78 views

Uploaded by dbrcs

Gives a print of the required glossary

- hicks1978.pdf
- 19790005847_1979005847
- impulseBrief.pdf
- Aero Background
- Jet Engine Fundamental of Theory Design Operations
- The Work of Wings
- QDB 15 POF
- Introduction to Lifting Line theory
- Technical Report Boost Gliders
- aerospace-03-00026
- Aerodynamics
- Buffet Envelope Prediction of Transport Aircraft
- 2002-04-19_airfoil
- Span Efﬁciencies of Wings at Low Reynolds Numbers
- Lecture 5 CCN2299 Assignment 2 2018 Rev2(1st)
- Drilling Case Study
- Donnelly DJ T 2010
- GPHYS Lecture 17 Fluids in Motion 2013F
- Drag
- Chapter 5

You are on page 1of 43

A most difficult aspect of wing design can be choosing the correct airfoil cross sectional

shape. Although most airfoil shapes can support flight, only the right one will save

thousands of dollars in operational costs over the life of the aircraft. MultiSurface

Aerodynamics is a digital wind tunnel that can compare the performance of many airfoil

shapes to make the airfoil selection process easy.

Aerodynamic Center

The aerodynamic center is a point along the airfoil or wing about which the moment

coefficient does not vary with an angle of attack change.

Airfoil

An airfoil is the cross section of a wing. The airfoil shape and variations in angle of

attack are primarily responsible for the lift and profile drag of the wing.

Angle of Attack

The angle of attack is defined as the angle between the plane of the wing (airfoil chord)

and the direction of motion (free stream velocity). The angle of attack can be varied to

increase or decrease the lift acting on the wing. An increase in lift often results in an

increase in drag.

Center of Pressure

A point along the airfoil about which the moment due to the lift is ero, i.e., it is the point

of action of the lift. The center of pressure will change its position when the angle of

attack changes.

Chord

The chord is the dimension of the airfoil from its leading edge to trailing edge.

Circulation

!irculation is a measure of the vorticity in the flow field. "or an inviscid flow field, the

lift is e#ual to the product of the circulation about the airfoil, the density and the velocity.

Computational Fluid Dynamics (CFD)

!omputational fluid dynamics is the term given to a variety of numerical mathematical

techni#ues applied to solving the e#uations that govern fluid flows and aerodynamics.

Modern CFD results can rival the accuracy of wind tunnels in testing airfoils, wings and

entire airplanes for certain test configurations.

Density

The mass of a substance contained in a given volume divided by the volume. "or a

incompressible fluid, the density is considered to be constant throughout the flow field.

$owever, for a compressible fluid, the density can vary from one location to the ne%t in

the flow field. The speed of sound in a fluid depends on the ratio of pressure changes to

density changes in the fluid.

Drag

&rag is an aerodynamic force opposing the direction of motion. &rag can be due to

surface viscosity (friction drag), pressure differences due to the shape of an ob'ect (form

drag), lift acting on an finite wing (induced drag) and other energy loss mechanisms in

the flow such as wave drag due to shock waves and inefficiencies in engines.

Drag Coefficient

The drag coefficient is defined as the drag((dynamic pressure ) reference area). The

reference area is usually the plan*form or flat pro'ection (the wing+s shadow at noon) area

of the wing.

Dynamic Pressure

The dynamic pressure is defied as the product of the density and the s#uare of the

velocity divided by two. The dynamic pressure has units of pressure, i.e. "orce(Area. The

dynamic pressure is used to non*dimensionalie forces and pressures in aerodynamics.

Flap Deflection Angle

The flap deflection angle is the angle between the deflected flap and the chord line. The

angle is positive for a downwards deflection of the flap. &eflect the flap downwards to

increase the airfoil+s lift.

Lift

The lift is a force acting perpendicular to the direction of flight. The lift is e#ual to the

fluid density multiplied by the circulation about the airfoil and the free stream velocity. ,n

level flight, the lift developed by an airplane+s must be e#ual to the weight of the entire

airplane.

Lift Coefficient

The lift coefficient is defined as the lift((dynamic pressure ) reference area). The

reference area is usually the plan*form area of a wing or horiontal pro'ection of the

wing.

ean aerodynamic chord

This chord is located along the wing and has the aerodynamic property of the two*

dimensional wing.

!ACA Airfoils

-A!A airfoils are wing cross section designs invented by the !ACA organiation.

-A!A eventually became -ASA (-ational Aeronautics and Space Administration). $ere

are a few popular airplanes that have -A!A airfoil wings.

Airplane "oot Airfoil Tip Airfoil

/eech 01 Twin /onana -A!A 23145.4 -A!A 23142

/*46 "lying "ortress -A!A 1142 -A!A 1141

!essna 402 -A!A 2542 -A!A 1142

!essna 462 4763*later -A!A 2542 -A!A 2542 mod

!essna 001 !itation ,, -A!A 23145 -A!A 23142

&ouglas &!*3 -A!A 2240 -A!A 2218

"airchild A#$% Thunderbolt ,, -A!A 8648 -A!A 8643

Sikorsky S*84 S$*3 Sea 9ing -A!A 1142 -A!A 1142

Panel ethod

This numerical method places singularities along the airfoil. ,n the case of :isual"oil or

3DFoil , the singularities are vortices. The vorticity is distributed linearly along the panel.

Plain Flap

A plain flap is a hinge attachment near the trailing edge of an airfoil. The length of the

flap is measured as a percentage of the chord and the deflection is measured in degrees.

Pressure Coefficient

The pressure coefficient is a non*dimensional form of the pressure. ,t is defined as the

difference of the free stream and local static pressures all divided by the dynamic

pressure.

"eynolds !um&er

The ;eynolds number is a non*dimensional parameter that compares the inertia to

viscous forces. ,f the ;eynolds number is low, then viscosity plays an importatant part in

the simulations.

'tall

At low angles of attack, the lift developed by an airfoil or wing will increase with an

increase in angle of attack. $owever, there is a ma%imum angle of attack after which the

lift will decrease instead of increase with increasing angle of attack. This is know as stall.

9nowing the stall angle of attack is e%tremely important for predicting the minimum

landing and takeoff speeds of an airplane.

'treamlines

!ontours in the flow field that are tangent to the velocity vector.

Wing Loading

The total weight of the airplane divided by the plan form area of the wing.

Wing 'pan

The span is the total length of the wing.

A&out Dr( )anley

&r. <atrick =. $anley, is the owner and founder of $anley ,nnovations, a small business

specialiing in the development of aerodynamics and fluid dynamics simulation software

for education and industry. &r. $anley earned his /.S. degree (summa cum laude) in

aerospace engineering from <olytechnic ,nstitute of -ew >ork and his S.M. and <h.&.

degrees from the department of Aeronautics and Astronautics of Massachusetts ,nstitute

of Technology (M,T). $e also completed a minor in the area of management of

innovation and technology at M,T+s Sloan School of Management.

After graduating from M,T, &r. $anley 'oined the Mechanical =ngineering faculty at the

?niversity of !onnecticut where he formulated and taught courses in aerodynamics,

compressible fluids, introductory fluid mechanics and heat transfer. As a faculty member,

he won the highly competitive -ational Science "oundation ;esearch ,nitiation Award,

the -ASA*AS== Summer "aculty "ellowship and three consecutive research awards

from -ASA @ewis ;esearch center to study compressible viscous flows in

turbomachinery using pseudospectral methods. This research led to the successful

education of four (5) <h.& students and four (5) Masters degree students. ,n addition &r.

$anley can be credited with a number of publications including the pioneering work in

multi*domain pseudospectral methods for compressible viscous flows entitled AA

Strategy for the =fficient Simulation of :iscous !ompressible "lows using a Multi*

domain <seudospectral MethodA which can be found in Bournal of !omputational

<hysics, :ol 41C, -o. 4, pp. 403*40C, September 4773.

ulti'urface Aerodynamics

TM

is the leading !A= (computer aided

engineering) product for multiple wing design D analysis on the <!.

MultiSurface turns your <! into a powerful wind tunnel D diagnostic

tool for creating D testing airfoils D 3& wing systems.The user*

interface is self*contained and has tools for design, analysis,

presentations and prototyping. MultiSurface computes the lift, total

drag, moments and stability derivatives of your custom wing system

in 'ust seconds on a Eindows <!. ?se different airfoils and wing

shapes to optimie your design on the fly. -o advanced aerodynamics

or computer programming skills are re#uired to use the software.

Design * Test Wings +n ,our

PC-

MultiSurface Aerodynamics features include.

stand*alone application (built*in pre D post processing tools)

easily enter and manipulate comple% wing systemF instantly

compute lift, drag (profile and vorte%), moments and stability

derivativesF calculate trim, neutral point and ground effect.

ability to read*in, modify and analye custom airfoil shapes to test

the effectiveness of original designs.

users do not need to have advanced aerodynamics or computer

programming skills

accurate solvers for realistic simulations

computer aided design interface to model and edit 3& wing

systems ('oined wing for e%ample)

interactive visualiations

graphs D tables for comparative analysis

ability to e%port 2& D 3& !A& files for rapid prototyping.

Lease./uy +ptions

/uy !o0

ore 12amples

Sailplane Analysis

Boined Eings

=ndplates (.pdf)

Sprint !ar Eing (.pdf)

MultiSurface :alidation

$ydrofoil Analysis

=lliptical Eing

Gther Eing Shapes

Applications

MultiSurface Aerodynamics is the ultimate desktop wind tunnel software. ?se it to design and test comple%

wing systems directly on your Eindows <!. ,t was developed by <atrick $anley, <h.&. for accuracy,

efficiency and ease of use and is by far the 3uickest method on the market for developing wings for

aircraft, water craft and automobiles. MultiSurface is used in the following areas.

@ight aircraft wing design D analysis. This includes kites, powered parachutes, parafoils and other

sporting e#uipment.

?A: design D analysis

Multi*plane wing configurations including 'oined wings

Sailboat(yacht keel and rudder design D analysis

?p*wind sail design D analysis

Ship(sailboat hydrofoil design D analysis

$uman powered hydrofoil design D analysis

Eater sport e#uipment design. This includes sailboard fins and sail configurations.

=#uipment for marine industry such as foils for stabiliing vessels D trawl doors.

Automobile wing, spoilers and surfaces designed for safety and fuel efficiency.

@ectures and presentations in the area of aerodynamics design and analysis.

+nline Tutorials * Demos

Tutorial # $ow to enter and analye a wing. :iew tutorial now.

&emo * $ow to improve a sailboat keel. :iew demo now

&emo * $ow to compare the performance of two keels. :iew demo now.

&emo * ?sing the mirror imaging feature for wing design. :iew demo now.

12amples

&esign and analye 4oined 0ings for ?A: applications. More information ...

?nmanned aerial vehicles design and analysis. More information ...

Sailboat keel D rudder analysis and design. More information ...

Analye hydrofoil with front and rear foils. More information ...

Analye hydrofoils in shallow water. More information ...

Analye an elliptical wing. More information ...

Analye a full*sied sailplane. More information ...

&esign D analye full*sied race car wings and spoilers. More information ...

Analye a full*sied sprint car wing with endplates (pdf file). More information ...

1ducational 5se

,ntroduce engaging e%periments, design content D capstone pro'ects into your lectures D labs. More

information ...

Who 5ses ulti'urface Aerodynamics

MultiSurface Aerodynamics is best suited for users who possess a good understanding of aerodynamics,

hydrodynamics and related engineering principles. Gur customers leverage their design D engineering

knowledge with MultiSurface Aerodynamics to develop outstanding products and services. The following

companies, consultants and universities are among the users of MultiSurface Aerodynamics.

AeroScience

!&= &anish Marine &esign ApS

!lyde /ooth (race car wing engineer)

$ugh Eelbourn (/ritish yacht designer)

,sland =ngineering, ,nc

9imokeo =ngineering

9orea Gcean ;esearch D &evelopment ,nstitute

Mike Mageria (race car wing designer)

Multi$ull <ower

<aradis -autica AS

Sea State

?nited States Air "orce

?niversitat <olitHcnica de !atalunya

?niversity of !alifornia, San &iego

?niversity of =dinburgh

Computer "e3uirements

The perpetual licence of MultiSurface Aerodynamics re#uires a <! running Eindows 70 or @ater. The

leased (online) versions re#uires a <! running Eindows 70 or @ater and a connection to the internet. The

leased software accesses the internet for login purposes only

+6er6ie0

:isual"oil <lus is used by engineers and designers to compute the lift, drag and moment coefficients for

airfoils in subsonic, transonic and supersonic flows. ,t generates accurate aerodynamics data for

e%perimental or e%isting airfoils where the information might be incomplete or unknown. The software can

analye and select airfoils for aircraft, water craft, industrial machinery and other products. :isual"oil<lus

produces graphs of lift, drag and moment coefficient versus angle of attackF surface pressure coefficient,

velocity, Mach number and temperature. The software produces contour plots of pressure, Mach number,

temperature and total pressure. :isual"oil <lus has a large library of built*in airfoils which includes -A!A

5, 0 D 8*digit airfoils. ,n addition, the user can enter custom airfoils for analysis and modify e%isting

airfoil shapes.

Please click here to purchase 7isualFoil Plus.

/enefits

=liminate steep learning curve for new engineers D reduce training costs.

Share airfoil data D results across your entire department

?nderstand and make critical design decision in transonic D supersonic flow regime.

?nderstand the behavior of airfoils used in the design of wings, helicopter blades, hydrofoils, fans,

propellers and other e#uipment.

=asy*to*use tool for designing more efficient wing D machine blade cross*sections.

Eindows based interfaces re#uires little or no training.

8ncompressi&le Flo0 'ol6er

@inear strength vorte% panel method

!oupled boundary layer solver

@ift, drag D moment coefficient calculations

<rediction of transition D separation points

Stall D ma%imum lift prediction

Surface graphs of pressure D velocity ratio

!ontour graphs of !p

Iraphs of !l, !d D !m versus angle of attack

Iraphs of !l vs !d.

Compressi&le Flo0 'ol6er

2*dimensional compressible =uler e#uations using a finite*volume solver

2nd order flu% vector splitting

2nd order flu%*difference splitting

5*stage ;unge*9utta time marching scheme using local time steps

Automatic o*grid generation

!ontour plots of pressure, Mach number, temperature D total pressure

Surface plots of !p, Mach number, velocity ratio D temperature

+ther Features of 7isualFoil Plus

,ncompressible flow solver using vorte% panel method coupled with boundary layer solver

!ompressible flow solver based on =uler =#uations

/uilt*in library of -A!A 5, 0 D 8*digit airfoils

!ustom airfoil analysis

!omputes !l, !d D !m

Stall Model

/uilt*in graphs

/uilt*in tables

=%port tables to .csv files

=%port airfoil to .d%f files

Airfoil plotting

Articles

$igh Speed /lade Screening (.pdf file)

:isual"oil <lus for <ropeller Analysis

Computer 'ystem "e3uirements

:isual"oil <lus is a stand*alone software package that re#uires Eindows 70 or later. Ee recommend a

pentium based <! running Eindows J<. The leased version of :isual"oi <lus also re#uires an internet

connection.

Purchase

)o0 to Purchase 7isualFoil Plus

!ote9 :isual"oil <lus is available e%clusively on hanleyinnovations.com.

7ersion Price Purchase

Lease for Three onths K870.11 /uy !& -ow

Lease for +ne ,ear K4,870.11 /uy !& -ow

Perpetual ('ingle 'eat) :;<==>(%% /uy CD !o0

Perpetual (? 5sers) :@<==>(%% /uy CD !o0

Telephone +rders. <lease call us at (?>;) @AB#CC@@ to place your credit card order by telephone. Ee can

provide you with a download ?;@ D password after the transaction is processed.

Fa2 ,our +rders. <lease click here to fa% your order. (.pdf order form)

+rder &y ail. !lick here to order by mail using a check, money order or credit card.

** The leased versions of VisualFoil Plus are identical to the full perpetual version. nli!e the perpetual

version, however, they re"uire an internet connection to enter a username # password..

!ote9 $ll prices su%&ect to change without notice.

Other Payment Methods

Ee also accept payment by /ank Transfer (SE,"T for e%ample). <lease email us for details.

"e3uest ore 8nformation9

"or more information, please email us at salesLhanleyinnovations.com. Ee can also be reached by

telephone at (302) 8C6*5588. Gur mailing address is $anley ,nnovations, Attn. :isual"oil <lus, <G /o%

C34045, Gcala, "l 355C3*4345.

'ee Also9

Multi=lement Airfoils, MultiSurface Aerodynamics

A&out Dr( )anley

&r. <atrick =. $anley, is the owner and founder of $anley ,nnovations, a small business specialiing in the

development of aerodynamics and fluid dynamics simulation software for education and industry. &r.

$anley earned his /.S. degree (summa cum laude) in aerospace engineering from <olytechnic ,nstitute of

-ew >ork and his S.M. and <h.&. degrees from the department of Aeronautics and Astronautics of

Massachusetts ,nstitute of Technology (M,T). $e also completed a minor in the area of management of

innovation and technology at M,T+s Sloan School of Management.

After graduating from M,T, &r. $anley 'oined the Mechanical =ngineering faculty at the ?niversity of

!onnecticut where he formulated and taught courses in aerodynamics, compressible fluids, introductory

fluid mechanics and heat transfer. As a faculty member, he won the highly competitive -ational Science

"oundation ;esearch ,nitiation Award, the -ASA*AS== Summer "aculty "ellowship and three

consecutive research awards from -ASA @ewis ;esearch center to study compressible viscous flows in

turbomachinery using pseudospectral methods. This research led to the successful education of four (5)

<h.& students and four (5) Masters degree students. ,n addition &r. $anley can be credited with a number

of publications including the pioneering work in multi*domain pseudospectral methods for compressible

viscous flows entitled AA Strategy for the =fficient Simulation of :iscous !ompressible "lows using a

Multi*domain <seudospectral MethodA which can be found in Bournal of !omputational <hysics, :ol 41C,

-o. 4, pp. 403*40C, September 4773.

As owner and chief software author of $anley ,nnovations, &r. $anley has written a number of software

packages including Airfoil/rowser, Airfoil Grganier, Science Iraphs, :isual"oil, Model"oil,

Aerodynamics in <lain =nglish, !enter of Iravity !alculator, EingAnalysis, SmockSoft, <erpetural<aper

amongst other

This is the html version of the file http://www.geodise.org/files/Papers/Man20441.pdf.

G o o g l e automatically generates html versions of documents as we crawl the we.

To lin! to or oo!mar! this page" use the following url: http://www.google.com/search?q=cache:IAhUq6ly_CIJ:www.geodise.org/files/Papers/a!"#$$%.pdf&'o(r!als&

)performa!ce&of&differe!t&airfoils&i!&cfd*+hl=e!+gl=i!+ct=cl!,+cd=%6

'oogle is neither affiliated with the authors of this page nor responsi%le for its content.

These search terms have een highlighted: performance different airfoils cfd

These terms only appear in lin!s pointing to this page: journals

Page 1

A Study of Shape Parameterisation Methods for

Airfoil Optimisation

Wenbin Song

This paper presents a study on parameterisation methods for airfoil shape optimisation

within a CAD-based design optimisation framework. The objective of the paper is to study

the effect of different methods on airfoil shape optimisation when using computational fluid

dynamics (CFD). Parameterisation of geometry is one of the essential requirements in shape

optimisation, and it presents further challenges when carrying out multidisciplinary design

optimisation, as it is critically important to maintain shape consistency between analysis

domains, while providing different analysis models from the same CAD definition. It is

usually the case that there are numerous possibilities in defining the parametric model,

and it will prescribe to a large extent the scope of the search space and landscape of the

objective function. This paper adopts design of experiments and optimisation approaches to

study several representative parameterisation methods in terms of flexibility and accuracy

of the methods for aerodynamic shape optimisation.

I. Introduction

A

ircraft design is a complex decision making process and according to Raymer,

1

can usually be broken

down into three phases, conceptual design, preliminary design, and detailed design. Aerodynamic design

occurs throughout these steps. Two different approaches are often employed in the aerodynamic design: 1)

inverse design and 2) direct numerical optimisation. The first method tries to solve for a geometry that

produces a prescribed pressure distribution. On the other hand, direct numerical optimisation methods

couple a geometry definition and aerodynamic analysis code in an iterative process to produce optimum

designs subject to various constraints. Depending on whether the goal is to improve on an existing design or

to create a completely new design, different parameterisation methods are often required. If the new design

only requires small changes to the initial geometry, a localized parameterisation approach is often used. But

when conducting a study of a radically new concept, the parameterisation method needs to accommodate a

wider range of new shapes.

Airfoils have been represented in a number of different ways in the past. For example, coordinates have

been directly used to fit airfoil shapes using B-splines and Bzier curves

2

via interpolation methods. Analytical

functions have also been derived to represent families of airfoils, for example, in the work reported by Hicks

and Henne.

3

In a more recent work,

4

Non-uniform rational B-splines (NURBS) were used first to approximate

existing airfoils, then adopted as a general parameterisation method to be used in optimisation. The concept

of using relatively few orthogonal functions to represent a large number of functions has also been exploited,

for example in a work reported by Robinson and Keane,

5

where a set of orthogonal functions was developed

using numerical methods. These functions were then used to represent a family of airfoils in a wing design

study. However, the basis functions derived by Robinson and Keane

5

were believed to be dependent on the

particular familiar of airfoils. Although these numerically derived basis functions can be used in the design

of a particular set of airfoils, other airfoils may not be adequately represented using them.

The choice of parameterisation method, when coupled with optimisation techniques to find desirable

shapes in terms of user-defined objective functions and constraints, has a major effect on the final results,

efficiency and effectiveness of particular search strategies. Giving the same CFD models, the parameterisation

Research Fellow, School of Engineering Sciences, University of Southampton, Highfield, SO17 1BJ, United Kingdom, and

AIAA Member.

Professor, School of Engineering Sciences, University of Southampton, Highfield, SO17 1BJ, United Kingdom.

1 of 8

American Institute of Aeronautics and Astronautics

Page 2

effectively defines the optimisation problem formulation, the topology of the design space, and the landscape

of the objective functions. Although it is vitally important, it is also very difficult to come up with a set of

effective criteria that can be readily used to evaluate the pros and cons of different parameterisation methods.

Wu

6

compared three geometric representations in three case studies of cascade blade design using adjoint

methods. After carrying out the optimisation, it has concluded that one of the methods using geometry

parameters is not suitable for two occasions.

There are a number of key issues that need to be addressed in the choise of parameterisation methods.

The first issue is the flexibility of any parameterisation method. Flexibility is interpreted here as the ability

to represent a wide range of different shapes. Some parameterisation methods, for example, coordinate-

based methods, can accurately represent a variety of dramatically different shapes and can also reflect subtle

changes in local areas, however it would be very difficult to use such an approach for optimisation problems

using high fidelity codes due to the large number of design variables and complexity of the design space. On

the other hand, methods using fewer variables may not be capable of generating shapes with high accuracy,

especially when used in inverse design problems where a target pressure distribution is sought. The second

issue when considering parameterisation methods is the accuracy or the optimum objective functions that

the final shape can achieve either in an inverse design study or direct optimisation work, respectively. The

accuracy should be measured in both geometric and aerodynamic senses. However, the optimal objective

function cannot be obtained without actually carrying out the optimisation, therefore, here an inverse design

approach is adopted to compare different parameterisations.

In this work, three datum airfoils, two from the NACA supercritical airfoils family (NACA0406 and

NACA0610) and the third being the RAE2822, are used as reference shapes to compare different param-

eterisation methods for airfoil design. The paper is organised as follows. Section two describes different

parameterisation methods for airfoil design. The geometry modelling and flow analysis of the airfoil prob-

lems are described in section three. Results and discussions are presented in section four, with conclusions

given in section five.

II. Parameterisation of Airfoil Geometry

Figure 1. The first three numerically derived or-

thogonal basis functions.

Geometry parameterisation methods have attracted

renewed interests in recent years, especially in the con-

text of multidisciplinary design optimisation (MDO).

Samareh

7

identified three categories of parameterisation

methods in the context of MDO. These include the dis-

crete approach, CAD-based approaches, and free-form de-

formation methods. Indeed, all these different approaches

could be implemented in most modern CAD systems.

Several parameterisation methods have been proposed

in previous papers for airfoil geometry, for example, the

NACA supercritical airfoils are defined as a series of y-

coordinates at prescribed chord wise locations.

8

The sec-

ond approach models the geometry as the linear combina-

tion of a basis airfoil and a set of perturbation functions,

defined either analytically

4

or numerically,

5

as shown in

Eq. (1). The coefficients of the perturbation functions in-

volved are then considered as the design variables. A set

of such orthogonal basis functions derived from a group

of base airfoils was developed by Robinson and Keane

5

to

provide an efficient means to define the airfoil for optimi-

sation study in preliminary design, for example.

y(x) =

0

y

0

(x) +

w

i

f

i

(x)

(1)

A third, and more geometrically intuitive method, is to use geometric parameters such as leading edge

radius, thickness-to-chord ratio or maximum thickness to define the airfoil shape. An airfoil parameterisation

using 11 geometry parameters was presented by Sobieczky

9

and used by Oyama etc.

10

A fourth method uses

the control points of Non-uniform Rational B-splines (NURBS) curves to define the airfoils.

4

This method is

2 of 8

American Institute of Aeronautics and Astronautics

Page 3

also used by Li

11, 12

in which a B-pline interpolation through 35 points is used to define the airfoil geometry.

The advantage of this approach is that free-form geometrical shapes can be accommodated with fewer design

variables compared to the direct use of coordinates. However, due to difficulties in controlling the relative

positions of the control points, free-form parameterisations are usually used in the inverse design approach,

where only a subset of control points are allowed to change in a relatively small range to meet a target

pressure distribution.

One of the key issues in deciding the parameterisation method is the balance the requirements of ro-

bustness and flexibility, and these decisions are also strongly dependent on the goal of the design activity.

Although free-form parameterisations may well be able to generate radical new shapes, this is not suitable

for designs where the aim is to meet a specific pressure distribution, due to the poor efficiency caused by the

large search space that arises in the optimisation process. Another disadvantage of free form parameterisa-

tion is the inherent difficulties encountered when trying to generate airfoil-like shapes: Usually additional

geometrical constraints need to be imposed. Two different parameterisation methods are implemented in

the current work to compare their effectiveness. The first approach is to use a set of numerically derived

basis function to define the airfoil,

5

in this case, only a small number of design variables are involved. The

basis functions used are illustrated in Figure 1. The second approach uses a B-spline interpolation based on

34 points as shown in Figure 2. The y coordinates of the points are used as design variables while the chord

wise coordinates of the points are fixed. Both methods are implemented in the CAD system ProEngineer.

13

Airfoil shapes from the family of supercritical airfoils

8

and RAE2822 are chosen in the current work as ref-

erence airfoils in the comparison. The number of parameters involved in the two parameterisation methods

are summarized in Table 1.

Table 1. Summary of different parameter methods for airfoils

Method

Number of parameters Description of the parameters

Numerical Basis Functions

5

Weights for the basis airfoil functions

B-spline interpolation

34

Point coordinates

III. Geometry Modelling and Flow Analysis

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.03

-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

0.015

0.02

0.025

0.03

Figure 2. Airfoil Modelling using B-spline inter-

polation.

To compare the flexibilities of different parameterisa-

tion methods in representing different airfoil shapes, three

existing airfoils NACA0406, NACA0610, and RAE4822

are chosen as the modelling targets for these parameter-

isations. The difference between the target airfoil and

approximated airfoil is defined as

diff =

abs(f

target

(x

i

) f(x

i

))

(2)

where x

i

(i = 1, ..., n) are the chordwise coordinates used

in the definition of the target airfoils. This quantity is

minimised using a global optimisation algorithm for dif-

ferent parameterisation methods and then used as an in-

dication of the flexibility of the methods. To remove the

effect of the optimisation techniques that may not pro-

duce the optimum results, a large number of iterations

has been carried out for the minimisation problem.

The airfoil models are all implemented using ProEngi-

neer and exported in the form of a STEP file, which is

then imported into Gambit. A boundary layer is attached

to the lower and upper surface of the airfoils, and size

functions are also used to give better control of the mesh and to reduce the computational time for the

problem. An unstructured mesh generated using Gambit

14

for solving the N-S equations is shown in Figure

3. The mesh contains 11356 cells (compared with 87305 cells without using the size function). In both cases,

3 of 8

American Institute of Aeronautics and Astronautics

Page 4

the node spacing on the airfoil surfaces and farfield circle are the same, with size functions giving better

control of the transition of size of the cells in between. The computation time is reduced from around 40

minutes to less than 20 minutes for most geometries on a Xeon 2.4Ghz compute node with 1Gb memory.

The flow model used in the current work is based on the Navier-Stoke model from Fluent.

14

The

pressure distribution of the upper and lower surfaces are used in the comparison. Here, the cruise condition

(M

= 0.73) is used when calculating the lift and drag values. The Spalart-Allmaras viscosity model is

used.

IV. Results and Analysis

Figure 3. Unstructured mesh used for solving the N-S

equations by Fluent

It is not straightforward to compare alternative

parameterisation methods. There are two impor-

tant considerations when a parameterisation model

is built around an existing geometry: the first is

the flexibility of the model, i.e., how many different

shapes can this model represent. The second is be

the robustness of the model, i.e., can the model gen-

erates the desired shape for large number of different

designs. In general models with more design vari-

ables will be able to represent more complex shapes,

and will be more likely to produce novel designs us-

ing optimisation. However, that will be more expen-

sive in the search as the design space will have higher

dimensions, and chances of failure or not generating

desired shapes will be higher.

The best results for approximations of the tar-

get airfoils are shown in Table 2. The results are

produced by minimising the objective function computed using (2). A genetic algorithm (GA) from OP-

TIONS

15

is used in the current work, however, the first population is not generated randomly, rather, it

is generated using a Design of Experiment (DoE) method plus one point describing a user specified base

design to obtain more uniform coverage of the design space as well as to provide the best possible guess.

The DoE method used here is a Latin Hyper Cube method. The base design specified by the user can play

an important role in accelerating the search process as the GA used in this work always maintains the best

solution in the population. A single base design is used in the orthogonal basis function approach for all

three target airfoils and in the B-spline approach for RAE2822 approximation. The two base designs used

in B-spline interpolation for NACA0406 and NACA0610 are NACA0403 and NACA0606, respectively.

It can be seen that the approximations using orthogonal basis function can produce better results for

the NACA supercritical airfoils NACA0406 and NACA0610 than for the RAE2822, this is not surprising,

as the set of basis functions used were derived from a family of airfoils containing these two. The errors are

believed to be caused by the smoothing process in the derivation of these basis functions.

5

For NACA0406

and NACA0610, the orthogonal basis functions also produce better results than the B-spline interpolation

approach, this is because more variables are used in the B-spline interpolation and so it would be much more

expensive to obtain the optimum if a comparable number of iterations were used for both cases. However, a

different story arises for RAE2822. Since it was not included in the process of deriving the basis functions,

this leads to greater error when compared with the B-spline interpolation approach. This indicates the wider

applicability of the B-spline approach, at the higher cost of reaching the optimum.

Table 2. Best Approximation of target airfoils for different parameterisations

Method

RAE2822 NACA0406 NACA0610

Numerical Basis Functions

0.2217

0.0582

0.1595

B-spline interpolation

0.1552

0.0758

0.1993

However, similarity in geometry does not always guarantee similar pressure distribution, especially for

4 of 8

American Institute of Aeronautics and Astronautics

Page 5

transonic flows where small perturbations in shape will lead to large variations in pressure. Therefore, the

pressure distributions of the approximated shape are also compared to that of the original airfoils, as shown

in Figure 4. It can be seen from Figure 4 that the orthogonal basis function approach always produces

smooth pressure distributions; also errors are generally bigger in the leading and trailing edge areas than in

the middle section of the airfoils.

Figure 5 shows the results of approximation using the B-spline interpolation approach. It can be seen that

close agreement can be achieved using B-spline interpolation through 34 points apart from the leading edge

area, which indicates that more points need to be placed within this area to achieve better results. Moreover

the chordwise coordinates can also be varied, but this would involve higher computational cost while not

necessarily increasing the accuracy of the approximation. Another advantage of the B-spline approach is its

ability to carry out local shape tunning by varying a subset of the coordinates, while it would be difficult

to perform this with the basis function approach, in which, any changes in the coefficients will change the

shape globally.

The approach adopted in the current work is essentially an inverse design method. However, it is not

used to seek a prescribed pressure distribution, as the definition of the pressure distribution itself is a design

problem and accurately re-generating the prescribed pressure distribution often leads to degradation of

performance in other conditions. This method can be used to evaluate different parameterisations before

carrying out optimisations using the high fidelity codes.

V. Conclusions

Two airfoil parameterisation approaches are studied in this paper to analyse their flexibility and robust-

ness in producing optimal shapes when used in optimisation studies. Global optimisation methods are used

to analysis the accuracy these two parameterisations can achieve when used to model three target airfoils.

The B-spline approach produces better results in terms of accuracy at a higher computational cost while

the basis function approach is more efficient while producing less accurate results. Further work will in-

volve the combination of the basis function approach in the initial stages of design combined with B-spline

interpolation for the final tuning of the shapes.

Acknowledgement

The work described here is supported by the UK e-Science Pilot project: Grid-Enabled Optimisation and

Design Search for Engineering (Geodise) (UK EPSRC GR/R67705/01). Financial support from EPSRC is

greatly acknowledged.

References

1

Raymer, D.P., Aircraft Design: A Conceptual Approach, AIAA Educational Series 1999.

2

Cosentino, G. B. and Holst, T. L., Numerical Optimisation Design of Advanced Transonic Wing Configurations, May,

1984.

3

Hicks, R. M. and Henne, P. A., Wing Design by Numerical Optimisation, Journal of Aircraft, Vol. 15, No. 7, 1978, pp.

407-413.

4

Lepine, J., Guibault, F., Trepanier, J.-Y., and Pepin, F., Optimized Nonuniform Rational B-spline Geometrical Repre-

sentation for Aerodynamic Design of Wings, AIAA Journal, Vol. 39, No. 11, 2001.

5

Robinson, G. M. and Keane, A. J., Concise Orthogonal Representation of Supercritical Airfoils, Journal of Aircraft, Vol.

38, No. 3, 2001, pp. 580-583.

6

Wu, H.Y., Yang, S.C., Liu, F. and Tsai, H.M., Comparison of Three Geometric Representations of Airfoils for Aero-

dynamic Optimization, AIAA 2003-4095, 16th AIAA Computational Fluid Dynamics Conference, June 23-26, Orlando, FL,

2003

7

Samareh, J. A., Survey of shape parameterization techniques for high-fidelity multidisciplinary shape optimisation, AIAA

Journal, Vol. 39, No. 5, 2001, pp. 877-884.

8

Charles D.H., NASA Supercritical Airfoils, A Matrix of Family-Related Airfoils, NASA Technical Paper 2969, Mar. 1990.

9

Sobieczky, H., Parametric Airfoils and Wings, Notes on Numerical Fluid Mechanics, edited by K. Fujii and G. S. Du-

likravich, Vol. 68, Vieweg Verlag, 1998, pp. 71-88.

10

Oyama, A., Obayashi, S., and Nakamura, T., Real-coded Adaptive Range Genetic Algorithm Applied to Transonic Wing

Optimisation, Springer, Paris, France, 2000, pp. 712-721.

11

Li, W., Huyse, L., and Padula, S., Robust airfoil optimisation to achieve drag reduction over a range of Mach numbers,

Structural and Multidisciplinary Optimisaiton, Vol. 24, 2002, pp. 38-50.

5 of 8

American Institute of Aeronautics and Astronautics

Page 6

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

0.06

RAE2822

Orthfoil-Approximation

(a) Approximation of geometry to RAE2822 by basis functions

approach

X/C

P

r

e

s

s

u

r

e

C

o

e

f

f

i

c

i

e

n

t

s

0

0.25

0.5

0.75

1

-1

-0.5

0

0.5

1

RAE2822

Orthfoil-approximation

(b) Comparison of pressure distributions for RAE2822 using

basis functions

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

NACA0406

Orthfoil-Approximation

(c) Approximation of geometry to NACA0406 by basis func-

tions approach

X/C

P

r

e

s

s

u

r

e

C

o

e

f

f

i

c

i

e

n

t

s

0

0.25

0.5

0.75

1

-1

-0.5

0

0.5

1

NAC0406

Orthfoil-approximation

(d) Comparison of pressure distributions for NACA0406 using

basis functions

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

NACA0610

Orthfoil-Approximation

(e) Approximation of geometry to NACA0610 by basis func-

tions approach

X/C

P

r

e

s

s

u

r

e

C

o

e

f

f

i

c

i

e

n

t

s

0

0.25

0.5

0.75

1

-1

-0.5

0

0.5

1

NACA0610

Orthfoil-approximation

(f) Comparison of pressure distributions for NACA0610 using

basis functions

Figure 4.

Comparisons of geometry and pressure distributions for approximations using orthogonal basis

functions

6 of 8

American Institute of Aeronautics and Astronautics

Page 7

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

0.06

RAE2822

B-Spline interpolation

(a) Approximation of geometry to RAE2822 by B-spline in-

terpolation

X/C

P

r

e

s

s

u

r

e

C

o

e

f

f

i

c

i

e

n

t

s

0

0.25

0.5

0.75

1

-1

-0.5

0

0.5

1

RAE2822

B-Spline Interpolation

(b) Comparison of pressure distributions for RAE2822 using

B-spline interpolation

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

NACA0406

B-Spline Interpolation

(c) Approximation of geometry to NACA0406 by B-spline in-

terpolation

X/C

P

r

e

s

s

u

r

e

C

o

e

f

f

i

c

i

e

n

t

s

0

0.25

0.5

0.75

1

-1

-0.5

0

0.5

1

NAC0406

B-Spline Interpolation

(d) Comparison of pressure distributions for NACA0406 using

B-spline interpolation

X/C

Y

/

C

0

0.25

0.5

0.75

1

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

NACA0610

B-Spline Interpolation

(e) Approximation of geometry to NACA0610 by B-spline in-

terpolation

X/C

P

r

e

s

s

u

r

e

C

o

e

f

f

i

c

i

e

n

t

s

0

0.25

0.5

0.75

1

-1

-0.5

0

0.5

1

NACA0610

B-Spline Interpolation

(f) Comparison of pressure distributions for NACA0610 using

B-spline interpolation

Figure 5. Comparisons of geometry and pressure distributions for approximations using B-spline interpolations

7 of 8

American Institute of Aeronautics and Astronautics

Page 8

12

Li, W., Profile Optimisation Method for Robust Airfoil Optimisation in Viscous Flow, NASA /TM-2003-212408, 2003.

13

ProEngineer, http://www.ptc.com, 2004.

14

Fluent, http://www.fluent.com, 2004.

15

Keane, A.J., OPTIONS Design Exploration System, http://www.soton.ac.uk/ajk, 2004

8 of 8

American Institute of Aeronautics and Astronautics

This is the html version of the file http://aero#comla.stanford.edu/Papers/!yoto$4.pdf.

G o o g l e automatically generates html versions of documents as we crawl the we.

To lin! to or oo!mar! this page" use the following url: http://www.google.com/search?q=cache:w!-alq./CowJ:aero-

comla/.sta!ford.ed(/Papers/,yoto0$.pdf&'o(r!als&)performa!ce&of&differe!t&airfoils&i!&cfd*+hl=e!+gl=i!+ct=cl!,+cd="#

'oogle is neither affiliated with the authors of this page nor responsi%le for its content.

These search terms have een highlighted: performance different airfoils cfd

These terms only appear in lin!s pointing to this page: journals

Page 1

A Two-Dimensional Multigrid-Driven Navier-Stokes

Solver for Multiprocessor Architectures

Juan J. Alonso, Todd J. Mitty, Luigi Martinelli, and Antony Jameson

Department of Mechanical and Aerospace Engineering

Princeton University

Princeton, New Jersey 08544 U.S.A.

Abstract

A two-dimensional unsteady Navier-Stokes solver has been parallelized using a domain decomposition

approach and the PVM message passing library. Several options for the treatment of multigrid and

implicit residual smoothing are examined. Results for the unsteady flow over a pitching NACA 64A010

airfoil are presented.

1 INTRODUCTION

In recent years, computational fluid dynamics (CFD) has been gaining acceptance as a design tool in industry.

Advancements in algorithm development and computational hardware have led to more complex modeling

of fluid flows. Although current inviscid models can accurately predict the coefficient of lift for an airfoil

in transonic flow, viscous effects such as shock wave/boundary layer interaction can significantly modify

important aspects ofa flow. Furthermore, unsteady phenomena ofpatent viscous character, such as buffeting,

can not be predicted with the help ofinviscid models. Such viscous phenomena directly impact the design

ofengineering configurations, and therefore, it is necessary to enhance the viscous prediction capability of

CFD tools.

Increasingly complex fluid flow models require high performance computing facilities. A cost effective

solution for problems of this type requiring fast CPUs and large internal memory is the use of a parallel

computing paradigm. For computational efficiency, one typically incorporates convergence acceleration tech-

niques such as multigrid and implicit residual smoothing. Message passing becomes necessary in this new

environment, and severely limits the performance of processes that are inherently communication intensive.

In this paper we present a parallelized version ofa well established Navier-Stokes solver, FLO103 [1].

This computer program has recently been enhanced to include Jamesons implicit multigrid approach [2] for

the efficient calculation ofunsteady viscous flows. Calculations are performed on an IBM SP1 multiproces-

sor computer with a domain decomposition approach, and message passing is handled by PVM (Parallel

Virtual Machine) software [3]. Several methods for implementing convergence acceleration techniques such

as multigrid and implicit residual smoothing are studied.

2 NAVIER-STOKES EQUATIONS DISCRETIZATION

The two-dimensional, unsteady, compressible Navier-Stokes equations may be written in divergence form for

a Cartesian coordinate system (x, y) as

w

t

+

f

x

+

g

y

=

R

x

+

S

y

,

(1)

where w is the vector offlow variables, f and g are the convective fluxes, and R and S are the viscous fluxes

in each ofthe coordinate directions. With Reynolds averaging, turbulence effects can be taken into account

with a turbulence model. In this work, a Baldwin-Lomax model was used. In integral form, Equation 1 can

1

Page 2

be applied to each finite volume ofa computational domain to yield a set ofcoupled first-order differential

equations ofthe form

d

dt

(w

ij

V

ij

) + E(w

ij

) + NS(w

ij

) + D(w

ij

) = 0

,

(2)

where E(w

ij

) are the convective Euler fluxes, NS(w

ij

) are the Navier-Stokes viscous fluxes, and D(w

ij

) are

the artificial dissipation fluxes added for numerical stability. For unsteady problems, Equation 2 is modified

by introducing a pseudo-time formulation to improve computational performance [2].

3 PARALLELIZATION STRATEGY

FLO103P is parallelized using a domain decomposition model, a SIMD (Single Input Multiple Data) strategy,

and the PVM Library for message passing. Flows were computed on a C-mesh of size n

i

n

j

= 102464. This

domain was decomposed into subdomains containing

n

i

N

p

n

j

points, where N

p

is the number ofsubdomains

used. Communication between subdomains was performed through halo cells surrounding each subdomain

boundary. A two-level halo was sufficient to calculate the convective, viscous, and dissipative fluxes for all

cells contained in each processor. In the coarser levels ofthe multigrid sequence, a single level halo suffices

since a simplified model ofthe artificial dissipation terms is used.

For problems with a low task granularity (ratio ofthe number ofbytes received by a processor to the

number of floating point operations it performs), large parallel efficiencies can be obtained. Unfortunately,

convergence acceleration techniques developed in the 1980s base their success on global communication in

the computational domain. Thus, current multigrid and implicit residual smoothing techniques [4] are bound

to hinder parallel performance in traditional mesh sizes. In order to effectively deal with the parallelization

ofthese two techniques, we propose several ideas.

In this paper, static load balancing is performed at the beginning of the calculation. Since the number

ofcells remains constant throughout the calculations, the domain can be partitioned into subdomains with

an equal number ofcells. Except for some domains where an additional message across the wake is required,

this provides for perfect load balancing.

4 PARALLEL MULTIGRID

The full approximation multigrid technique enhances the convergence rate of a scheme by performing com-

putations on a series ofincreasingly coarser meshes. The calculations performed in each ofthese meshes are

driven by the residuals at the previous finer level, and the results obtained are interpolated to the corre-

sponding finer mesh. The explicit time-stepping acts as a smoothing operator for the high frequency errors

in each level, thus damping the dominant error mode at the finest level and accelerating convergence. A

more detailed discussion of this procedure can be found in [4]. If domain decomposition is used for parallel

calculations, the size ofthe meshes contained in each processor at the coarser levels ofthe multigrid cycle

is quite small (typically 8 or 16 cells). Most ofthe CPU time is then wasted sending information back and

forth between processors. This situation worsens for multigrid W-cycles, where equilibrium in the coarser

meshes is established repeatedly before traversing the series back to the finest level. Previous authors [5]

have attempted to deal with this problem in the Euler equations by limiting message passing to only some

stages in the Runge-Kutta time-stepping scheme, or passing the boundary data exclusively at the finest level

in the series. This usually led to a decrease in the convergence rate ofthe numerical algorithm, presenting a

clear tradeoff between the improved parallel performance and the increasing number of cycles required for a

similar level ofconvergence. In this work, we examine three different approaches to the implementation ofthe

multigrid algorithm. First, the full multigrid algorithm is implemented with message passing at all required

points such that the parallel program exactly recovers the results ofthe serial code. Second, multigrid is

applied independently within each subdomain to completely avoid inter-processor communication. Third,

and last, at coarse multigrid levels where communication overhead dominates CPU time, a single processor

will be used to gather, compute, and scatter information. Parallel speed-up curves for these three cases are

also presented.

2

Page 3

"DUMB" PROCESSOR W-CYCLE

Parallel

T

T

T

T

T

T

T

Fine

Coarse

Single Proc.

"LAZY" PROCESSOR W-CYCLE

Parallel

T

T

T

T

T

T

T

Fine

Coarse

No work

Figure 1: Lazy-Dumb Multigrid Approach.

4.1 Full multigrid approximation

This first approach was an attempt to implement a parallel program that exactly reproduced the output of

the single processor code in which message passing is not necessary. In order to achieve this goal, boundary

information was passed among processors on all multigrid levels at the beginning of all five stages of the

Runge-Kutta time-stepping. Additional messages were required in order to process convergence information,

calculate force coefficients, and compute the eddy viscosity in the turbulence model. This procedure recovers

the serial version convergence rate, at the expense ofpoor parallel performance.

4.2 Implicit multigrid within subdomains

The second approach used to deal with the multigrid algorithm was to completely decouple the subdomains

on the coarser levels ofthe series in order to minimize the number ofmessages passed when very little

computation is being performed. As expected, this approach restores parallel efficiency to higher levels, but

it suffers from a decreased convergence rate. Clearly, the success of the multigrid technique is due to the

increased communication between different parts ofthe computational mesh at all levels, and the transfer of

this information is limited by the isolation of the coarser levels in each subdomain. The degradation of the

convergence rate is especially bad when interdomain boundaries lie close to regions ofhigh gradients (such

as shock waves and stagnation points). In some ofthese cases, a converged solution cannot be obtained.

4.3 Lazy-Dumb multigrid approach

In order to preserve the convergence rate ofthe original multigrid algorithm, and somehow improve the

parallel efficiency ofthe first approach we propose the following procedure: calculations on the finer levels are

performed as in 4.1; when the algorithm shifts the solution to a user-specified coarse mesh, all the processors

in the calculation pass their flow variables and grid locations to a single processor which computes the coarser

levels of multigrid without the need for message passing. This processor is termed dumb since it performs

everyone elses work. The rest ofthe processors in the computation wait until the calculation needs to be

performed on the finer levels, at which point they receive the information from the dumb processor and

proceed once more in parallel. These processors are called lazy since they avoid carrying out part ofthe

work that was, in theory, assigned to them. In our program, the level at which the transfer is done can be

specified as an input, allowing for investigations of the optimal location of this transition point. Figure 1

presents this procedure graphically. Note that this construction can be extended to a hierarchy ofdumb

processors for larger calculations involving more processors.

3

Page 4

5 IMPLICIT RESIDUAL SMOOTHING

Implicit residual smoothing (or averaging) is a technique that couples the residuals at any given point with

those ofall the other cells in the domain, increasing the support ofthe scheme, and allowing a larger time step

than that permitted by the Courant-Friedrichs-Lewy (CFL) restriction. Once more, this is a communication

intensive procedure that negatively impacts parallel performance in a multiprocessor environment.

5.1 Fully implicit residual smoothing

Serial implementations ofthis technique usually split the problem into the implicit coupling along each of

the two coordinate directions. The direction which is normal to the airfoil surface presents no difficulties

since all the required data resides in the appropriate processors (in this decomposition), allowing calculations

to proceed in parallel. In the coordinate direction that is parallel to the airfoil surface, the residuals that

are to be coupled reside in different processors. Moreover, the solution procedure (Thomas algorithm for a

tridiagonal system) is inherently serial. A brute force method allows only one processor to be active at a

time in the forward elimination and back substitution phases, while the remaining processors are idle. Since

the residual smoothing procedure consumes about 30% ofthe time ofthe total calculation, this idle time

causes parallel performance to drop-off considerably.

5.2 Implicit residual smoothing within subdomains

In a very similar fashion to the multigrid performed implicitly within blocks, and in an attempt to reduce

the amount ofmessages passed, residual smoothing was performed implicitly within blocks, with no global

coupling ofthe residuals. Once more, as expected, although parallel performance improves, the convergence

rate degrades to unacceptable levels. As in the corresponding multigrid procedure, this lack ofcoupling

between domains often led to instability in the calculations.

5.3 Iterative implicit residual smoothing

An alternative to the previous approach is to perform an iterative solution of the smoothing problem.

Messages containing the boundary residuals are passed at the beginning ofeach iteration, and the relaxation

process proceeds in parallel. It has been observed in practice that in order to obtain the increase in CFL

number provided by the fully implicit version (from 3 to about 6), at least two iterations are required.

In these calculations, three iterations were performed and a CFL number of 6 was used without stability

problems. The matrix problem is setup in each subdomain and a Jacobi relaxation procedure is performed

on all subdomains concurrently.

6 RESULTS

A summary of the results for the different multigrid approaches is presented in Figure 2. The full multigrid

approach produces results with an optimum convergence rate, but with a parallel performance that degrades

for a large number of processors. One must take into account, that for these meshes (n

i

n

j

= 1024 64)

granularity becomes too large as the number ofprocessors increases. This makes efficient parallel calculations

not viable for a number of processors larger than 8. The multigrid performed implicitly within blocks (see

Section 4.2) exhibits better parallel performance, but at a very high cost. When the total cost to achieve

a solution converged to the same level through these two procedures is computed, the second technique

requires about twice the number ofcycles making this approach less than desirable. Finally, the lazy-

dumb version ofthe algorithm performs slightly better when the dumb processor computes the two

coarsest levels ofthe sequence. When the three coarsest levels are assigned to the dumb processor, parallel

performance degrades since the amount of time required by the dumb processor exceeds that needed for

all processors (with poor parallel performance). Notice that wall time was used to compute the parallel

efficiencies in all cases. Ifthe unused CPU time ofthe lazy processors were to be factored in, the gain

would be larger. It must also be mentioned that the implementation ofthis approach introduces a higher

level ofcomplexity to the multigrid algorithm, since the memory management on both types ofnodes differs

4

Page 5

considerably from the traditional one. As a payoff for the additional work, this approach exhibits the exact

same rate ofconvergence as the full multigrid algorithm.

Evaluating the performance of multigrid in meshes that are relatively small places a strong restriction

on the parallel efficiencies that can be achieved. Calculations using very large meshes (three-dimensional

calculations, two-dimensional LES calculations, etc) will benefit from the full multigrid algorithm and will

achieve parallel performances on the order of 90% since most of the time will be spent on the finer levels of

the sequence. Nevertheless, for engineering calculations, reasonable speed-ups can be obtained through this

method.

The performance results of the different methods used to implement residual smoothing are presented in

Figure 3. The fully implicit approach is clearly unacceptable even for a small number of processors. While

the approach that uses implicit residual smoothing within blocks exhibits a parallel performance of 96% for

8 processors, the degradation ofthe convergence rate disqualifies it as a possible candidate. Finally, the

iterated residual smoothing preserves both the convergence rate and a reasonable parallel speed-up, and is

chosen as the candidate for engineering calculations.

Figure 4 presents a detail of the domain decomposition for a calculation involving four processors for a

typical airfoil section. Figure 5 shows the pressure contours at different points on the oscillation period of

a NACA 64A010 airfoil at a Reynolds number of 10

6

, M

c

= 0.202.

These results were obtained with the O-mesh version ofthe program. One can clearly see how the shock

waves strengthen and weaken as the airfoil pitches up and down.

The lines that intersect the pressure contour lines are the interprocessor boundaries ofthe O-mesh.

Perfect continuity ofthese contour lines validates the accuracy ofthe code even for the cases where strong

shocks are close to the interprocessor boundaries. For more details, please refer to [6].

7 CONCLUSIONS

A parallelized, two-dimensional, unsteady Navier-Stokes flow solver has been developed. Parallelization

was realized using a domain decomposition approach to achieve proper load balancing and computational

efficiency. PVM communication software was used for message passing between processors. Strategies for

dealing with two convergence acceleration techniques, namely implicit residual smoothing and multigrid, and

the performance ofeach ofthese techniques have been evaluated.

It is observed that for meshes of sizes typically used in engineering calculations, acceptable parallel

performances can be achieved with up to 8 processors. A larger number of processors is not suitable for this

type ofcalculation. For larger meshes, the multigrid technique is still quite favorable even in multiprocessor

architectures. Implicit residual smoothing can be performed in an iterative fashion without an observable

impact in the convergence rate while retaining good parallel performance. All calculations used the public

distribution ofthe PVM software developed at Oak Ridge National Labs. Preliminary results with the IBM

optimized version ofthis message passing standard (PVMe), on both the SP1 and SP2 platforms, confirm

the expected trends in which parallel performances improve considerably. Finally, results for the unsteady

pressure field around a pitching NACA 64A010 at M

References

[1] Martinelli, L., Jameson, A., Validation ofa Multigrid Method for the Reynolds Averaged Equations,

AIAA Paper 88-0414, AIAA 26th Aerospace Sciences Meeting, Reno, January 1988.

[2] Jameson, A., Time Dependent Calculations Using Multigrid, with Applications to Unsteady Flows

Past Airfoils and Wings, AIAA Paper 91-1546, AIAA 10th Computational Fluid Dynamics Conference,

Honolulu, June 1991.

[3] Geist, A., Beguelin A., Dongarra J., Jiang W., Manchek, R., Sunderam V., PVM 3 Users Guide and

Reference Manual, Oak Ridge National Laboratory, May 1993.

5

Page 6

1

2

3

4

5

6

7

8

1

2

3

4

5

6

7

8

Number of Processors

Parallel Speedup

Comparison of Parallel Speedups of Multigrid Methods

solid Full multigrid

Implicit within blocks

.. LazyDumb 2 levels in dumb processor

..... LazyDumb 3 levels in dumb processor

Figure 2: Summary ofParallel Speed-ups for Differ-

ent Approaches to the Multigrid Technique.

1

2

3

4

5

6

7

8

1

2

3

4

5

6

7

8

Number of Processors

Parallel Speedup

Comparison of Parallel Speedups of Residual Smoothing Methods

solid fully implicit

.. implicit within blocks

explicit iteration

Figure 3: Summary ofParallel Speed-ups for Differ-

ent Implicit Residual Smoothing Approaches.

129x6 5

129x6 5

129x6 5

129x6 5

Figure 4: Domain Decomposition for a NACA 0012 Airfoil at a 10 degree Angle of Attack.

[4] Jameson, A., Transonic Flow Calculations, Princeton University Report 1651, March 1984, in Numerical

Methods in Fluid Dynamics, edited by F. Brezzi, Lecture Notes in Mathematics, Vol. 1127, Springer-

Verlag, 1985, pp. 156-242.

[5] Yadlin, Y. and Caughey, D. A., Block Multigrid Implicit Solution ofthe Euler Equations ofCompressible

Fluid Flow, AIAA Journal, 29(5):712-719.

[6] Alonso, J. J., Martinelli, L., and Jameson A., Multigrid Unsteady Navier-Stokes Calculations with

Aeroelastic Applications, 33rd AIAA Aerospace Sciences Meeting, AIAA Paper 95-0048, Reno, NV,

January, 1995.

6

Page 7

1

2

3

4

5

6

Figure 5: Mach Number Contours. Pitching Airfoil Case. Re = 1.0 10

6

, M

= 0.796, K

c

= 0.202. Read

figures by lines.

7

This is the html version of the file http://www.aoe.vt.edu/%mason/Mason&f/'onfig(ero)i*ift.pdf.

G o o g l e automatically generates html versions of documents as we crawl the we.

To lin! to or oo!mar! this page" use the following url: http://www.google.com/search?q=cache:/h12et/e1a1J:www.aoe.3t.ed(/4maso!/aso!_f/Co!figAero5i6ift.pdf&'o(r!als&

)performa!ce&of&differe!t&airfoils&i!&cfd*+hl=e!+gl=i!+ct=cl!,+cd=$7

'oogle is neither affiliated with the authors of this page nor responsi%le for its content.

These search terms have een highlighted: performance different airfoils cfd

These terms only appear in lin!s pointing to this page: journals

Page 1

W.H. Mason

4/5/06

8. High-Lift Aerodynamics

8.1 Introduction: Why high lift?

For transonic transports, the high-lift system design is a critical part of the configuration

design.

To achieve reasonable field performance while also obtaining efficient transonic

cruise the

design will require a fairly sophisticated high lift system. From a paper by Boeing

aerodynamicists

1

: (presumably referring to the B-777)

A 0.10 increase in lift coefficient at constant angle of attack is equivalent to reducing

the

approach attitude by one degree. For a given aft body-to-ground clearance angle, the

landing gear may be shortened for a savings of airplane empty weight of 1400 lb.

A 1.5% increase in maximum lift coefficient is equivalent to a 6600 lb increase in

payload

at a fixed approach speed

A 1% increase in take-off L/D is equivalent to a 2800 lb increase in payload or a 150

nm

increase in range.

For fighters, devices are also scheduled allow efficient maneuver.

High-lift systems are also critical for STOVL and V/STOL aircraft. They also use the

propulsion

system to help generate the lift. It always seemed to me peculiar to design fighter aircraft,

or

virtually any military aircraft, to operate from traditional runways. The one thing the

adversary is

going to know is the exact location of your runways. So a STOVL capability seems to be

critical

in a serious confrontation.

Current status: Typical values of C

Lmax

are shown in Table 1. They come from papers by Brune

and McMasters,

2

Roskam and Lan,

3

and Sanders.

4

Table 1. Values of C

Lmax

for some airplanes.*

* Note that there is a significant variation of values from different sources.

Clearly the 727 emphasized short fields, and thus required a higher C

Lmax

Anyone who ever

looked out the window while landing in a 727 noticed the elaborate high lift system

employed.

Model

C

Lmax

B-47/B-52

1.8

367-80/KC-135

1.78

707-320/E-3A

2.2

727

2.79

DC-9

3.0

737-200

3.2

747/E-4A

2.45

767

2.45

777

2.5

Page 2

8-2 W. H. Mason, Configuration Aerodynamics

Some Key Aspects:

Compressibility can be important early.

Reynolds number scaling from WT to flight may be problematic.

Today simple high lift systems are critical, the high manufacturing cost for high lift

systems

is important.

Classes of problems

High lift for a single element airfoil

Multi-element airfoils

Use of blowing in some form: Powered Lift

Computing:

Requires consideration of viscous immediately. (unlike typical cruise airfoil analysis

and

design, where some insight can usually be gained ignoring viscous effects).

Predicting high-lift is done almost entirely with Navier-Stokes (RANS) codes

(exception,

Prof. Mark Drelas MSES

5

code). A recent summary of the computational capability is by

Rumsey and Ying.

6

Single element airfoils

The key example of how to obtain high lift on a single element airfoil is the story of

Liebecks high lift airfoil

7

and the Stratford pressure recovery shape of the pressure

distribution. This introduces the classic paper by A.M.O. Smith.

8

See section 8.5 below.

Multi-element airfoils:

Understanding the physics: Section 6.3 of A.M.O. Smiths paper

8

is critical to understanding

the physics. Know what is meant by: 1. The slat effect, 2. The circulation effect, 3., The

dumping effect, 4. Off-the-surface pressure recovery, and 5. The fresh boundary layer

effect.

Note: some people combine the circulation and dumping effects and call it the vane

effect.

See section 8.5 below for details.

Design: See the recent survey by C.P. van Dam,

9

Existing systems on commercial transports: Rudolph has surveyed the high-lift systems

on

current subsonic transport aircraft.

10

8.2 Types of Trailing Edge Devices

To begin we include a number of examples of high-lift devices as drawn by Dick Kita of

Grumman.

11

These are fairly realistic drawings, as opposed to many of the drawings in

textbooks, which are cartoonish. Kita worked in high lift for many years. Im aware of his

work

on the Gulfstream II and F-14, but he worked on many other Grumman aircraft high-lift

systems

as well.

Of the large number of papers addressing high lift, several deserve mention. Pepper, et al

12

provides a more current look at high lift, while the Boeing 777 high lift system

development, as

well as the overall design process, is available in the paper by Nield.

13

A valuable description of

high lift on transports in contained in Gratzer.

14

The use of powered lift is covered in the survey

by Korbacher.

15

Somewhat dated but valuable resources are the book by Hoerner and Borst,

16

and the book by McCormick

17

(recently reissued unchanged from the 1967 edition as a Dover

paperback). Perhaps the best chapter on high lift in a basic text is the chapter in Shevell.

18

Page 3

High-Lift Aerodynamics 8-3

a) basic devices

Figure 8-1. Examples of typical trailing edge devices. Note that the Fowler flap also adds

area,

so that part of the C

Lmax

increase is simply due to the use of the original reference area

in computing the C

L

. (from Dick Kitas Grumman talk, Feb. 1985)

Page 4

8-4 W. H. Mason, Configuration Aerodynamics

b) other trailing edge devices

Figure 8-1. Examples of typical trailing edge devices.

(from Dick Kitas Grumman talk, Feb. 1985)

Page 5

High-Lift Aerodynamics 8-5

8.3 Types of Leading edge devices

Figure 8-2. Typical leading edge device concepts.

(from Dick Kitas Grumman talk, Feb. 1985)

Page 6

8-6 W. H. Mason, Configuration Aerodynamics

Figure 8-3. The actual high-lift system employed on the F-14.

(from Dick Kitas Grumman talk, Feb. 1985)

Notice that the F-14 has a fairly elaborate scheme, where the cove region is smoothed

with a

moving flap, and the upper surface of the slat also has a movable piece to fair the flap.

Also note

the spoilers. Many fighter airplanes use spoilers instead of ailerons for roll control,

although their

use varies from company to company.

Page 7

High-Lift Aerodynamics 8-7

8.4 Aerodynamics of Leading and Trailing Edge Devices

Figure 8-4. Typical effect of flaps on lift (from Dick Kitas Grumman talk, Feb. 1985)

This figure illustrates the effects of flap deflection on lift. Note that the angle of attack for

C

Lmax

actually decreases as the flap is deflected.

Page 8

8-8 W. H. Mason, Configuration Aerodynamics

Figure 8-5. Effect of flap extension on lift.

(from Dick Kitas Grumman talk, Feb. 1985)

Flaps that extend in a Fowler motion also benefit from additional wing area. Because we

continue to use the same reference area, the C

Lmax

value increases.

Page 9

High-Lift Aerodynamics 8-9

Figure 8-6. Typical variation of C

Lmax

with Mach and Reynolds Number.

(from Dick Kitas Grumman talk, Feb. 1985)

In general, we expect C

Lmax

to increase with Reynolds number as shown here for a clean airfoil.

However, sometimes the projection is not so straightforward, and C

Lmax

may even decrease.

19

Considering the adverse effect of Mach number on C

Lmax

, notice the low Mach numbers at which

the effects take place. Its rather surprising to the uninitiated.

Page 10

8-10 W. H. Mason, Configuration Aerodynamics

Figure 8-7. Effect of leading edge slats.

(from Dick Kitas Grumman talk, Feb. 1985)

Leading edge slats work to protect the leading edge from separation. Therefore, they

dont really

do anything until you reach the angle of attack where the leading edge flow would let

go

without the slats for protection. Thus the slats allow the lift to continue to rise to higher

angles of

attack.

Page 11

High-Lift Aerodynamics 8-11

Figure 8-8. Effects of various types of leading edge devices on C

Lmax

.

(from Dick Kitas Grumman talk, Feb. 1985)

Different leading edge devices differ in their effectiveness. This is Kitas estimate of

how each

type of device affects the performance of the wing.

Page 12

8-12 W. H. Mason, Configuration Aerodynamics

Figure 8-9. Estimated performance of various types of high lift systems.

(from Dick Kitas Grumman talk, Feb. 1985)

This is Kitas estimate of the typical best performance you can get from various types

of high-

lift systems. You can see that sweeping the trailing reduces the effectiveness of all the

systems.

The curve labeled advanced is typical of projections made in advanced departments,

where the

assumption is that an advanced technology development effort can improve the

performance of

any system. This may or may not be true.

Page 13

High-Lift Aerodynamics 8-13

Figure 8-10. Effects of flap deflection on drag.

(from Dick Kitas Grumman talk, Feb. 1985)

In addition to lift, flaps make a large change in drag. Clearly, you dont want the flaps

deployed

at low lift, and thus flaps arent deflected in cruise. As lift increases, there may be an

optimum

flap deflection schedule, and this is done, for example, on the F-18, where the flaps are

scheduled with angle of attack and Mach number. This was also done on the Grumman

X-29.

Page 14

8-14 W. H. Mason, Configuration Aerodynamics

Figure 8-11. Flap effects on pitching moment.

(from Dick Kitas Grumman talk, Feb. 1985)

Flap deflection also produces a large change in pitching moment. This is an important

consideration, since you need to be able to trim this pitching moment. In the case of the

Beech

Starship, the canards actually changed sweep to be able to generate the required force. So

this

effect cannot be ignored in developing the high lift system.

Page 15

High-Lift Aerodynamics 8-15

Figure 8-12. Definition of gap and overlap.

(from Dick Kitas Grumman talk, Feb. 1985)

In developing the high lift system, the selection of the gap between the slot and main

element,

and the overlap have been found to be two of the key parameters. A lot of wind tunnel

time,

and more recently computer resources, are spent trying to identify the values of these

parameters

that produce the highest lift. Figure 8-12 provides Kitas definition of these parameters.

8.5 Physics of high lift: A.M.O. Smiths analysis of the high lift aerodynamics

Now that weve surveyed the characteristics of high lift systems, we need to examine the

physical basis for the operation and limits of high lift systems. A.M.O. Smith wrote the

book

on the physics of high-lift systems

8

and is required reading. His message: you need to carry as

much lift (load) as you can on the airfoils upper surface without separating the boundary

layer.

His classic paper describes the physics associated with the high-lift characteristics we

described

above, and how to achieve the available high lift performance. We summarize his

description

here.

To obtain insight into the characteristics of pressure distributions as they affect boundary

layer separation, Smith introduced the use of a canonical pressure distribution. He felt

strongly

that this was necessary to understand and compare possible separation on different

airfoils. It is

Page 16

8-16 W. H. Mason, Configuration Aerodynamics

essentially another type of dimensionless or scaled pressure distribution. For boundary

layer

investigations it is found that the best scaling factor is the velocity just before the

deceleration

begins. In Smiths canonical system, C

p

= 0 represents the start of the pressure rise and C

p

= +1

the maximum possible value, that is, u

e

= 0.The velocity at the start of the pressure rise is u

0

.

Thus, the canonical pressure distribution is defined as

C

p

= 1

u

e

u

0

2

The next step is to examine the best way to specify the pressure distribution to allow the

pressure to recover to as close as possible to C

p

= +1. Smith made a parametric study of various

possibilities to gain insight into the best way to prescribe a pressure distribution to

delay

separation. However, here we look at his limiting case. It makes use of an analysis by

Stratford

that was done to estimate separation before the days when boundary layer computer

programs

were available and their use routine (1959). Using Stratfords analysis, it is possible to

define a

pressure distribution where the boundary layer is everywhere just on the verge of

separation. To

do this, we manipulate the Stratford criteria:

C

p

x

dC

p

dx

10

6

R

(

)

1

10

1

2

= S

Note that originally, Stratford used this relation to say that separation occurred when the

quantity

on the LHS of the equation reached the value of S (typically 0.35). However, we can

define a Cp

distribution using this relation that is everywhere equal to S, just on the verge of

separation, and

this pressure distribution is the best way to achieve a very large pressure recovery without

separation. Figure 8-13 shows the resulting pressure distribution. Examining this pressure

distribution, several key observations can be made. The initial slope is infinite, and then

decreases. Thus, when the boundary layer is thin, it can withstand a very large pressure

gradient.

As the boundary layer thickens (either when it starts to recover, or as it recovers) it

cannot

sustain the large pressure gradient, and the pressure gradient to maintain attached flow

decreases.

This illustrates the idea that thick boundary layers are more likely to separate than thin

boundary

layers. Note also that the Reynolds number effect is relatively weak. Finally, the

boundary layer

could recover all the way to C

p

= +1, but to attain this, x would need to go to infinity. These

shapes are the best possible pressure distributions to use to recover the pressure without

separating the boundary layer. as Smith notes, the only way to do better is to use some

sort of

active boundary layer control (suction or blowing).

Page 17

High-Lift Aerodynamics 8-17

-0.20

0.0

0.20

0.40

0.60

0.80

1.0

0.0

0.50

1.0

1.5

2.0

Canonical

Cp

x - feet

solid line U

0

/v = 10^6

dashed line U

0

/v = 10^7

Figure 8-13. Stratford Limiting flows for two different Reynolds numbers.

Single element airfoils: The key is how to obtain high lift on a single element airfoil, and

is

essentially the story of Liebecks high lift airfoil

7

and the Stratford pressure recovery shape of

the pressure distribution described above, as told by Smith.

8

The question of how much lift you

can obtain on a single element airfoil involves how low the pressure can be on the upper

surface,

and how the pressure can recover to a positive pressure coefficient at the trailing edge

and keep

the boundary layer attached. Smith describes two aspects of the problem. In the first case,

he

explains the limit of the pressure coefficient in terms of the vacuum C

p

when a zero pressure is

specified on the airfoil upper surface. Thus, using the definition of C

p

,

C

p

=

p p

1

2

2

we can obtain an alternate form using q = 1

2

2

=

2

P

2

, that is

C

p

=

p p

2

p

2

Page 18

8-18 W. H. Mason, Configuration Aerodynamics

and vacuum C

p

occurs when the pressure is zero:

C

p

vac

=

2

M

2

Then, he points out that only a value of 70% of the vacuum C

p

has been achieved in practice.

This results in a C

p

limit of M

2

C

p

= 1 for of 1.4.

Next, Liebeck and Smith used the analytical analysis by Stratford illustrated above to

specify

a pressure distribution that allows the most lift to be obtained. Given this pressure

distribution,

an inverse method is used to obtain the associated airfoil shape. The result is the Liebeck

family

of high lift airfoils.

Multi-element airfoils: Understanding the physics: read section 6.3 of A.M.O. Smiths

paper.

The five ideas are:

1. The slat effect. The slat protects the leading edge of the main element. Thats why its

effect

is only observed near C

Lmax

of the single element. Thought of as a point vortex, the slat velocity

acts to reduce the velocity around the leading edge of the main element.

2. The circulation effect. The downstream element causes the upstream element to be in a

high

velocity region, inclined to its mean line. To meet the Kutta condition, the circulation has

to be

increased. Instead of the airfoil deflecting as a plain flap, the trailing edge is placed in an

inclined

flow, something else (the downstream element) turns the flow.

3. The dumping effect. The trailing edge of the forward element is in a region of velocity

appreciably higher than the freestream velocity. Thus, the boundary layer can come off

the

forward element at a higher velocity. You dont have to recover back to Cp = +0.2 for

attached

flow, relieving the pressure rise on the boundary layer, alleviating separation problems

and

permitting increased lift. The suction lift can be increased in proportion to U

TE

2

for the same

margin against separation.

4. Off-the-surface pressure recovery. The boundary layer leaves the trailing edge faster

than the

freestream, and now becomes a wake (a viscous phenomena). The recovery back to

freestream

velocity can be more efficient way from contact with the wall. Wakes withstand more

than

boundary layers. Note that the wake can actually separate out in the flowfield. Note: for

a well

designed high lift system the local boundary layers and wakes remain separate. If they

merge,

everything is more complicated.

5. The fresh boundary layer effect. Thin boundary layers can sustain a greater pressure

gradient

than a boundary layer. Thus, three thin boundary layers (on three airfoil elements) are

more

effective than one thick boundary layer (single element).

Note: some people combine the circulation and dumping effects and call it the vane

effect.

8.6 Computational methods for high lift

Significant effort has been devoted to improving prediction capabilities for high lift

systems. As

stated in the introduction, the best recent survey is by Runsey and Ying.

6

In the meantime, for

low speed predictions of the maximum lift for a single element airfoil, XFOIL can be

used.

experience shows that its predictions are slightly higher than experimental results.

Nevertheless,

this is rea remarkable capability of a code than be run on a laptop PC.

Page 19

High-Lift Aerodynamics 8-19

8.7 Passive and active boundary layer control

Passive Boundary Layer Control: The boundary layer can be prevented from separating

by the

use of vortex generators, snags and fences.

20

Active boundary layer control: If suction or blowing is used to suppress boundary layer

separation, the blowing (which is generally preferred to suction) is known as boundary

layer

control (BLC), if the amount of blowing exceeds the value required for BLC, then the

blowing is

know as powered lift. The key parameter used to describe the amount of blowing is the

blowing coefficient, defined as:

C

=

m

j

V

j

qc

where the subscript j refers to the jet, and q is the dynamic pressure. Blowing for BLC

was used

often in early fighters, but is not used nearly as much today. The F-4 Phantom originally

had

blowing on both the leading edges and over the trailing edge flap. However, to improve

transonic

maneuver characteristics and to improve resistance to departure, the leading edge

blowing was

replaced by leading edge slats.

21

8.8 Powered Lift

A huge class of concepts have been tried to increase maximum lift using high pressure air

from

the engine. Some examples of powered lift concepts are:

propeller slipstream deflection (Brequet 941/McDonnell Model 188)

externally blown flaps (McDonnell DouglasYC-15/C-17)

internally blown flaps

upper surface blowing (Boeing YC-14, NASA QSRA, Ball-Bartoe JetWing)

vectored thrust (AV-8 Harrier)

jet flaps (Hunting H.126)

jet augmentor wings (NASA-deHavilland Augmentor Wing Aircraft)

circulation control (advocated by Hokie Bob Englar

22

, A-6 CCW)

8-9 Configuration Integration issues

The best airplane C

Lmax

you can achieve with a mechanical high lift system is about 3 3.5

The military and civil air regulations require a margin between C

Lmax

and the operating C

L

of the airplane. This must be accounted for during design. For example, the approach

speed must be 1.3 times the stall speed. This would suggest that the maximum approach

C

L

is only 59% of the C

Lmax

. However, some relief is available because the measured stall

speed, V

smin

is usually about 0.94 times the stall speed in 1g steady flight (the wind tunnel

case). This means that you can use a C

L

of about 67% of the C

Lmax

.

9,23

Increased span can be used to reduce the induced drag, so a bigger flap angle can be

used

before a climb limit is encountered, but this is a heavy solution.

Sweep decreases max lift

Page 20

8-20 W. H. Mason, Configuration Aerodynamics

2D to 3D: lots of losses dont be mislead by 2D C

Lmax

values, the real 3D value will be

much less.

The maximum C

L

available for takeoff and landing for many swept-wing airplanes is

actually the limit on angle of attack to avoid tailscrape.

The best high-lift configuration integration description also involves powered lift. The

YC-14 AIAA Case Study

24

is highly recommended.

Finally, issues not covered but worth mentioning: the concept of the Gurney flap

7

and the need

for accuracy around the leading edge.

8-10 Exercises

1. Examine the predictive capability of XFOIL for C

Lmax

. Use the data from Abbott and von

Doenhoff supplied previously for the NACA 0012 and 4412 airfoils at a Reynolds

number of

6 million and free transition. Comment on your results.

2. Read the high lift paper by A.M.O. Smith. Summarize what you learned in one page.

Pay

special attention to the details of single and multielement airfoils described in Sections 3-

6.

8-11 References

1

Garner, P.L., Meredith, P.T., and Stoner, R.C., Areas for Future CFD Development as

Illustrated by Transport Aircraft Applications, AIAA Paper 91-1527, 1991.

2

G.W. Brune and J.H. McMasters, Computational Aerodynamics Applied to High Lift

Systems, Chapter 10 of Applied Computational Aerodynamics, P. Henne, Ed. Progress

in

Astronautics and Aeronautics, Vol. 125, AIAA, Washington, 1990.

3

Jan Roskam and C-T Edward Lan, Airplane Aerodynamics and Performance,

DARcorporation,

Kansas, 1997, pp. 343.

4

Karl L. Sanders, High-Lift Devices, A Weight and Performance Trade-Off

Methodology,

Society of Allied Weight Engineers (SAWE) Technical Paper 761, May 1969.

5

Mark Drela, Design and Optimization Method for Multi-Element Airfoils, AIAA Paper

93-

0969, Feb. 1993.

6

Christopher L. Rumsey and Susan X. Ying, Prediction of high lift: review of present

CFD

capability, Progress in Aerospace Sciences, Vol. 38, pp. 145-180, 2002. (note that

articles in

Progress in Aerospace Sciences are available for download through the Virginia Tech

University

Library if you search Addison and have a vt.edu address)

7

Robert H. Liebeck, Design of Subsonic Airfoils for High Lift, Journal of Aircraft, Vol.

15,

No. 9, Sept. 1978, pp. 547-561.

8

A.M.O. Smith, High-Lift Aerodynamics, 37

th

Wright Bothers Lecture, Journal of Aircraft,

Vol. 12, No. 6, June 1975

9

C.P. van Dam, The aerodynamic design of multi-element high-lift systems for transport

airplanes, Progress in Aerospace Sciences, Vol. 38, pp. 101-144, 2002. (note that

articles in

Progress in Aerospace Sciences are available for download through the Virginia Tech

University

Library if you search Addison and have a vt.edu address)

10

Peter K.C. Rudolph, High-Lift Systems on Commercial Subsonic Airliners, NASA CR

4746, Sept. 1996.

Maintaining an accurate leading edge contour is critical at high lift conditions. Once after a navy depot had

repainted an F-14 wing, a small ridge was left on the leading edge where the upper and lower surface paint

overlapped. This was enough to cause early stall. Ed Heinemann reported a similar experience on a

Douglas airplane

during World War II in his autobiography.

Page 21

High-Lift Aerodynamics 8-21

11

Dick Kita, Mechanical High Lift Systems, Grumman Aerodynamics Lecture Series,

Grumman Aerospace Corporation, Feb. 1985.

12

C.P. van Dam, J.C. Vander Kam, and J.K. Paris, Design-Oriented High-Lift

Methodology for

General Aviation and Civil Transport Aircraft, Journal of Aircraft, Vol. 38, No. 6,

November-

December 2001, pp. 12076-1084.

13

B. N. Nield, An overview of the Boeing 777 high lift aerodynamic design,

Aeronautical

Journal, Nov. 1995. pp. 361-371.

14

L.B. Gratzer, Analysis of Transport Applications for High Lift Schemes, AGARD LS

43,

1971.

15

G.K. Korbacher, Aerodynamics of Powered High-Lift Systems, Ann. Rev. of Fluid

Mech.,

1974.

16

S.F. Hoerner and H.V. Borst, Fluid Dynamic Lift, 1975.

17

B.W. McCormick, Jr., Aerodynamics of V/STOL Flight, Dover, 1999.

18

Shevelle, Fundamentals of Flight, 2

nd

ed., Prentice-Hall, 1989.

19

John H. McMasters and M.D. Mack, High Reynolds Number Testing in Support of

Transonic

Airplane Development (Invited Paper), AIAA Paper 92-3982, July 1992.

20

D.G. Mabry, Design features which influence flow separations on aircraft,

Aeronautical

Journal, Dec. 1988, pp. 409-415.

21

D.H. Bennett and W.A. Rousseau, Seven Wings the F-4 Has Flown, Evolution of

Aircraft

Wing Design Symposium, Dayton, OH, AIAA Paper 80-3042, March 1980.

22

Englar, Robert J., Smith, Marilyn J., Kelley, Sean M., Rover, Richard C., III,

Development of

circulation control technology for application to advanced subsonic transport aircraft,

AIAA

Paper 93-0644.

23

A. Flaig and B. Hilbig, High-Lift design for large civil aircraft, in AGARD High-Lift

System Aerodynamics, R 415, 1993.

24

John K. Wimpress and Conrad F. Newberry, The YC-14 STOL Prototype: Its Design,

Development, and Flight Test, AIAA Case Study, AIAA, Reston, 1998.

- hicks1978.pdfUploaded byVinicius Moda
- 19790005847_1979005847Uploaded byYRDscs
- impulseBrief.pdfUploaded byARYAYESHU SINGH
- Aero BackgroundUploaded bySenthil Kumar
- Jet Engine Fundamental of Theory Design OperationsUploaded bybaseer12
- The Work of WingsUploaded byWilliam Robertson
- QDB 15 POFUploaded byant
- Introduction to Lifting Line theoryUploaded byEthan Loza
- Technical Report Boost GlidersUploaded byvoraze
- aerospace-03-00026Uploaded byJesus Alfredo Soria Sanchez
- AerodynamicsUploaded byRahul Verma
- Buffet Envelope Prediction of Transport AircraftUploaded byf_mashayekhi
- 2002-04-19_airfoilUploaded byPraveen P Jose
- Span Efﬁciencies of Wings at Low Reynolds NumbersUploaded byAlexandros Haf
- Lecture 5 CCN2299 Assignment 2 2018 Rev2(1st)Uploaded byDickson So
- Drilling Case StudyUploaded byAhmed Elkady
- Donnelly DJ T 2010Uploaded byxeyran
- GPHYS Lecture 17 Fluids in Motion 2013FUploaded bydhaval2011
- DragUploaded byvishal
- Chapter 5Uploaded byRichardHoldich
- 58 SafetyUploaded byRamesh Selvaraj
- 1-s2.0-S0140700702000166-mainUploaded byMuhammadRiady
- PhysRevLett.93.144501Uploaded byErin Morales
- Stem Lesson Ld04 Aerodynamics SarradetUploaded byYadana1
- on of Turbulent Gas-solid FlowUploaded bymarmaduke32
- advisory cercularUploaded byankonmahmud
- Day 1 - Complete Tech Quiz Questions - Finalised and PrintedUploaded byShabbir Bhojani
- U91 Principles of Helicopter Flight and Aerodynamics AM2Uploaded byYannis
- A Computational Investigation of Ground Simulation for a Saloon CarUploaded byvolo87
- Wind Tunnel TechniqueUploaded byRaj Manova

- RBI indiaUploaded bydbrcs
- JMEMS_2015Uploaded bydbrcs
- Hanging Crane Model and Lagrange's EquationsUploaded bydbrcs
- Drift removalUploaded bydbrcs
- Lecture 6 Modeling and Linearization(1)Uploaded bydbrcs
- 2005 - Advanced Robotic Manipulation_OussamaKhatibUploaded bynachoflores214793
- Graphs PrintUploaded bydbrcs
- Differential EquationsUploaded bydbrcs
- Lecture 20 StabilityUploaded bydbrcs
- Routh-Hurwitz Stability CriterionUploaded bydbrcs
- Doc1Uploaded bydbrcs
- APL Origami Solar 2014Uploaded bydbrcs
- Book. 2010. Introduction to Embedded Systems Using Ansi C and the Arduino Development EnvironmentUploaded byfrsalazar
- 51Uploaded bydbrcs
- Scientific Reports PreprintUploaded bydbrcs
- Time Response Analysis lecture notesUploaded bydbrcs
- Lecture 6 Modeling and Linearization(1)Uploaded bydbrcs
- Lecture17 Midsem Exam Analysis 24sep09Uploaded bydbrcs
- Differential EquationsUploaded bydbrcs
- Mil Handbook 5hUploaded bydarkwing888
- Paper 16Uploaded bydbrcs
- Students Developing Low-cost Portable Braille PrinterUploaded bydbrcs
- 2014 Baja SAE India RulesUploaded bydbrcs
- New Snake Discovered in OdishaUploaded bydbrcs
- New Snake Discovered in OdishaUploaded bydbrcs
- New Snake Discovered in OdishaUploaded bydbrcs
- Root Locus PlotsUploaded bydbrcs
- Block Diagram Examples for LCSUploaded bydbrcs
- Linearization notesUploaded bydbrcs
- Digital Fingerprints 20oct09Uploaded bydbrcs

- Viscosity Exp updated.docx1111.docxUploaded byIbrahim
- 101174973.pdfUploaded bytonysawan
- Abet Course Syllabus - CHEN231Uploaded bysyed waheed ul hasan
- Fm Imp Questions & AnsUploaded byrajeshkumarpuvvala
- 50508_appe.pdfUploaded bywearplay
- IRJET- Performance Analysis of Winglet at Different Angle of AttackUploaded byIRJET Journal
- Slipstream TheoryUploaded byPallab Banik
- Hk BernoulliUploaded by123habib123fikri
- How to calculate tension in phsysicsUploaded byaristo_onan2546
- SP 8001 - Buffeting During Atmospheric AscentUploaded byElumalai Srinivasan
- 6dof Rigid Motion CfdUploaded byDaniel Coelho
- CFD Analysis of a Three Bladed H-Rotor of Vertical Axis Wind TurbineUploaded byIRJET Journal
- Fluid MechanicsUploaded byRegineDagumanFuellas
- Equilibrium ProblemsUploaded byhassan_m2222
- Bath Flapping WingUploaded bynd1977
- VDHS-0 Vehicle DynamicsUploaded bySeshan R.V
- Universal GravitationUploaded byMark Prochaska
- Motion Along a Straight LineUploaded byahmednalla
- 19650013440_1965013440Uploaded byernestozag
- Heat Transfer in Spiral Tube Coils - IJTS Altac & AltunUploaded byCostynha
- PhDThesis_rethore_phd_100209_RISOUploaded bypierre_elouan
- Computational Fluid DynamicsUploaded byMohammad Hosein Khanesaz
- Kinetics of Rigid BodyUploaded byLily Antonette Agustin
- Comput_Fluid_Dynamics.pdfUploaded byJinoyVargheseTharu
- Session1FundamentalsofVIV DaltonUploaded byAhmed Ibrahim
- Adequecy Check Report 5610 NC 06 BA 1617_Rev C1Uploaded byskzubs
- Turbulent flowUploaded bycrisjrogers
- Viscous FlowUploaded byAamir Qutub Alig
- Fluid Mechanics AMIEUploaded bysayhigaurav07
- Bernoulli_EquationUploaded byJosh Sam