

^IdW

a. = strength of the jth doublet element on

airfoil surface

W A(j). = strength of the jth doublet element on wake

Di = normal downwash at the ith control point due

to the freestream velocity and motion of the

airfoil

One may now recognize the product, [A(i,j)] ja.f , as

the normal induced velocity on S due to the disturbance

field created by the presence of S itself.

r W • 1 ( W)

[A (i,D)j]A4) { is the downwash on S due to the wake and

|D.| is the downwash due to the relative freestream fluid

and airfoil motions.

Page 25: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

17

Recall that the airfoil and freestream are started

impulsively such that the wake doublet strengths, h^^'-| ,

are known for that instant and all later ones. Therefore,

it is convenient to define

Bi = total downwash array

so that Eq.(2.13) becomes

[A(i,j)] {c } = {B^} . (2.15)

This linear equation set may be solved for the airfoil

surface doublet distributions, {a.}, i.e..

\o.\ =[A(i,j)] -' \B.\ . (2.16)

It should be noted that the normal velocity influence

coefficient matrix, [A(i,j)], does not change with time

since the airfoil maintains a fixed geometry with respect to

a body fixed coordinate system.

Once the unknown doublet strengths have been

determined, the potential for any number of points in the

field may be found from a matrix expression of Eq.(2.8) or

U\ = [C(i,j)] ja } + [c"(i,j)] |.." I (2.1")

Page 26: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

where C(i,j) = potential influence coefficient at the

ith control point due to the jth source

element on S

C^(i,j) = potential influence coefficient at the

ith control point due to the jth source

point on W

The potential influence coefficient matrix, [C(i,j)], like

the normal velocity influence coefficient matrix is

independent of time.

2.4 Evaluation of the Influence Coefficients

As noted in the previous section, the representation of

a general doublet distribution on a surface element by a

uniform distribution permits the definition of normal vel­

ocity and potential influence coefficients. These were giv­

en by

A(i,j) = normal induced velocity coefficient at the ith

control point of S due to the jth source element.

i_ / r _ f i ) dS 2n i' 5n3v \R

(2.17)

C(i,j) = potential influence coefficient at the ith con­

trol point due to the jth source element.

2n g- .v \R i 1 dS (2.18)

Page 27: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

19

The direct evaluation of the integrals of Eq.(2.17) and

(2.18) may be avoided by taking advantage of the analogy

between surface distributions of doublets and vortices. It

may be shown (Oler & Strickland, 1980, pp. 87 - 99) that a

general distribution of doublets may be represented by a

distribution of vortices on the surface. The strength of

the vortex sheet at any point is equal to the gradient of

the doublet strength with the vortices oriented normal to

that gradient. For the particular case of a surface element

having a uniform doublet distribution, an equivalent

representation is that of a vortex ring on the boundary of

the element with strength equal to the element doublet

strength. This is illustrated for a two-dimensional surface

element in Fig.2.3

The influence coefficients may be determined by

evaluating the vortex equivalents of the doublet elements.

Referring to Fig.2.3, the potential influence coefficient

may be written as

, . . , el 62 C(i,D ) ^ ^ - —.

2r tan

-1 / 1 y - tan . T • e - 1 / 2

(2.19)

ri-ex/ V^2*ex^J

The normal velocity influence coefficient, may written as

Page 28: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

A(i,j) = •1 e.

2 71 2n r n

2 ' -•

(2.20)

2n

e xr z 1

r 1

e xr^ z 2 I -^ I n

Since the doublet strength for the elements in the wake

are known for all times, it is possible to utilize an

equivalent discrete vortex representation of the influence

due to the wake.

The discrete vortex approach is only a slight variation of

the doublet panel method and offers the benefit of a

significant reduction in computation time.

As described in the previous paragraphs, the net

influence of a two-dimensional uniform strength doublet

panel is equivalent to the combined influences of discrete

vortices at its boundaries with strengths equal to plus or

minus the panel strength. For two adjacent doublet panels,

the contribution to the total influence by the vortex at

their common boundary is proportional to the difference in

the two panel strengths. Since the strengths of the wake

elements are known for all times in the step-by-step

solution procedure, it is possible to replace the panelized

wake with an equivalent discrete vortex representation. An

Page 29: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

21

%=-(r

Fig. 2.3 Evaluation of Element Influence Coefficients

Page 30: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

22

important advantage of this approach is that the influence

of an individual vortex on a particular control point is

calculated only once instead of twice as would be required

in the finite element representation. This is particularly

important for the wake calculations since the wake geometry

changes with each time step and all wake influence

coefficients must be recalculated.

2.5 Calculation of Airloads

The pressure at any point in an irrotational, ideal

flow may be found with the unsteady Bernoulli equation.

P = P. - P J M - I ^^^^)^|- (2.21)

As shown by Summa(1976), this may be rewritten in body fixed

coordinates as

P(r,t) = P - P oo

3 o (r , t)

t -i-

[U -\J^'Z{t)-r^'l?{r,t) + i[vMr,t)]2}. (2.22)

The primary difference in Eq.(2.21) and Eq.(2.22) is the

additional increment to 3t on the body due to the ' .ction

of the body through the potential field. The significance

Page 31: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

23

of this term may be appreciated by recognizing that even if

the potential field was steady, d(^/dt would not be equal to

zero unless the potential field were also uniform.

Eq.(2.22) may be rewritten in a more convenient form

for numerical computation by expressing v^ as

9 6-^ ^ s^ 8n n (2.23)

where v is the surface gradient. Recall that the surface s

tangency condition was written in body fixed coordinates as

a 4) Tn H = -{ CO X ?) n (2.2U)

With Eq.(2.22) , Eq.(2.23) may be expanded to yield

P = P - p CX)

H + (3„ - "B - B " ' • 's*

9n °° B B X r ) n

^ ('s'' ^1 3^ \9) (2.25)

By substituing Eq.(2.24) into Eq.(2.26), we arrive at

Page 32: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

p = p (3* (V - u B

- 03 B

X r) V i> S^

24

- T [(U^ - U^ - .. X ? B B

) . n]

^ 1 (v^^)^} (2.26)

Eq.(2.26) provides the advantage of reducing the computation

of the gradient of the disturbance potential to the

computation of its surface gradient. With Eq.(2.26), the

airloads on the airfoils may be calculated by integrating

the pressure force vector components over the surfaces.

2.6 Separated Flow Model

For modeling purposes, it is assumed that the wake may

be adequately represented by two sheets of potential discon­

tinuity. One surface extends from the trailing edge while

the other originates at the boundary layer separation point

as illustrated in Fig.2.4.

The rates at which vorticity is shed into the two wake

surfaces are related by the Kelvin-Helmholtz theorem to the

rate of change of the vorticity bound to the airfoil

surface. The theorem requires that the rate of change of

net vorticity in the flow field is zero, i.e.,

Page 33: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

25

-J^-G

F i g . 2 . 4 S e p a r a t e d F l o u Model

Page 34: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

or

26

dV net 3t

= 0

arb 3rw srs ^ bt 3t

(2.27)

Here, the net vorticity has been divided into three

components: the vorticity bound to the airfoil surface,

r, , the vorticity shed from the boundary layer separation

point, VQ , and the vorticity shed from the trailing edge,

r,, . The time derivatives of r,, and r represent the rate

of vorticity shedding to the respective wake surfaces.

A simple vorticity flux analysis may be utilized to

estimate the vorticity shedding rate from the boundary layer

separation point.

dr 6 / V i U at f^u[^ - ^ ] d y ^VI^ ^

5 1 |_ (u2) dy 2 9y (2.28)

U

2 •

Ue is the velocity at the edge of the boundary layer or the

surface velocity calculated by the potential flow routine.

Page 35: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

27

The assumption is made that 100^ of the vorticity contained

in the boundary layer is injected into the inviscid flow

field at the separation point.

If "b is the vorticity per unit length along the

airfoil surface and a is the distribution of potential

discontinuity or doublet strength along the surface, then,

referring to Fig.2.5, the bound vorticity may be written as

^b = A^^ ^b ^S

= /^ dS A- dS °^ (2.29)

= Ac TE •

The rate of change of bound vorticity may be expressed in

terms of the difference in surface doublet strength at the

trailing edge or potential jump across the trailing edge:

b ^ £_ dt dt " TE •

(2.30)

Substituting Eq.(2.28) and (2.30) into Eq.(2.27)

provides an expression for the rate of shedding of vorticity

to the wake surface extending from the trailing edge.

dr w dt

d_ dt

( a_ ) + TE

U 2

2 I (2.3')

Page 36: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

28

Recall that Eq.(2.31) was based on the Kelvin-Helmholtz

vorticity conservation theorem. The same result may be

arrived at by applying the dynamic free surface boundary

condition at the airfoil trailing edge. Specifically, the

pressure difference across the infinitely thin surface must

be zero since it cannot sustain a load. This leads to

P - P, = 0 u 1

It ^^ - h^ ^ 1 (*u - '^V ' (2.32)

Recognizing that 7cf) = u for the fluid fixed reference

frame, then

2 u u|) = 3t ^^"TE^ (2.33)

From Fig.2.6, it is noted that (u2-u2)/2 is the net rate

of vorticity shedding from the boundary layers on the upper

and lower surface of the airfoil at the trailing edge.

Then,

dr W dt dt ^^^TE^

(2.3^)

Page 37: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

29

Fig. 2.5 The Bound Vorticity on an Airfoil

Page 38: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

30

By calculating the circulation about a curve, as shown in

Fig.2.7, it is apparent that

dt --^TE^ dF" HF" (2.35)

So,

W dt dt + dt

(2.36)

which is equivalent tc the result obtained from th(

Kelvin-Helmholtz theorem.

An important conseq lence of boundary layer separation

may be noted by applying the dynamic free surface condition

to the wake surface exteniing from the separation point. Let

points A and B be located an infinitesimal distance ahead of

and behind the boundary layer separation point. The

pressure difference across the two points must be zero which

results in

ft ^^A - ^B^ ^ 7 ^^*A - ^^B^ = '' (2.37)

As in Eq.(2.33),

O 1

}t 1 r A 2

2 ^^^A V - ) (2.38)

Substituing into Eq.(2.39) yields

Page 39: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

31

10

Ui

c TO •o O

X

>>

u

+>

o >

o

o +> Id

0)

•D

u c

fd i .

CO •

OJ

Hi

X

Page 40: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

32

O

u a: a E 3

»->

4J

c •p o

Q.

«*-

O

o

c 10 X

u o o +> Id

Q: I)

X

U

C

Id I .

h-

o X

Page 41: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

33

3(})

Tt B

3(j).

w dr

i

dF (2.39)

Therefore, it is noted that behind the boundary layer

separation point, there is an additional increment to 3(j)/3t

equal to the rate of vorticity shedding from the separation

point.

The same observation may be made by considering the

rate of change of the potential jump across the trailing

edge as described by Eq.(2.35) which may be rewritten as

9 9 u

3i 1

dr dr

at 3t + dt + dt (2.40)

For the case of a steady, stalled airfoil, the average rates

of change of ^^ and r^ are zero, yet the d(}) /dt is not zero

due to the vorticity being shed from the boundary layer

separation point.

The additional contribution to d^/dt in the separated

region is important in the calculation of the pressure

distribution around the airfoil. Without its inclusion, a

pressure jump would be indicated across the two wake

sur faces.

Page 42: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

34

_2.2 Outline of Computer Program

The governing equations which are described in Eq.(2.8)

through (2.12) have been incorporated into a computer code

which follows the step-by-step solution method discussed in

section 2.1. The program consists of a MAIN routine which

controls the general flow of the computations and 19 su­

broutines. The functions of the subroutines are as follows:

CPDIST .... calculates the airloads

DECOMP .... decomposes the influence coefficient matrix

DWSH calculates the downwash due to the free

stream

GEOM calculates the geometric description of the

airfoil

GRAD finds the surface gradient of the velocity

potential

INFO computes the influence coefficient matrix

MATRIX .... controls the matrix calculations

NUGEOM .... finds the new geometry for the wake

SETUP inputs the data and initializes the data

arrays

SHEDWK .... calculates the position of the wake

elements shed at the current time step

SOLVE .... solves the set of linear equations

TMSTEP .... controls the excution with time increren^s

Page 43: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

35

VELWK computes the velocity vectors for the wake

vortices

WAKINF .... adds the contribution to normal downwash

due to the airfoil motion and free-stream

MOTION .... finds the location and velocity of the foil

at every time step ARCTAN .... find the

angle in the range (0.0, 2PI) whose tangent

is tan(Y/X)

SIZE calculate the viscous core size for the

wake vortices

VISCOR .... computes the local velocity induced by a

viscous wake vortex

PRTPLT .... plots the calculation results on a line

printer

Fig.2.8 is a flow chart of the calculation procedure.

Page 44: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

36

( START

RERD THE DRTR RND SETUP THE GEOMETRY OF THE RIRFOIL

EVRLURTE THE INFLU­ENCE COEFFICIENTS

Yes

GENERRTE THE NEW WRKE ELEMENTS RND

FIND THEIR NEW LOCR-TION

COMPUTE THE DOWN-WRSH DUE TO THE WRKE ELEMENTS

CRLCULRTE THE DOWN-WRSH DUE TO THE

FREESTRERM

No

Yes

STOP

F i g . 2 .8 The Floiu Char t

Page 45: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

CHAPTER III

RESULTS

The potential flow model has been evaluated for three

basic flow conditions: steady flow, unsteady nonseparated

flow, and unsteady separated flow. The calculation results

have been evaluated on the basis of comparision with experi­

mental data and analytical solutions.

3_>^_ Steady Flows

Steady solutions may be obtained in two different ways:

(1) Allowing the flow to develop until the unsteady in­

itial effects fade away.

(2) Forcing the initial condition to be equivalent to

the steady state condition.

Even though the first method is a viable approach to

the calculation of steady state solutions, it is computa­

tionally inefficient due to the v very long computation tir.es

required. It has been found that approxi.r.ately 10 chord

37

Page 46: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

38

lengths of travel or 100 time steps are required for an im­

pulsively started rectilinear motion to approach 90% of its

steady solution. An additional 10 chord lengths are re­

quired to bring the calculations to 955& of the steady va­

lues. Consequently, the calculation of steady flows by this

method has been found to be impractical on the basis of the

computation time requirements.

The doublet strength distribution and circulation about

the airfoil are constant with respect to time for a steady

flow. Consequently, the rate of vorticity shedding from the

trailing edge to the wake is zero and the only vorticity in

the wake is the starting vortex which is assumed to be an

infinite distance behind the airfoil. The previously de­

scribed steady potential on the airfoil may be calculated

simply by forcing the vorticity on the airfoil surface to go

to zero at the trailing edge. This is done by setting the

influence of the trailing edge vortex to zero in the formu­

lation of the influence coefficient matrices. The complete

steady solution can then be obtained with a single pass

through the program.

To get a crude idea of the convergence requirements on

the number and size of the panel elements, several

calculations have been made for the circular cylinier. The

comparision of the pressure distributions for different

Page 47: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

39

4>

c o E o

»— o

•> c o E o

*— c

(9 S CU en a>

Q. O

1 .

u

Id

3 O (.

o fd

c o

c o

3 JO

L. 4>

Q

L. D

O

a. >s

"O tJ O

o X

Page 48: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

40

E c

CU

c o •» c 3 O O 4>

0) 3

•TO O O 03 •» 3 *> Q. O £ m O X

o 3 •P

a. o E « O X

C3 (\J

• •

I I

O O

a. o -«LL.

I I

o 1.

3 o 3 O

»-)

c o c o

JO

C L. 3 O (1 O L.

Q.

T3 Id O 4* 0)

c

X

OJ

•3

iZ

Page 49: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

41

O

c o o

o

+• o

JQ JO

tr c o

o

T3 V t.

c en •o v N

c

•a u

+> 3

a £ O

u

-p Id

in « - H

(S G}

CE U cc z iO

£ O

£

. ^ •

E 0)

•— 0)

O G} * -+> 3

X 1 —

L. 4-> C9 •^ Q

0) i . 3 O CO 0)

in • ^ ^

JC u id •p p CE

<f-o 0)

r—

O) c

1 - CE Q.

>> T l Id 0)

0) 0) L. O) «)

•p n (/)

ID X h-

cn »

n •

en -m—

u.

cs

Page 50: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

42

+> a

o \ X

in

(S

CE

u

,r

Q.

o

c 0

••"

bu

t

L.

d **

Q

0 L. 3

0 L.

Q.

X Id c

4> OD

0 X K

^ •

05

• 05

^

u.

OJ

o Id •P •P (E

O

o m C

CE

Vi

c i .

C9 •

0)

Page 51: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

43

U

**-

o c o

Q

o S3 JO

C

o

u •o o L.

a. o +*

c en

<o N

&. It) 0)

c

-o u

• J

3 a. E O

u

^

J

i^"—1»—I 1 t

J

Id

in •—«

C3

CE U

to

o 5 0)

3 w

1 .

I

Q

Id 0) p

X

in

0)

0)

o Id

CE

0)

c CE

1-U) 0)

CD

Page 52: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

44

o \ X

u

p to

m ^ cs cs CE o CE z

a

c 0

c o •^^ •P 3 X •»-L. •P d

• i -

n c &_ 3 M

y ^

• £ u — o (9 OJ .

^ o Id •p p CE

«*-o o ^-O) c

M CE O t. a.

X "D Id

(1 o o 1. O) o

C Q P CO cn

0 X 1-

(D

n

• u.

• — »

Page 53: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

^5

numbers of doublet elements is plotted in Fig.3.1. 20 or 30

equally divided elements are noted to yield very good

agreement with the exact theoretical solutions.

To test the capability of the model to represent the

flow about a lifting body, the symmetric Joukowski airfoil

was examined. In Fig.3.2, the computed pressure

distribution around a 20 element representation of the

Joukowski airfoil is plotted against the exact solution.

The elements are spaced so as to have a denser element

distribution near the nose of the airfoil where flow

variations occur most rapidly. The element spacing has a

strong influence upon the calculated pressure distributions.

This is especially true around the nose of the airfoil.

The pressure distribution for the NACA0015 airfoil is

tested in the same fashion as with the Joukowski airfoil,

with the results illustrated in Figures 3.3 through 3.6.

3.2 Unsteady Flows

The case of an impulsively started flat plate airfoil

provides a useful test of the potential flow model in an un­

steady flow. Fig.3.7 exhibits the numerically calculated

distribution of potential jump along the airfoil at the

starting instant together with the exact solution. Fig.3.?

Page 54: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

46

illustrates the indicial circulation and lift experienced by

the flat plate during the impulsively started rectilinear

motion. Comparision is made with the linearized analytical

solution. The presented flat plate computation uses 10

equally spaced elements and employs a time step equivalent

to the time needed for the airfoil to travel 0.1 chord

lengths or the length of the trailing edge element. This

follows the suggestions of R. H. Djojodihardjo and

S. E. WidnalK1969) . Fig.3.9 shows the instanteneous wake

geometry after 0.5 and 1.0 chord lengths of the travel. The

beginning of the rollup of the wake is observed after 1

chord length of tavel in Fig.3.9-

To illustrate the realistic manner in which the wake

geometry may be predicted, the calculated wake behind a

peoridically plunging flat plate airfoil is presented in

Fig.3.10.

The pressure distribution on an implusively started

NACA0015 airfoil, after 1 and 5 chord lengths of travel, is

illustrated in Fig.3.11- It can be observed in Fig.3-11

that the pressure gradually changes its distribution to the

steady state condition.

Page 55: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

47

^0

1.0 r

.5

.5 1.0 X/c

Fig. 3-7 Potenttal Jump Distribution on an Impulsively Started Flat Plate

Page 56: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

48

• • «*. «•• • o u •» <•-^» J

— c « o •" • -o •* - a "V c C 3

M L .

•o » O M *» L. 3 • a c E CO o m uz

c 0

•» «

3 o L ^ U C

o ^^ ^» • • * •^ 3 o — — o -DO} c M *

I I •o o o c •* o 3»-> a • E H o • UQ:

o \

3

Ci«

1 CD

CD

S

n (9

• cn

(S •

(U

8) >

3 a E

C Id

c o c o o

— +i +> Id /O —

— CL 3 O -P L. Id

u LZ

C 0)

•p Id «p -p

-J

Id

o

c

CD

DO

L.

Page 57: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

49

in Q

I I

u \ • * :•:

U \ • » 4c

ID Z3

<5 O

>

3 CL E

C Id

c X o O -P

CO to

O Q. Q. Id

X en

Id

Id

2

TJ O P i -

CD Id C P - cn

Id

X

D3

Lx.

Page 58: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

u \ X

50

Id o

o 1. o

Q.

c

X o o +*

m Id

Q. Id +*

X Id 05 —

U. O

J£ D) « C

O) c C 3

^ CL

to

o X

eg

cn

m

Page 59: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

51

O C9 •

'-

1

u \ *> :*:

• in 1

o \ +» 4c

3 ID

a O

P c o E a o o >

n Id

O -P — CE o

(S3 O

L. 3

i*G) CE Id Q)

CS

d

c X

O)

u.

u

c o

Page 60: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

52

2*2. Separated Flow

When the boundary layer separates prior to the

trailing edge, an additional wake surface must be generated.

Figures 3«12 and 3»13 illustrate a typical calculated geome­

try for the wake in a separated flow and the type of diffi-

culies which have been encountered in making these calcula­

tions. From the figure, it is apparent that the wake

surface has been allowed to across the airfoil surface.

This is a consequence of the combined effects of modeling

the airfoil and wake surfaces with discrete vortices and

representing an unsteady flow with a step-by-step solution

method utilizing finite time increments.

Fig.3.14 depicts the variation of lift and drag with

respect to angle of attack. For these calculations, it was

necessary to assume a location for the boundary layer sepa­

ration point as illustrated in Fig.3-15. The data points

for the separated flow were taken from the time step previ­

ous to the event of the wake crossing the airfoil surface.

Several modifications to the potential flow model are

currently being evaluated. These changes are directed at

solving the wake crossing problem and improving the overall

accuracy of the model .

Page 61: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

53

09 c n

CU CD CVJ

* v . *

CM • CVI •-*

P

£

o

Id

P Id Id a in •D 0) P Id

3 O

Id U OJ

cn

U.

Page 62: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

54

cn o

Q

CU (D cn

• ^ 4

CU • CU •«-•

I I ^ o

\ p

J

c p o E o o

3:

V P Id L . Id Q.

o in •a o

p Id

3 O

<J U

d

cn

i l

Page 63: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

55

The wake's crossing of the airfoil surface is primarily

the consequence of the descritization of a continuous fluid

flow problem. For instance, the airfoil surface is modeled

by a finite set of doublet panels whose strength are

determined such that the flow is required to be tangent to

the airfoil surface at a finite number of control points.

Between the control points, the flow is not tangent to the

surface and the streamlines go in and out of the airfoil at

various points. A free wake vortex could be convected

through the airfoil at critical points between the control

points. Possible remedies to this particular problem are to

use more elements to model the airfoil surface or to use

higher order doublet distributions on the elements so that

the number of control points can be increased. Schemes

involving the interpolation of velocities between control

points are also being considered.

Recall that the wake surfaces are modeled by discrete,

potential vortices, which induce infinite velocities as

their centers are approached. As a consequence,

unrealistically high induced velocities are indicated when

vortices happen to be convected very close to one another.

This problem may be alleviated by utilizing Rankine vortices

which have viscous cores instead of the inviscid, ootenUal

vortices (Fig. 3.16). By reducing the induced velocity and

Page 64: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

56

— a 3 0 — O 10

ck: ^ c

o £ 3 ft. CL O

E a O X

O Ui

< I

\ \

\ \

\ \

\

\ \

\ \

\

s\ s

\ \

\ \

\ \

\

\

\

\

in CVI

\

\

\ - .

\ . .

\

\

s eg (Vi

s •

in

eg CO

^ i .

\ in

\ \

\ \

•o

— o U

\ \

03 U) (Vi (Vi

I

M 13 Id O i-

£ X •p < ^

o c o

•p Id

3 U

Id U u

X H-

cn

L.

Page 65: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

57

X/o

1.5 , -

1.0

.5

10 20 30 J a

40

F i g . 3 . 1 5 Rssumed S e p a r a t i o n P o i n t by K a t z ' s D a t a

Page 66: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

58

e c o u tA 3 O O (A

.»->

O \ 3

O L.

o o c ^

>s > S 4 ' 4* -^ — O o o o —

— o o > >

o ft- c « ^ o jm c c

«i - Id

— oc

3 \

« E

3

/ »

I I

I /

X o

•p t.

o >

cn

*p o Vi

«4-

o a. •p

u o

>

cn

Page 67: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

59

consequently the distance traveled by a vortex during a

finite time step, the chance of its crossing the airfoil

surface is reduced.

At the current time, the flow near the boundary layer

separation point is only crudely modeled. Ideally, a Kutta

condition similar to that enforced at the trailing edge

should be utilized at the separation point in the linear

equation set for the solution at any instant in time. This

would guarantee a stagnation point at the boundary layer

separation point. Instead, vortices are injected just above

the airfoil surface at each instant in time which have the

effect of retarding the flow in that region. A better

representation is the generation of a uniform strength

vortex element from the boundary layer separation point at

each time step. The total vorticity in the elements are

lumped into discrete vortices as they are convected away

from the surface and new elements are created.

Page 68: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

CHAPTER IV

CONCLUSIONS AND RECOMMENDATIONS

The two-dimensional airfoil model described in this

thesis is theoretically capable of the prediction of air­

loads for both nonseparated and separated flows. However,

the calculation results reveal that, in its present form,

the model is only partially successful. On the basis of

those calulations, the following conclusions and recommenda­

tions may be made:

1. The model does a satisfactory job of calculating

steady and unsteady nonseparated flows.

2. Accurate predictions of the airloads on airfoils with

boundary layer separation can not be made until sig­

nificant modifications of the model are implemented.

3. One refinement needed in the model is a more precise

representation of the flow in the vicinity of the

boundary layer separation point.

4. In addition, the airfoil surface should be modeled

differently so that the velocity along the surface is

more continuous than the present model allows.

60

Page 69: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

REFERENCES

1 .

2.

3.

4.

5.

6.

7.

8.

9.

10

Deffenbaugh, F. D. and Marshall, F. J. "Time Development of the Flow about an Impulsively Started Cylinder," AIAA J., Vol. 14, July 1976, pp. 908 - 913-

Djojodihardjo, R. H. and Widaal, S. E. "A Numerical Method for the Calculation of Nonlinear, Unsteady Lifting Potential Flow Problems," AIAA J., Vol. 7, 1969, pp. 2001 - 2009.

Gerrard, J. H. "Numerical Computation of the Magnitude and Frequency of the Lift on a Circular Cylinder," Philosophical Transaction of the Royal Society of London, Vol. 261, Jan. 1967, pp. 137 - 162.

Ham, N. D. "Aerodynamic Loading on a Two-Dimensional Airfoil During Dynamic Stall," AIAA J., Vol. 6, 1968, pp. 1927 - 1934.

Karamcheti, K. "Principles of Ideal-Fluid Aerodynamics," John Wiley and Sons, 1966

Katz, J. "A Discrete Vortex Method for the Non-Steady Separated Flow Over an Airfoil," J. Fluid Mech, Vol. 102, 1981, pp. 315 - 328.

Lamb, Horace "Hydrodynamics," Dover Publications, 1932

Kraus, W. "Panel Methods in Aerodynamics," Numerical Methods in Fluid Dynamics, McGraw-Hill, Editors Wirz and Smolderen, pp. 237 - 297.

McCroskey, W. J. "Inviscid Flowfield of an Unsteady Airfoil," AIAA J., Vol. 11, 1973, PP• 1130 - 1137.

McCroskey, W. J. and Phillippe, J. J. "Unsteady Viscous Flow on Oscillating Airfoils," AIAA J., Vol. 13, 1975, pp. 71 - 79.

61

Page 70: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …

62

11. Oler, J. W. "An Investigation of a Numerical Method for the Exact Calculation of Unsteady Airloads Associated with Wing Intersection Problems," M.S. Thesis, University of Texas - Austin., 1976

12. Oler, J. W. & Strickland, J. H. "Dynamic Stall Regulation of the Darrieus Turbine," Progress Report on Sandia Contract No. 7^-1218, 1980, pp. 87 - 99

13. Sarpkaya, Turgut "An Inviscid Model of Two Dimensional Vortex Shedding for Transient and Asymtotically Steady Separated Flow Over an Inclined Plate," J. Fluid Mech., Vol. 68, part 1, 1975, pp. 109 - 128.

14. Sarpkaya, T., and Schoaff, R. L. "Inviscid Model of Two-Dimensional Vortex Shedding by a Circular Cylinder," AIAA J., Vol. 17(11), 1979, PP• 1193 - 1200.

15. Summa, J. M. "Potential Flow About Impulsively Started Rotors," J. Aircraft, Vol. 13, 1976, pp. 317 - 319.

Page 71: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …
Page 72: A NUMERICAL METHOD FOR THE CALCULATION OF UNSTEADY …