You are on page 1of 92

Flight Simulator Database Population from Wind Tunnel and CFD

Analysis of a Homebuilt Aircraft

A Thesis
Presented to the Faculty of
California Polytechnic State University
San Luis Obispo

In Partial Fulfillment
Of the Requirements for the Degree of
Master of Science in Aerospace Engineering

By
Robert P. Little
May 2006
© Copyright 2006
Robert P. Little
All Rights Reserved

ii
Approval Page

TITLE: Flight Simulator Database Population from Wind


Tunnel and CFD Analysis of a Homebuilt Aircraft

AUTHOR: Robert P. Little

DATE SUBMITTED: May 2006

Dr. David D. Marshall (AERO) ______________________________

Adviser and Committee Chair

Dr. Jin Tso (AERO) ______________________________

Committee Member

Dr. Jon Tarantino (AERO) ______________________________

Committee Member

Dr. Lanny Griffin (CENG) ______________________________

Committee Member

iii
Abstract

Flight Simulator Database Population from Wind Tunnel and CFD Analysis of a
Homebuilt Aircraft

Robert P. Little

Current industry methods of utilizing both wind tunnel and CFD tools to populate flight

simulator databases are implemented on a homebuilt RV-7 kit aircraft in order to obtain

initial estimates of lift, drag, and moment coefficients. The resulting data are used to

populate portions of the Cal Poly Motion Simulator data tables. The methods for creating

an aerodynamically accurate solid model of the RV-7 aircraft in SolidWorks using

aircraft construction drawings provided by the RV-7 kit manufacturer are discussed first.

Techniques for creating a 1/15th scale wind tunnel model using rapid prototyping

technology are then discussed. Wind tunnel testing is conducted at 30 meters per second

and various angles of attack in the Cal Poly 3 by 4 foot low-speed wind tunnel in order to

obtain lift, drag, and moment coefficient data. The aerodynamic model created in

SolidWorks is then imported into the preprocessor GAMBIT 2.2.30 where a

computational grid of the RV-7 aircraft and surrounding flow volume is created for CFD

analysis. The CFD tool FLUENT 6.2 is used to obtain estimates of lift, drag, and moment

coefficients at 30 meters per second flow velocity and various angles of attack. A grid

independence study is then conducted to determine the accuracy of the computational

grid used for CFD analysis. The wind tunnel and CFD results are compared and

appropriate error bands placed on the data. Finally, data from both wind tunnel and CFD

testing is compiled into a final data set for use in the Cal Poly Motion Simulator.

iv
Acknowledgements

The author would like to thank Dr. David Marshall of the Aerospace Engineering

Department for his continual support and insight with this project. The support of Dr. Jin

Tso was instrumental in conducting wind tunnel testing. The help of Mr. Larry Coolidge

of the Mechanical Engineering Department in rapid prototyping a wind tunnel model was

critical to the breadth and completion of this project. Numerous students within the

College of Engineering deserve recognition for their assistance in various aspects of this

project including SolidWorks and CFD assistance, as well as model preparation and wind

tunnel testing. These students include Francesco Giannini, Stephen Kubik, Edward

Clements, Kiel Carreau, Ryan Huthmacher, Ted Garbeff and Julie de la Montanya. The

author would also like to thank his family for their love and support throughout this

project.

v
Table of Contents

LIST OF FIGURES….……………………………………….………………………. viii

LIST OF TABLES…….………………………………………………………………... x

NOMENTCLATURE………………………………………………………………….. xi

CHAPTER 1: INTRODUCTION…………………………………………………… 1

1.1 INVESTIGATION OF PREEXISTING WORK……………………………… 1

1.2 OVERVIEW OF WORK PRESENTED………………………………………. 5

1.3 ABOUT THE CAL POLY RV-7 MOTION SIMULATOR PROJECT………. 6

CHAPTER 2: GEOMETRY MODELING USING SOLIDWORKS…………….. 8

2.1 OBTAINING INFORMATION FROM AIRCRAFT HOMEBUILT

AIRCRAFT CONSTRUCTION PLANS…………………………………..….. 9

2.2 MODELING THE FUSELAGE……………………………………………… 10

2.3 MODELING THE EMPENNAGE…………………………………………… 14

2.4 MODELING THE WINGS………………………………………………….. 18

2.5 COMBINGING INDIVIDUAL COMPONENTS TO CREATE A

COMPLETE AIRCRAFT…………………………………………………….20

CHAPTER 3: WIND TUNNEL TESTING OF 1/15TH SCALE MODEL……… 22

3.1 TEST APPARATUS AND PROCEDURE………………………………….. 25

3.2 WIND TUNNEL ANALYSIS……………………………………………….. 27

3.3 WIND TUNNEL RESULTS………………………………………………… 30

CHAPTER 4: CFD ANALYSIS USING GAMBIT AND FLUENT……………. 35

4.1 GEOMETRY IMPORT AND SETUP………………………………………. 36

4.2 MESHING THE AIRCRAFT BODY SURFACES………………………….. 38

vi
4.3 MESHING THE FLOW VOLUME…………………………………………. 42

4.4 FLUENT 6.2 CASE SETUP………………………………………………….. 48

4.4.1 Defining the Models, Materials, and Boundary Conditions………………. 48

4.4.2 Solution Initialization and Setup……………………………………………… 50

4.5 GRID INDEPENDECE STUDY……………………………………………... 53

4.6 FLUENT RESULTS………………………………………………………….. 55

CHAPTER 5: WIND TUNNEL AND CFD DATA COMPARISON……………. 60

5.1 ERROR ESTIMATES………………………………………………………... 60

5.1.1 Wind Tunnel Error Calculation Procedure…………………………........... 60

5.1.2 CFD Error Analysis…………………………………………………………… 64

5.2 DATA COMPARISON……………………………………………………… 64

CHAPTER 6: CONCLUSIONS AND FUTURE WORK……………………….. 69

6.1 FUTURE WORK……………………………………………………………. 69

BIBLIOGRAPHY…………………………………………………………………….. 71

APPENDIX A: CONSTRUCTION DRAWINGS USED FOR SOLID MODEL … 72

APPENDIX B: WIND TUNNEL DATA TABLES………………………………….. 77

APPENDIX C: WIND TUNNEL SAMPLE CALCULATIONS…………………… 80

vii
List of Figures

Figure 2.1: Reference Plane Locations for Fuselage Cross Sections……………………10


Figure 2.2: Scanned Cross Sections Scaled and Located………………………………..11
Figure 2.3: Fuselage Cross Sections as SolidWorks Sketches…………………………. 12
Figure 2.4: Aft Fuselage is Now a Solid Body, Front Fuselage Ready for Lofting……..13
Figure 2.5: Completed Fuselage Solid Body………………………………………….... 14
Figure 2.6: Horizontal Stabilizer Root Chord Dimension Scheme……………………... 15
Figure 2.7: Horizontal Stabilizer Root and Tip Locations……………………………… 16
Figure 2.8: Final Right Horizontal Stabilizer……………………………………………17
Figure 2.9: Vertical Stabilizer…………………………………………………………... 18
Figure 2.10: Right Wing Solid Body…………………………………………………… 19
Figure 2.11: Completed RV-7 External Geometry Model in SolidWorks…………….. 21
Figure 3.1: SolidWorks Model for Rapid Prototyping………………………………… 23
Figure 3.2: Four Sections of Rapid Prototyped RV-7 with Dowels Inserted………….. 24
Figure 3.3: Finished Wind Tunnel Model of the RV-7 Attached to Sting Balance……. 25
Figure 3.4: Sting Balance Calibration Plot…………………………………………….. 31
Figure 3.5: Lift Coefficient versus Angle of Attack for 30 m/s Flow…………….…… 32
Figure 3.6: Drag Coefficient versus Angle of Attack for 30 m/s Flow…………….…... 33
Figure 3.7: Moment Coefficient versus Angle of Attack for 30 m/s Flow…………..… 34
Figure 4.1: Half Aircraft Geometry used for CFD and Corresponding Clean
Geometry in GAMBIT……………………………………………………... 37
Figure 4.2: Aircraft and Flow Volume Geometry in GAMBIT………………………… 38
Figure 4.3: Aircraft Faces Meshed in GAMBIT……………………………………….. 42
Figure 4.4: Aircraft Faces and Flow Volume Meshed………………………………….. 44
Figure 4.5: Representative Slice of Final Mesh in GAMBIT………………………….. 45
Figure 4.6: Boundary Condition Summary…………………………………………….. 47
Figure 4.7: Lift Curve from FLUENT Data……………………………………………. 56
Figure 4.8: Drag Curve from FLUENT Data…………………………………………… 57
Figure 4.9: Moment Curve from FLUENT Data……………………………………….. 58
Figure 5.1: Wind Tunnel and CFD Lift Curve Comparison……………………………. 65
Figure 5.2: Wind Tunnel and CFD Drag Curve Comparison…………………………... 66
Figure 5.3: Wind Tunnel and CFD Moment Coefficient Comparison…………………. 67

viii
Figure A.1: Firewall Bulkhead………………………………………………………… 72
Figure A.2: Canopy Frame…………………………………………………………….. 72
Figure A.3: F-706 Aft Fuselage Bulkhead…………………………………………….. 73
Figure A.4: F-708 Aft Fuselage Bulkhead…………………………………………….. 73
Figure A.5: F-712 Aft Fuselage Bulkhead…………………………………………….. 74
Figure A.6: Side View for Guide Curves……………………………………………… 74
Figure A.7: Top View for Guide Curves………………………………………………. 75
Figure A.8: Front View………………………………………………………………… 75
Figure A.9: Forward Fuselage for Locating Bulkheads…..……………………………. 76
Figure A.10: Aft Fuselage for Locating Bulkheads and Measuring Empennage……… 76

ix
List of Tables

Table 4.1: FLUENT Setup Summary…………………………………………………... 53


Table 4.2: Grid Study Summary………………………………………………………... 54
Table 4.3: FLUENT Results Summary 30 Meters per Second…………………………. 58
Table 4.4: Airspeed Effects Summary………………………………………………….. 59
Table 5.1: Wind Tunnel Error Summary……………………………………………….. 63
Table 5.2: CFD Error Summary………………………………………………………… 64
Table 5.3: Flight Simulator Data Summary…………………………………………….. 68
Table B.1: General Test Information…………………………………………………… 77
Table B.2: Calculated Test Conditions…………………………………………………. 77
Table B.3: Calibration Data…………………………………………………………….. 77
Table B.4: Raw Data Model and Strut………………………….………………………. 78
Table B.5: Raw Data Strut Only…………………………………………………………78
Table B.6: Final Aircraft Values………………………….…………………………….. 79
Table B.7: Final Calculated Coefficients……………………………………………….. 79

x
Nomenclature

A Axial Load T Temperature

c Mean Aerodynamic Chord U∞ Free Stream Velocity

CD Coefficient of Drag VNE Never-Exceed Speed

CFD Computational Fluid Dynamics VX Best Angle of Climb Airspeed

CG Center of Gravity VY Best Rate of Climb Airspeed

CL Coefficient of Lift α Angle of Attack

CM Coefficient of Moment ℓ Moment Arm Length

D Drag Force ρ Density

E Sting Balance Reading ∆ Difference of Value, ∆ = final - initial

Hg Mercury
Subscripts
lbs pounds
A Axial
L Lift Force
CG Center of Gravity
LSR Least Scale Resolution D Drag

M Moment Load
L Lift
mV Millivolts
M Moment
m/s Meters per Second
N Normal
N Normal Load
Sting Sting Balance Value
p Pressure
Strut Strut Assembly Value
q∞ Dynamic Pressure Wing RV-7 Wind Tunnel Value

R Ideal Gas Constant = 1717.548


0 m/s Wind Tunnel Setting of 0 m/s
RN Sting Balance Calibration Slope
30 m/s Wind Tunnel Setting of 30 m/s
S Wing Area
∞ Freestream

xi
Chapter 1: Introduction

Computational fluid dynamics (CFD) is used extensively in the aerospace field to provide

designers and engineers further insight into design issues that may arise at various stages

in the design process. In addition, CFD can be used to provide useful information on pre-

existing designs such as homebuilt aircraft. In aerospace applications, CFD can be used

as the sole means of analysis, or to complement additional analysis techniques and

processes. For example, because of time and/or monetary considerations, a design team

may choose to use CFD for all of the aerodynamic analysis of an aircraft prior to flight

testing. Similarly, it may in the best interest of the design team to collect both CFD and

wind tunnel data. In this instance, data may be collected for the same flight conditions

and compared, or the two methods may be used to collect data in different areas of the

flight regime. This project involves collecting data from both CFD and wind tunnel

experiments. Both techniques are used throughout the majority of the flight envelope for

positive angles of attack, but with more data points taken at higher angles of attack during

wind tunnel testing than for the CFD cases. This is due to the increased complexity in

numerical solutions for three-dimensional flows at high angles of attack. In addition, the

landing gear and propeller were not included in the wind tunnel or CFD models due to

the complexities involved with obtaining CFD and wind tunnel data with them attached.

1.1 Investigation of Preexisting Work

In both CFD and wind tunnel testing, aerodynamic forces are measured and can be

compiled for various applications. For example, depending on the complexity of the tests

being conducted, various aircraft aerodynamic coefficients can be estimated from

1
experimental data. An investigation into previous CFD work done on the RV-7 yielded

no results, thus pointing to the need for CFD testing to be conducted for this project.

The accurate modeling of the appropriate geometry is an essential first step for both CFD

and wind tunnel testing. After extensive research, the author was unable to find any

preexisting complete solid models of the RV-7 in a format that could be used to make a

wind tunnel model or for use in CFD. Flight Factory Simulations provides an RV-7 flight

simulator model for Microsoft Flight Simulator 2004. However, the actual solid model

used in the flight simulation model purchased from Flight Factory Simulations can not be

extracted for use in wind tunnel testing or CFD. A flight simulation model of the RV-7

also exists for the X-Plane flight simulator. However, the model is similar to the one

created by Flight Factory Simulations in that the actual solid model can not be extracted

from the flight simulator files. Therefore, a solid model of the RV-7 needed to be created

using a solid modeling program for both wind tunnel and CFD testing. 1

Before beginning further work on the RV-7 model, it was imperative to learn of pre-

existing analysis of the RV-7. Through extensive research, it was found that a flight test

program of an RV-6 aircraft was conducted by Comparative Aircraft Flight Efficiency, or

CAFE. The RV-6 aircraft is the predecessor to the RV-7 and the two aircraft are quite

similar. According to the manufacturers website, the RV-7 differs from the RV-6 in that

it is capable of handling larger engines, legroom, headroom, and useful load are

increased, the never-exceed speed (VNE) is increased to 230mph, and the fuel capacity is

increased from 38 gallons to 42. 2

2
The flight testing of the RV-6 conducted by CAFE included, among other things, useful

aerodynamic data on the aircraft. This included a drag summary plot for the aircraft at

various flight speeds. Various aircraft speeds such as stall speed in different flap

configurations, best rate of climb (Vy), and best angle of climb (Vx) were also recorded.

In addition, roll rates and stick forces were also recorded at various airspeeds. This data is

useful as a future reference for wind tunnel and CFD tests, but does not provide a

thorough analysis of CL, CD, and CM at various angles of attack. It is clear that there is a

need to develop methods for finding these coefficients for the RV-7 in order to provide a

foundation for future analysis of the aircraft. 3

Because this project involves both wind tunnel and CFD data, it was imperative to

investigate how CFD and wind tunnel analyses are used to obtain aerodynamic data and

how these data are used in a flight simulation database. Though this investigation does

not include all preexisting cases of projects involving mixed fidelity CFD and wind

tunnel data, the cases studied provided valuable information on how mixed fidelity data is

used and integrated in a project.

The Boeing Company currently uses both CFD and wind tunnel testing to obtain data for

their flight simulator databases. The data provided by both CFD and wind tunnel testing

must be very accurate because once the flight simulator database for a particular aircraft

model has been populated, the dataset is used in airline flight simulators for pilot

training.4,5

The present role of CFD at the Boeing Company is expanding to include analysis of the

full flight envelope through the use of Navier-Stokes codes. Previously, CFD use was

3
limited to mostly high-speed cruise design and was not used extensively in cases of

significant flow separation. In the past as well as the present, large amounts of data for

flight simulator database population are obtained from costly wind tunnel experiments.

However, as CFD and computing capabilities increase, it is becoming apparent that CFD

may be an efficient tool for populating flight simulator databases in various other regions

of the flight envelope that were previously only analyzed in wind tunnels.5

According to Tinoco, et al., CFD is not yet capable of replacing wind tunnel analysis and

the challenge today remains how best to combine CFD and wind tunnel testing. The

specifics for the optimum use of wind tunnel and CFD data to populate flight simulator

databases are beyond the scope of this report. However, the study of the use of wind

tunnel and CFD data at The Boeing Company indicates that while wind tunnel data may

be used to validate CFD results, it may be most efficient to use CFD to rapidly generate

data for flight simulator databases. Because of the uncertainties that remain in analyzing

massively separated flows, CFD will be used in this thesis at low to moderate angles of

attack prior to massive flow separation.5,6

A paper by S. Saephan, et al. discusses comparing wind tunnel and CFD results of a

rotary aircraft at high angles of attack. In this paper, the authors use experimental wind

tunnel data to confirm findings from the CFD investigation. By comparing the percent

difference for various cases of CFD data compared to experimental wind tunnel data, the

authors were able to show that stability derivatives can be efficiently and accurately

obtained from CFD. This report further confirmed that wind tunnel and CFD can be used

together to provide accurate results for this thesis project.7

4
In a Master’s Thesis presented by Dale R. Turley, the author presents various coefficient

plots versus angle of attack for both flight test and wind tunnel data of a PA-30 twin

engine aircraft. Each plot depicts flight test and wind tunnel data, as well as the

corresponding uncertainty levels for the flight estimates. This report shows the possible

discrepancy in mixed fidelity data and how error estimates can be used to provide the

researcher with a reasonable range of data values. It is evident from this investigation that

an error analysis will be crucial in determining the accuracy of the results found from

wind tunnel and CFD tests.8

After investigating previous CFD testing done on the RV-7, the availability of preexisting

solid models to conduct these investigations, and using mixed fidelity wind tunnel and

CFD data, it is clear that an investigation of the RV-7 stability derivatives using a model

created in SolidWorks for wind tunnel and CFD testing is warranted. In addition, the use

of both CFD and wind tunnel testing is warranted based on the lack of existing data on

the RV-7 for use as a reference. Therefore, it is essential to use more than one method for

obtaining aerodynamic characteristics to ensure accuracy of the final data set.

1.2 Overview of Work Presented

The purpose of this thesis is to provide a methodology for rapidly obtaining initial

stability derivatives of homebuilt aircraft for implementation into the Cal Poly Motion

Simulator and provide a foundation from which further testing can be done. The test

aircraft for this project is a Van’s Aircraft RV-7. This type of aircraft is currently being

built by Cal Poly students and is the basis for the creation of the Cal Poly Motion

Simulator. The first phase of this thesis project consists of using the building plans for

5
the aircraft to create an accurate solid model of the aircraft external geometry in

SolidWorks. The solid model is then imported into a CFD program (FLUENT) using a

pre-processor for meshing (GAMBIT) and tested at various flight conditions. The Cal

Poly 3x4 foot low-speed wind tunnel is used to validate CFD data and collect data at

higher angles of attack. The data taken from CFD and wind tunnel tests is then compared

and compiled into a single table for use by the Cal Poly Motion Simulator.

1.3 About the Cal Poly RV-7 Motion Simulator Project

The Spring Quarter of 2003 academic school year was the first time a class on building

experimental kit-built aircraft was taught at Cal Poly. The class began as a result of a

student’s interest to share his background and expertise in aircraft construction with other

students. Much of the Spring Quarter was spent teaching students the basics of building

an RV-7. Students learned skills necessary to construct an aluminum aircraft such as

drilling, dimpling, countersinking, and riveting techniques.

Interest in the class increased as students found out about the opportunity to obtain hands-

on aircraft building experience and thus a new perspective on their aeronautical

engineering education. As more students became involved with the project, an increasing

number of ideas based on the RV-7 came forth. One of those ideas was to build upon the

foundation of flight simulation knowledge already available within the Aerospace

Engineering Department and create a motion simulator of the RV-7.

In order for the motion simulator to function accurately, various aerodynamic data for the

RV-7 must be obtained. This thesis is meant to be a foundation for obtaining these data

6
and will cover methods for obtaining CL, CD, and CM values of the aircraft at various

flight conditions. Upon completion of this project, it may be used as a starting point for

researchers to obtain additional data on the RV-7. This paper is also intended as a

representation of current industry practice of using both wind tunnel and CFD data to

populate flight simulator tables.

7
Chapter 2: Geometry Modeling Using SolidWorks

In order to obtain CFD and wind tunnel results, the aircraft external geometry was

modeled using SolidWorks. Pre-existing aircraft solid models similar to that of the

aircraft to be tested were investigated in order to find out if the external geometry of the

RV-7 was already available in a usable form. To be of any use, the aircraft model must be

available as a certain type of file that accurately conveys the aircraft geometry, such as an

IGES, ACIS, or STEP file. If these files already exist and are accessible, the experimenter

may not need to reproduce the aircraft geometry.

The investigation of pre-existing geometry yielded several models of the RV-7. The first

model was created by Factory Flight Simulations for use in Microsoft Flight Simulator

2004. The full flight model is available for download from their website. Despite the

accuracy of the flight model, the files attained were for use in Microsoft Flight Simulator

and did not contain an aircraft geometry file in an appropriate format for this project.

Several other flight simulator models were found, including a model for flight simulator

X-Plane. However the file formats of these models were also not consistent with those

needed to conduct CFD and wind tunnel analysis. Therefore, the aircraft geometry for the

RV-7 was created in SolidWorks.1

This chapter presents a method for using aircraft building plans and other sources to

accurately reconstruct a solid model of a homebuilt aircraft in SolidWorks. The aircraft

will be initially modeled in components, then assembled together to create the final

geometry. The resulting solid model will serve as the baseline model for all aerodynamic

testing techniques discussed in this thesis. It should be noted that the methodology

8
presented in this chapter represents one of possibly multiple approaches that could be

employed to accurately model an RV-7 in SolidWorks. The material in this chapter

should be used as a reference for future modeling of homebuilt aircraft and not

necessarily as an acknowledged standard procedure.

2.1 Obtaining Information from Homebuilt Aircraft Construction Drawings

The first step in creating accurate aircraft geometry of a homebuilt aircraft is to gather as

much information about the aircraft as possible. Possible sources of information include

online databases or the aircraft manufacturer’s website, as well as information published

in aerospace articles. For homebuilt aircraft, construction plans purchased from the

aircraft manufacturer provide a wealth of information and are used extensively in this

methodology.

Several characteristics are vital in determining the applicability of a construction plan

drawing to the modeling process. In general, any depiction of the aircraft external

geometry with dimensions is useful, such as a standard three-view drawing. In addition,

cross section views throughout the aircraft are valuable in creating various aircraft

components, such as the fuselage. Because aircraft building plans are being used, cross

section views may be in the form of a bulkhead front view, in which case the outline of

the bulkhead may be used as a cross section. The next section shows how this technique

is used to create part of the RV-7 geometry.

It is also important to know the correct location of the cross section or bulkhead along the

length of the fuselage. In general, before beginning modeling or taking measurements

from drawings it is helpful to set a universal origin for all measurements. In addition, for

9
each drawing to be used, the drawing scale should be noted and the appropriate scaling

factor applied to all measurements taken from the drawing. Once all the relevant

drawings are compiled and applicable measurements taken from the drawings,

construction of the solid model may begin.

2.2 Modeling the Fuselage

This section outlines the method used for modeling the fuselage of the RV-7. The

fuselage is modeled as several loft features in SolidWorks using cross sections and guide

curves as loft references. The relevant drawings from the manufacturer for this portion of

modeling are the fuselage layout showing fuselage bulkhead locations, a dimensioned

three-view, and several drawings showing front views of various bulkheads. The latter

two are scanned and saved as picture files (such as .jpeg) for import into SolidWorks.

As a first step, the location of the fuselage bulkheads along the length of the fuselage are

measured and reference planes corresponding to these fuselage locations are created in

SolidWorks. Figure 2.1 shows a summary of the reference planes used to locate fuselage

bulkheads and cross sections.

Figure 2.1: Reference Plane Locations for Fuselage Cross Sections

10
Once the reference planes are created, the scanned bulkhead images can be imported into

SolidWorks on the appropriate reference plane. Once imported, each image must be

adjusted until it represents the appropriate scale. For instance, if the width of an actual

bulkhead is 40 inches, the image of that bulkhead in SolidWorks should be scaled such

that the width of the bulkhead in SolidWorks is also 40 inches. The image must also be

centered properly both horizontally and vertically to correspond to its actual location on

the aircraft. Figure 2.2 shows the five cross sections taken from construction plans and

the reference lines used to properly locate the images. It should be noted that the second

cross section from the left is not a bulkhead but is actually a cabin frame used to support

the sliding canopy of the aircraft, and thus requires additional work to be used as a loft

reference in SolidWorks.

Cabin Frame

Figure 2.2: Scanned Cross Sections Scaled and Located

11
Once all of the cross sections are located properly, a new sketch is created in SolidWorks

for each cross section image. Because SolidWorks does not recognize the images as

entities from which parts can be created, the outline of each cross section must be traced

in a SolidWorks sketch. This is done using spline curves and straight lines. Once the

cross section outlines have been traced, the scanned images are no longer needed and can

either be deleted or suppressed in SolidWorks. In SolidWorks there are now five distinct

profiles that define the majority of the fuselage.

The work involving the cross sections needed for the fuselage requires several more steps

to be complete. First, using the F-706 bulkhead as a reference (see Figure 2.1 for F-706

location), the cabin frame is extended to represent a complete cross section. Also, an

additional cross section must be created to fully define the forward limit of the aircraft

canopy. This is done using the preexisting cross sections and other drawings as guides.

Finally, the front view of the three-view drawing is used to construct a cross section of

the front of the aircraft cowling. Figure 2.3 shows a summary of fuselage cross sections.

New Cross Section


New Cross to Define Canopy
Section of
Cowling Front

Cabin Frame Extended

Figure 2.3: Fuselage Cross Sections as SolidWorks Sketches

12
The aft portion of the fuselage may now be made into a solid body by creating two lofts

between the aft three cross sections. This is shown below in Figure 2.4. In order to

accurately model the complex geometry of the canopy and engine cowling, additional

information is needed to direct SolidWorks to create accurate shapes. This is done by

using the top and side view from the three-view drawing as guide curves. The top and

side view drawings are imported, scaled, located, and traced in a similar fashion as the

fuselage bulkhead drawings. It should be noted that in order for SolidWorks to accurately

acknowledge the top and side view curves as guide curves, a pierce relation must exist in

SolidWorks between the guide curves and any cross sections they intersect. Figure 2.4

depicts the addition of these guide curves.

Figure 2.4: Aft Fuselage is Now a Solid Body, Front Fuselage Ready for Lofting

Once all the guide curves are in place and correctly related to the fuselage cross sections,

a loft can be created between cross sections. The completed fuselage section, shown

below in Figure 2.5, consists of five lofts total. Note that the transparency of one of the

faces in the canopy region was adjusted in SolidWorks to simulate the look of a canopy

on the aircraft.

13
Figure 2.5: Completed Fuselage Solid Body

The spinner can also be created by tracing the outline of half the spinner from the side

view of the three-view drawing and revolving the resulting spline curve about the

fuselage centerline. The spinner is added to the fuselage in a later step.

2.3 Modeling the Empennage

Construction drawings from the aircraft manufacturer are also helpful in modeling the

vertical and horizontal stabilizers. In this section, the three-view drawing of the RV-7 is

used to trace the outline of the airfoil cross section for both the horizontal and vertical

stabilizers. The cross sections depicting the root and tip are then lofted to create the solid

body. Finally, cuts are made in the solid bodies to more accurately depict the vertical and

horizontal stabilizers. Because the methods for modeling the vertical and horizontal

stabilizers are very similar and differ only in the final steps when cuts are made to the

solid body, only the details for creating one the horizontal stabilizers will be discussed.

Note that the airfoil cross sections used for the loft can also be created using JavaFoil or a

14
similar program if the type of airfoil is known. This method is used to model the wings

and is shown in the next section.

The first step in modeling one of the horizontal stabilizers is measuring the root and tip

chord lengths from the top views of the three-view drawing. Measurements may also be

taken from other construction plan drawings for the horizontal stabilizer if available. For

the root chord dimension, the lines depicting the leading and trailing edges of the

horizontal stabilizer should be extrapolated until they intersect the fuselage centerline in

the top view drawing. The root chord is measured as the length along the fuselage

centerline between these two extrapolated lines. A summary of the dimensions measured

from the top view of the three-view drawing is shown in Figure 2.6.

Fuselage Centerline

Horizontal Stabilizer
Root Location

Figure 2.6: Horizontal Stabilizer Root Chord Dimension Scheme

An image of the side view from the three-view drawing is scanned into SolidWorks and

scaled such that chord length of the horizontal stabilizer cross section corresponds to the

15
measured root chord length. The outline of the horizontal stabilizer airfoil cross section is

then traced as a spline curve from the side view of the drawing. This process is repeated

for the horizontal stabilizer tip section, but on a new plane in SolidWorks that represents

the tip location. Note that because the RV-7 horizontal stabilizer is tapered, the cross

section of the tip must be scaled and properly located in the chord direction in relation to

the root chord. This is shown in Figure 2.7.

Tip

Root

Figure 2.7: Horizontal Stabilizer Root and Tip Locations

Once the root and tip cross sections are scaled and located properly, a loft is created

between the two cross sections in SolidWorks. The horizontal stabilizer is now a solid

body, but must be trimmed in order to replicate the clearance between the elevator and

rudder on the actual aircraft. This is done by completing an extruded cut in the

appropriate shape from the aft portion of the horizontal stabilizer body. Recall that the

root chord was measured from the fuselage centerline. Therefore, an additional extruded

cut is made on the forward portion of the body to delete the portion of the horizontal

stabilizer contained within the fuselage. A summary of the final horizontal stabilizer solid

body is shown in Figure 2.8.

16
Fuselage Rudder
Clearance Clearance
Figure 2.8: Final Right Horizontal Stabilizer

This process is repeated for the left vertical stabilizer and a similar process creates the

vertical stabilizer. Depending on the method used to assemble the aircraft components, it

may not be necessary to create a separate left vertical stabilizer. If the aircraft is

assembled in assembly mode in SolidWorks, a separate left vertical stabilizer must be

created. If the aircraft is assembled in parts mode and multiple components are combined

together to make a single part, the right horizontal stabilizer and right wing may both be

mirrored about the fuselage centerline and the creation of separate components for the

right side of the aircraft is not necessary. It is possible that a similar mirror feature could

be used in assembly mode, although this was not explored. The process for creating the

vertical stabilizer differs in the airfoil cross section used, the dimensions, and the final

cuts made to the solid body. The final vertical stabilizer solid body is shown in Figure

2.9.

17
Figure 2.9: Vertical Stabilizer

2.4 Modeling the Wings

The wings of the RV-7 are modeled in SolidWorks by first importing the airfoil cross

section from JavaFoil, then extruding the airfoil cross section to create a solid body, and

finally adding a dome feature to the wingtip. The three-view drawing is used to measure

the wing semi-span and wing chord. The wing dimensions may also be available from

other sources, such as the manufacturer’s website. This process is repeated for each wing

if the parts are being combined as an assembly. Recall that if the individual aircraft

components, such as the fuselage, wing, and empennage parts, are being combined into a

single part in SolidWorks, a second wing does not need to be modeled but rather can be

mirrored about the fuselage centerline.

Tracing an outline of an airfoil cross section from a scanned image provides a quick and

reasonably accurate means to model wings or other aircraft entities. However, an

18
alternate technique is comparable in simplicity while adding a greater amount of accuracy

to the overall solid body being modeled. In this section, the free program JavaFoil is used

to create the airfoil data points, which are then scaled to the appropriate wing chord

length, and finally imported as a curve into SolidWorks.

The specific airfoil used for the RV-7 wing was obtained from the manufacturer. The

airfoil used for the RV-7 wing is a NACA 23013.5. Once the airfoil type is known,

JavaFoil is used to create a list of the airfoil coordinates. The airfoil coordinates are then

copied into Microsoft Excel and each coordinate multiplied by the wing chord length of

58 inches. In SolidWorks, the list of scaled airfoil coordinates is imported as a “Curve

Through XYZ Points.” Using the extrude feature in SolidWorks, the wing solid body is

created by extruding the imported airfoil curve. The extrude distance should correspond

to the length of a single wing. Once the wing solid body is created, a dome feature is

added to the wingtip to decrease the amount of sharp edges at the wingtip and to more

accurately portray practical wingtip geometry. The final wing solid body of the right

wing is shown in Figure 2.10.2

Figure 2.10: Right Wing Solid Body

19
2.5 Combining Individual Components to Create a Complete Aircraft

Several methods are available for combining the previously created components into a

complete aircraft. The components may be joined together as an assembly in SolidWorks,

or as multiple parts combined into one part while in SolidWorks part mode. The final

aircraft geometry used in this investigation was created using the latter method.

In SolidWorks, with the RV-7 fuselage part open, the individual aircraft components are

inserted into the fuselage part document and aligned with the fuselage. First, the right

wing solid body part is inserted and located through a series of translations and rotations

in SolidWorks. Construction drawings are referenced to determine the wing location in

relation to the fuselage as well as the wing dihedral and incidence angles. The RV-7 has a

wing dihedral of 3.5 degrees and an incidence angle of one degree. Once the right wing is

properly located in relation to the fuselage, the part is mirrored about the fuselage

centerline. This creates a left wing that is already scaled and located correctly on the

fuselage. The two wing parts are then combined with the fuselage part to create a single

SolidWorks part. By inserting a combine feature in SolidWorks as an “Add” operation,

any part of the wings that carry through to the interior of the fuselage are deleted. The

wing root shape for each wing now conforms to the fuselage and no portion of the wing

remains inside the fuselage. This simplifies the meshing process for CFD analysis.

The horizontal and vertical stabilizers are both combined to the fuselage in a similar

manner, with one horizontal stabilizer mirrored about the fuselage centerline to create the

second horizontal stabilizer. Finally, the spinner is added to the fuselage and combined

with the remainder of the aircraft to make a single part. The aircraft external geometry

20
model in SolidWorks is complete and can be saved as a variety of file types that are

appropriate for CFD use. The final SolidWorks model of the entire aircraft to be used for

wind tunnel and CFD analysis is shown in Figure 2.11.

25’-0” 20’-2”

Figure 2.11: Completed RV-7 External Geometry Model in SolidWorks

21
Chapter 3: Wind Tunnel Testing of 1/15th Scale Model

In order to conduct wind tunnel testing of the RV-7, a model was rapid prototyped from

the aircraft model created in SolidWorks. Because the rapid prototype machine operated

by the Cal Poly Mechanical Engineering Department is limited to eight by ten inch

blocks, the SolidWorks model of the RV-7 was cut into four pieces, as shown in Figure

3.1. Note that the center fuselage section is portrayed in two different orientations, but

only one center fuselage part was manufactured. Because of the difficulty associated with

accurately accounting for the 3.5 degrees of dihedral and one degree of wing incidence

incorporated by the RV-7, the center fuselage section was created with portions of the

wing already attached to the fuselage. In addition, two holes were modeled in the cross

section of the wing so that dowels could be placed in the wings for support. The center

fuselage section to be rapid prototyped also featured a flat plate modeled into the bottom

of the fuselage with six holes to accommodate the sting balance. Because of this added

feature, there is a two degree difference between the axis of the sting balance and the

longitudinal axis of the aircraft. This angle difference is taken into account in the data

processing after the test is conducted. Finally, the size of the nuts used to secure the

10/32nd size sting balance screws was measured and slots allowing sized to accommodate

the nuts were modeled into the back face of the fuselage bulkhead. The solid model was

then saved as a .stl file, imported into the rapid prototyping software, and scaled to be

1/15th the size of the actual aircraft. The wind tunnel model has a wingspan of 20 inches.

22
Figure 3.1: SolidWorks Model for Rapid Prototyping

The aft fuselage and two outboard wing sections of the model were made at the same

time, taking approximately 30 hours to be rapid prototyped. Because of the dimensions of

the center fuselage piece, it was made separately in approximately 23 hours. Each piece

came out of the rapid prototype machine fairly smooth, but some additional work was

required to get the model ready for the wind tunnel. Figure 3.2 shows the four rapid

prototyped pieces prior to assembly.

23
Figure 3.2: Four Sections of Rapid Prototyped RV-7 with Dowels Inserted

Once the model was rapid prototyped, two holes were drilled into each outboard wing

section to accommodate the wooden dowels inserted in the corresponding inboard wing

sections already attached to the fuselage. Five-Minute epoxy was then used to attach the

outboard wing sections to the center fuselage with dowel supports. Because it expands to

several times its original volume as it dries and since interior of all rapid prototyped parts

was a plastic honeycomb material, Gorilla Glue was used to attach the dowels in the

wings as well as in the fuselage. Six nuts were then inserted in the slots created by the

rapid prototype machine and located in the center fuselage over the sting balance attach

holes. The aft and foreword portions of the fuselage were then drilled with several holes

so that dowels could be inserted. The two fuselage sections were then glued together with

Five-Minute epoxy. Finally, the entire aircraft was coated with a spray-type primer and

sanded to create a smooth finish. The final wind tunnel model is shown in Figure 3.3.

24
Figure 3.3: Finished Wind Tunnel Model of the RV-7 Attached to Sting Balance

3.1 Test Apparatus and Procedure

The test apparatus was based around the Cal Poly 3 by 4 foot wind tunnel. This is of the

open cycle type in which air is sucked through by a fan in the aft portion of the tunnel

which is driven by an electric motor of approximately 100 horsepower capacity. A

calibration bar was connected to the Aerolab sting balance to calibrate the sting balance.

The RV-7 wind tunnel model was attached to the Aerolab sting balance and placed in the

test cell. The sting balance was connected to the Angle of Attack Indicator, with an LSR

of 0.1 degrees, and a Data Precision Multimeter 3600 with a LSR of one microvolt. A

Princo Instruments thermometer/barometer, with and LSR of two degrees Fahrenheit and

0.01 inches Hg, was used to record the atmospheric temperature and pressure. A Model

number 40HE35 slant manometer, with an LSR of 0.01 inches, was used in conjunction

with a SQUARED Altivar Speed Controller to measure the flow speed.

25
The experiment began by checking that the wind tunnel was clean and unobstructed. The

barometer/thermometer was used to record the atmospheric temperature and pressure.

A calibration bar was then connected to the Aerolab sting balance and the output voltages

as displayed on the Data Precision Multimeter were recorded. A one pound weight was

then added to the sting balance and the resulting output voltages were recorded. One

pound weights were then added in one pound increments up to five pounds and the

corresponding output voltages for each weight increase were recorded. The output

voltages were then plotted against the input forces and moments to determine the

conversion factor for normal, axial, and pitching forces. Next, the calibration bar was

removed from the sting balance and the strut used to attach the aircraft to the sting

balance was attached to the sting balance without the aircraft attached to it. The axial

force, normal force, and pitching moment were recorded at angles of attack ranging from

negative eight degrees to positive 18 degrees in increments of two degrees up to an angle

of attack of 10 degrees, and in one degree increments from 10 degrees to 18 degrees. The

wind tunnel was then turned on to a flow speed of 30 meters per second, which was

verified from a dynamic pressure value of 2.177 inches of water using the Mariam Model

40HE35 inclined manometer and the same forces were recorded. Next, the RV-7 model

was mounted to the sting balance using the strut and the process was repeated, with

forces being collected with the wind tunnel off and then the wind tunnel on.

After all the data was collected and transferred into an Excel spreadsheet, the values of

lift coefficient were calculated and plotted against their corresponding angle of attack.

Next, the drag coefficients were calculated and plotted against their corresponding angle

of attack. Because the Cal Poly Motion Simulator references the aircraft center of gravity

for moment calculations, the moment coefficients about the center of gravity were then

26
calculated and plotted against their corresponding angle of attack. Because the RV-7 is a

homebuilt aircraft in which the builder has a wide variety of options available to

customize the aircraft, the center of gravity ranges for each aircraft individual will vary.

Therefore, an average center of gravity location of 72 inches aft of the aircraft spinner is

used. Applying the 1/15th scale factor of the wind tunnel model, this translates into 4.8

inches. 3, 8

Because the axis of the root chord of the model is not parallel to the sting balance axis,

the angles of attack recorded during the experiment were relative to the sting balance

axis. The angle difference between the sting balance axis and the root chord of the model

was measured with an inclinometer and found to be approximately 2 degrees. This value

was added to all angles of attack in this report to obtain the corresponding angle of attack

of the model.

3.2 Wind Tunnel Analysis

Before any calculations could be completed, the data obtained needed to be converted to

useful information. Using fixed amounts of applied load and measuring the

corresponding readings yielded by the sting balance voltmeter in volts, a calibration

curve used to reference readings directly to pound force or foot pound forces for

moments. The following equation is used to convert sting balance readings to forces:

N = RN E N [2.1]

27
where N is Normal, Axial, or Pitching. RN is the slope from the sting balance calibration

and EN is the reading reported by the sting balance in volts. With this, the forces on the

model were found with,

[ ]
N wing = N wing + strut −30 m / s − N wing + strut −0 m / s − [N strut −30 m / s − N strut −0m / s ] [2.2]

Nwing is the normal force on the aircraft due to the free stream velocity. Nwing+strut is the

normal force on the aircraft and strut at 30 meters per second and at 0 meters per second.

Nstrut is the normal force due to the free stream velocity on the strut at 30 meters per

second and 0 meters per second. A similar equation is used for the axial forces and

pitching moments.

With the normal, axial, and pitching moment forces converted to pounds, the lift, drag,

and moment about the aircraft center of gravity are calculated using the equations below.

L = N cos α − A sin α [2.3]

D = N sin α + A cos α [2.4]

M CG = M MC − N l N ,CG − Al A,CG [2.5]

Where L is the lift, D is the drag, N is the normal force, and A is the axial force all in

pounds. The angle of attack in degrees is α. MCG is the moment about the aircraft center

of gravity and MMC is the measured moment. l N ,CG is the moment arm in the normal

direction and l A,CG is the moment arm in the axial direction, both referencing the

approximate center of gravity location of the aircraft.

28
The atmospheric conditions recorded during the test must be taken into account prior to

calculation of lift, drag, and moment coefficients. The Perfect Gas Law is used to

calculate the density, which is used in the calculation of dynamic pressure.

p
ρ= [2.6]
RT

The pressure (p) is measure in inches H2O of Mercury which is then converted pounds

per square foot to obtain standard units for the equation. Ambient temperature (T) taken

from the thermometer is in Fahrenheit and is also converted to Rankine to satisfy the

units of the Gas constant (R) of foot times pounds per slugs per degrees Rankine.

The calculated density is substituted into the dynamic pressure equation to obtain the last

necessary portion of the coefficient of lift, drag and moment equations listed below. The

dynamic pressure is obtained by multiplying the previously found density (ρ) in slugs per

feet cubed by one half and the free stream velocity (U∞) squared in feet per second.

q∞ = ½ρU ∞2 [2.7]

The coefficient of lift and drag are calculated from the lift and drag, wing surface area,

and the dynamic pressure. Lift and drag coefficients are calculated using the following

equations

L
CL = [2.8]
q∞ S

29
D
CD = [2.9]
q∞ S

Cl is the coefficient of lift and Cd is the coefficient of drag. L is lift and D is drag both in

pounds. The dynamic pressure in pounds per foot squared is q∞ , and S is the wing

planform area in square feet.

The moment coefficients are calculated in a similar fashion, but the denominator is

multiplied by an additional term, the mean aerodynamic chord. Because the wing of the

RV-7 has no taper, the mean aerodynamic chord ( c ) is equal to the wing chord. The

moment coefficient is found using the following formula:

M CG
CM ,CG = [2.10]
q∞ Sc

3.3 Wind Tunnel Results

The sting balance was calibrated to find a conversion factor between volts and pounds for

normal, axial, and pitching forces. The conversion factor is based on the slope of the

resulting curve of the voltage plotted against calibration load. The conversion plot and the

corresponding conversion factors are shown in Figure 3.4.

30
2.0000

1.8000 y = 0.313x

1.6000
Sting Balance Reading (V)

1.4000

1.2000
Pitching
1.0000 Axial
Normal
0.8000

0.6000
y = 0.0948x
0.4000
y = 0.054x
0.2000

0.0000
0 1 2 3 4 5 6 7
Load (lbf)

Figure 3.4: Sting Balance Calibration Plot

The resulting lift coefficient versus angle of attack found from the wind tunnel

experiment is shown below in Figure 3.5. The wind tunnel data indicates that the aircraft

stalled at an angle of attack of approximately 14-15 degrees. The corresponding

maximum lift coefficient was approximately 1.146. The zero lift angle of attack of -2

degrees for the entire aircraft corresponds to the zero lift angle of attack of the airfoil

when analyzed using JavaFoil. The variability of the lift coefficient values at angles of

attack beyond the stall angle show the expected variability of data collected at high

angles of attack.

31
1.4

1.2

1.0

0.8

0.6

CL
0.4

0.2

0.0
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
-0.2
α
-0.4

-0.6

Figure 3.5: Lift Coefficient versus Angle of Attack for 30 m/s Flow

The drag coefficient at various angles of attack found from the wind tunnel experiment is

shown in Figure 3.6. An increase in the drag coefficient trend at the stall angle was

observed and can be seen in Figure 3.6. This is expected because of the increased amount

of flow separation when the aircraft is in a stalled condition. Drag is also increased

because as lift increases with angle of attack, the induced drag of the aircraft also

increases.

32
0.45

0.40

0.35

0.30

0.25
CD
0.20

0.15

0.10

0.05

0.00
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
α

Figure 3.6: Drag Coefficient versus Angle of Attack for 30 m/s Flow

The moment coefficient about the aircraft center of gravity is shown in Figure 3.7. The

angle of attack at which the moment coefficient about the center of gravity equaled zero

was approximately negative one degree. This makes sense because the NACA 23013.5

airfoil has a positive camber. The negative slope of the moment curve indicates the RV-7

exhibits positive pitch stability, which is expected for a homebuilt general aviation

aircraft. The figure also indicates the maximum pitching moment of approximately -0.85

occurred at the highest angle of attack measured, which also makes sense.

33
0.2

0.1

0.0
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
-0.1 α

-0.2
Cm,cg

-0.3

-0.4

-0.5

-0.6

-0.7

Figure 3.7: Moment Coefficient versus Angle of Attack for 30 m/s Flow

The wind tunnel analysis of the 1/15th scale RV-7 model provides an initial data set for

lift, drag, and moment coefficients. The complete data tables from the wind tunnel test

can be found in Appendix B. A summary of the coefficient values found from wind

tunnel testing is shown in Appendix B. In a later chapter of this report, an error analysis

will be conducted on the data collected from wind tunnel testing. The wind tunnel data

will then be compared to values of lift, drag, and moment coefficient obtained from CFD.

34
Chapter 4: CFD Analysis Using GAMBIT and FLUENT

In order to obtain initial estimates of lift, drag and moment coefficients of the RV-7 using

CFD, the aircraft geometry first must be meshed into elements using a preprocessor. In

addition, a flow volume containing the meshed aircraft surfaces must be created in the

preprocessor. For this analysis, the preprocessor GAMBIT 2.2.30 was used. In addition,

to ensure that the mesh created is sufficient to accurately model the flow behavior around

the aircraft at various flight conditions, a grid independence study was conducted. This

chapter discusses the process of creating a mesh of the aircraft and flow volume,

checking the grid independence in FLUENT, and the steps taken using the CFD program

FLUENT 6.2 to obtain lift, drag, and moment estimates at various flight conditions.

Because the purpose of this thesis is not to explore the inner workings of CFD programs,

but rather to provide detailed methods for obtaining usable data from CFD, the specifics

of the mathematics solved by FLUENT will not be discussed in this thesis. However,

more information on this subject can be found in the FLUENT Users Guide and the

general approach used by FLUENT is discussed below.10

FLEUNT solves the Navier-Stokes equation to obtain the results found in this report.

FLUENT also solves the Euler equations, which is more appropriate for inviscid flows

and will not be discussed since this report deals with viscous flows. The Navier-Stokes

equation includes expressions for the conservation of mass, momentum, pressure, species

and turbulence. Because the flow conditions explored in this report are incompressible,

the compressible form of the Navier-Stokes equations is not shown. For incompressible

flow, the Navier-Stokes equation solved by FLUENT to obtain the results in this paper is

35
written as a continuity equation (eq. 4.1) and as a momentum equation (eq. 4.2). Note that

the equations shown below are the incompressible Navier-Stokes equations solved by

FLUENT.14
v
∇ U = 0 [4.1]

v
∂U vv v
+ ∇  (UU ) = −∇p + ∇  (ν∇U ) [4.2]
∂t
v
In the above equations, U is the flow velocity, p is the static pressure, ν is the

viscosity, and ∇ represents the partial derivative of a quantity with respect to all

directions in the specified coordinate system.14

4.1 Geometry Import and Setup

In order to transform the aircraft geometry into a series of meshed elements to be read by

FLUENT, the solid model of the aircraft geometry created in SolidWorks must first be

exported as either a .STEP, .ACIS, or .IGES file. Because computing time can be

decreased by only analyzing half the aircraft geometry, the solid model created in

SolidWorks was first cut in half so that only the left half of the aircraft existed. For this

analysis, the half aircraft geometry was saved in SolidWorks as a .STEP file, and then

imported into GAMBIT. The aircraft geometry used for CFD analysis is shown in Figure

4.1. Note that in order to accurately model the wing/fuselage interface, the wing root

conforms to the aircraft fuselage and does not carry through into the fuselage, as is the

case with the actual aircraft wing. Once the aircraft geometry was imported into

GAMBIT, several steps were taken to clean up the aircraft geometry in order to simplify

36
the meshing process. Using the “clean up small faces” command in GAMBIT, several

small faces contained within the aircraft geometry were deleted and merged with existing

faces, thus creating new face entities. This is done to simplify the meshing process and

decrease the possibility of skewed mesh elements. A high number of skewed, or

flattened, elements in the mesh may result in a diverging solution in FLUENT and thus

should be avoided.9

Figure 4.1: Half Aircraft Geometry used for CFD and Corresponding Clean

Geometry in GAMBIT

Once the aircraft geometry was simplified by merging and creating new faces, a flow

volume was created in GAMBIT using the “create brick” command. The brick size (in

inches) was (600,600,2320) using Cartesian Coordinates, with the negative X direction

along the wingspan direction, positive Y pointing in the lift (upwards) direction, and

positive Z in the flow direction, or direction of the wing chord. This coordinate system

can be seen in Figure 4.2. To ensure that the flow at the outlet boundary condition is

close to free stream, the flow volume brick is created such that the flow exit face is

approximately 20 chord lengths behind the aircraft. Boundary conditions will be

37
discussed in greater detail later in this chapter. The volume brick was then aligned with

the symmetry plane of the aircraft. In order to merge the aircraft volume with the flow

volume, the brick volume is first deleted while retaining the brick faces. Using the

“connect faces” command in GAMBIT, the symmetry plane of the aircraft is attached to

the corresponding face of the brick. Finally, using the “stitch faces” command, the flow

volume that now incorporated the aircraft geometry is created. This is shown in Figure

4.2. The geometry of both the aircraft and the flow volume is now ready for the meshing

process. 11

Figure 4.2: Aircraft and Flow Volume Geometry in GAMBIT

4.2 Meshing the Aircraft Body Surfaces

Once the aircraft geometry is simplified to reduce element skewness and the flow volume

is created, the external aircraft faces are ready for meshing. Creating a quality mesh with

few skewed elements and appropriate element spacing is critical in achieving accurate

38
results in FLUENT. The meshing process is iterative and not trivial, and will vary for

different geometries. The steps outlined in this section represent the steps taken to

achieve a quality mesh for the RV-7 geometry that contains sufficient elements to

accurately model the flow behavior around the aircraft. The basic concepts presented are

applicable to similar geometries and should be used as a guide.

Size functions are used to accurately portray the RV-7 external geometry as a series of

mesh elements. A size function allows the user to control mesh intervals on faces or

edges and element spacing on faces and volumes. Size functions achieve this by

controlling the characteristics of the mesh in the area of the entities to which they are

attached. In order to place a higher density of mesh elements in areas of the aircraft

geometry where the flow characteristics are rapidly changing without reducing the

overall size of all the mesh elements, a curvature size function is attached to the faces of

the aircraft geometry. A curvature size function allows the user to place a limit on the

angle allowed between outward facing normals between any two adjacent mesh elements

adjacent to the surface of the source edge, face, or volume.12

The user has several options when creating a curvature size function. The source, or

entities to which the size function will be applied to, must be specified. For the RV-7

geometry, the sources for the curvature size function include all the faces that constitute

the external aircraft geometry. Next, the user specifies the entities in which the size

function will be attached. Once again, all of the aircraft faces should be included as

attachment entities. It should be noted that the symmetry face of the aircraft geometry

previously used to attach the aircraft faces to the symmetry plane of the flow volume

should not be included as either a source or attachment entity.

39
Once the source and attachment entities are specified for the curvature size function, the

angle allowed between outward facing normals, growth rate, and size limit of the

elements are specified. Using the GAMBIT aircraft geometry tutorial as a guide, an angle

of 15 degrees was specified for the curvature size function. The growth rate defines the

rate at which the mesh element edge length increases with each layer of elements from

the attachment edge or face. For example, a growth rate of 1.1 represents a 10% increase

in mesh element edge length with each succeeding layer of mesh elements. To ensure that

a sufficient amount of mesh elements are placed in areas of rapidly changing flow

conditions such as the wing leading edge, a growth rate of 1.1 was specified for the RV-7

geometry. The size limit specification allows the user to control the maximum mesh

element size to be used on the attachment entities. For the RV-7 geometry, the size limit

on the curvature size function was set at 2. Finally, the “Apply” button in GAMBIT

applies the curvature size function to the specified aircraft faces. It should be noted that

the exact values specified for the curvature size function may vary depending on the

quality of mesh produced by the size function. For different geometries, it may be

necessary to set different values for angle, growth rate, and size limit in order to achieve a

mesh without highly skewed elements.11

Once the appropriate size functions are specified for the aircraft geometry, the “Mesh

Command Button” in GAMBIT allows the user to specify several final options before the

aircraft is meshed. Clicking on the “Face Command” button under the “Mesh Command

Button” menu in GAMBIT brings up the menu for meshing faces. The user first specifies

the faces to be meshed. Note that the face defining the aircraft symmetry plane should not

be included in the list of faces to mesh because it is not part of the actual aircraft

40
geometry and thus does not need to be accounted for. It is helpful to turn off the faces

defining the flow volume when selecting the faces to be meshed. This is accomplished by

selecting the “Specify Model Display Attributes” in GAMBIT and turning off the

“visible” option for the flow volume first, and then the six faces associated with the flow

volume.

Once the aircraft faces are selected for meshing, the user specifies the type of meshing

scheme to be used. The options available are Quad, Tri, Quad/Tri, and Unspecified.

Going into great detail about each meshing scheme is beyond the scope of this thesis.

However, in general one should choose a scheme that allows for a quality mesh (one

without highly skewed elements) while accurately placing a higher density of elements in

regions where the flow conditions are rapidly changing. The specific combination of

appropriate size functions and meshing schemes will vary for different geometries and

the user should explore these combinations until a desirable mesh is achieved. For the

RV-7 case, it was found after exploring the other options that a Tri meshing scheme

produced the highest quality mesh while placing a higher concentration of elements in

areas of rapidly changing flow conditions. Once the Tri meshing scheme is chosen, the

“Apply” button takes into account the previously created size functions and applies the

specified mesh to the aircraft faces. The resulting aircraft mesh is shown in Figure 4.3.

41
Figure 4.3: Aircraft Faces Meshed in GAMBIT

4.3 Meshing the Flow Volume

There are several options in GAMBIT for meshing the flow volume once the aircraft

faces have been meshed. The first option involves attaching a “mesh” size function to the

aircraft and flow volume symmetry planes and meshing the symmetry plane of the flow

volume. The flow volume is then meshed in a separate step. The alternate to this

approach involves attaching a “meshed” size function to the flow volume and using the

meshed aircraft faces as source entities. The first option allows the user to have more

control over the distribution of mesh elements immediately adjacent to the aircraft face

entities. According to the GAMBIT Tutorial 14 involving meshing aircraft geometry, the

second option produces a smoother distribution of mesh elements but significantly

42
increases the meshing time and number of elements used. Although it is desirable to

achieve the highest quality mesh with the least amount of elements, after multiple trials

utilizing the first option, it was found that a higher quality mesh could be obtained by

employing the second option.11

To mesh the flow volume using the second option mentioned above, a “meshed” size

function is created. The faces that define the aircraft external geometry are specified as

the size function source, and the flow volume is specified as the attachment entity. Next,

the user must specify the growth rate and size limit associated with the size function. In

order to minimize the total number of elements used, the maximum element size should

be quite large. In addition, there is no need to have small mesh elements in areas of the

flow volume that experience little or no change, such as regions far from the aircraft

surfaces. For these reasons, a size limit of 150 was specified. The growth rate

specification should be chosen such that the elements are allowed to increase in size from

the meshed aircraft surface at a reasonable rate while maintaining a fine enough mesh

near the aircraft surfaces to accurately capture flow effects in those regions. However, if

the growth rate is too small, an unnecessary amount of elements will be created in the

region adjacent to the aircraft. The effects on the final solution of varying the growth rate

are discussed in the grid independence section. A growth rate of 1.2, or 20%, was

specified for the meshed size function. The “Apply” button in GAMBIT applies the mesh

size function.13

Once the appropriate size functions are created, the flow volume is ready to be meshed.

The “Volume Command Button,” found under the “Mesh Command Button” in

GAMBIT, brings up the menu for applying volume meshes. The available meshing

43
schemes include Hex, Hex/Wedge, and Tet/Hybrid. Detailed information on each of the

meshing schemes can be found in the GAMBIT Modeling Guide, Section 3.4. As with

meshing the aircraft faces, it is important to choose a meshing scheme that will accurately

portray the flow volume with the least amount of elements while providing a high quality

mesh. After exploring the various volume mesh options, the Tet/Hybrid scheme of the

TGrid type was found to be most effective for the RV-7 case. After selecting the flow

volume as the entity to be meshed, the “Apply” button creates the specified volume mesh.

Even with a large size limit of 150, 1,627,345 total elements were produced for the RV-7

mesh. Figure 4.4 depicts the meshed flow volume.13

Figure 4.4: Aircraft Faces and Flow Volume Meshed

Once the flow volume mesh is successfully created, the “Examine Mesh” command may

be used to examine the mesh elements and mesh quality. It is important to examine the

skewness of mesh elements to ensure that there are no highly skewed elements present in

44
the mesh. Recall that highly skewed elements may cause solution convergence problems

in FLUENT. Although a highly skewed element in GAMBIT is one that has a skewness

greater than 0.97, it is generally favorable to have a maximum skewness of 0.85-0.9. The

maximum skewness for the RV-7 mesh is 0.85. To examine the element skewness in

GAMBIT, the “Examine Mesh” subpad is used. By clicking on the range option and

specifying the skewness range to be examined, the user can locate any mesh elements

that have a skewness within the specified range. GAMBIT provides the number of

elements within the specified skewness range and highlights only those elements in the

graphics window. In addition to checking skewness, the “Examine Mesh” menu may be

used to examine the mesh from various perspectives. Figure 4.5 shows a representative

slice of the final mesh. Note the density of cells near the aircraft surface and the large size

of the cells far away from the aircraft.14

Figure 4.5: Representative Slice of Final Mesh in GAMBIT

45
After examining the mesh for skewed elements, boundary conditions may be set in

GAMBIT. A velocity inlet boundary condition is used to define the flow velocity at flow

inlets, as well as the scalar properties of the flow. The front face of the flow volume is set

as a velocity inlet boundary condition to specify the flow properties coming into the

volume domain. For velocity inlet boundary conditions, the flow stagnation properties are

not fixed and thus will rise to the necessary value to provide the specified flow

conditions. Therefore, in addition to the front face, the left, top, and bottom faces are also

specified as velocity inlets in order to ensure that the flow properties were close to free

stream in areas of the flow volume that are far from the aircraft surfaces. It should be

noted that velocity inlet boundary conditions are meant for use with incompressible

flows. Because all flow Mach Numbers to be examined for the RV-7 are below 0.3, the

use of velocity inlet boundary conditions is warranted.10

The exit plane of the flow is specified as an outflow boundary condition. A pressure

outflow boundary condition may also be used to specify the exit plane, but is more

sensitive to the flow having returned to free stream conditions. The outflow boundary

condition is used when the details of the flow velocity and pressure are unknown at the

exit boundary. As with velocity inlet boundary conditions, the outflow boundary

condition is suited only for incompressible flow.10

The aircraft surface is specified as a wall boundary condition. Wall boundary conditions

are used to differentiate between solid and fluid regions. For viscous flows, such as those

investigated in this report, by default a no-slip boundary condition is enforced at the

walls. Finally, the symmetry plane of the flow volume is specified as a symmetry

46
boundary condition. A summary of the boundary conditions used for the RV-7 case is

shown in Figure 4.6.

Figure 4.6: Boundary Condition Summary

A discussion of the calculations associated with each boundary condition is beyond the

scope of this report. Information on the calculations used for each boundary condition

may be found in Chapter 7 of the FLUENT Users Guide.10

Once the boundary conditions have been specified in GAMBIT, the entire mesh is ready

to be exported. Using the File-Export As command and selecting “Export Mesh,” the

mesh is saved as a .msh file and is ready for import into FLUENT.

47
4.4 FLUENT 6.2 Case Setup

To begin a case in FLUENT, the program is initiated and the user is asked to choose a

solver type. The options are 2D single-precision, 2D double-precision, 3D single-

precision, and 3D double-precision. The FLUENT Users Guide states that “For most

cases, the single precision solver will be sufficiently accurate.” A list of possible

applications that would benefit from a double-precision solver does not include anything

that resembles the type of case for the RV-7. Therefore, a 3D single-precision solver is

used for all RV-7 cases in this report.10

Once the solver is chosen, the program loads and is ready to read in a mesh file. Under

“File-Read-Case”, the user can find the appropriate .msh file and FLUENT imports the

previously created mesh (grid). Once the grid is read into FLUENT, a grid check should

be ran to ensure that there are no negative volumes contained within the grid as FLUENT

is unable to provide a solution to a grid with negative volumes. Because the grid was

created in inches and the default grid setting in FLUENT is in meters, the grid must be

scaled to inches using the “Grid-Scale” command.

4.4.1 Defining the Models, Materials, and Boundary Conditions

The next step in setting up a case to run in FLUENT is to choose the appropriate models

FLUENT will run from the “Define-Models” menu. The solver should be set as

“Segregated.” The Segregated solver in FLUENT solves the governing equations

sequentially and is traditionally used for incompressible flows. Under the “Solver” menu,

the “Time” option should be selected as “Steady.” Modeling the flows as steady-state

48
allows pertinent flow features to be captured while not placing an excessive

computational load on the computer.10, 14

The Spalart-Allmaras turbulence model is used for all flow cases because of its simplicity

and applicability to the problem. This simple, one- equation model was created for

aerospace applications and solves a modeled transport equation for the turbulent

viscosity. In this model it is not necessary to calculate a length scale related to the local

shear layer thickness, thus simplifying the case setup process. According to the FLUENT

Users Guide, the Spalart-Allmaras model is also the best choice for coarse meshes that

may not completely represent a region of turbulent flow.10

The Spalart-Allmaras turbulent model was originally a low-Reynolds-number model

which required the region of the flow affected by viscous effects to be fully resolved. In

FLUENT, however, the turbulence model has been changed such that wall functions are

used when the mesh resolution is not sufficient to capture the viscous effects of rapidly

changing flow conditions, such as separated flow. Therefore, the Spalart-Allmaras

turbulence model may not accurately capture the effects of massively separated flow

found at high angles of attack. According to the FLUENT User’s Guide, the Spalart-

Allmaras turbulence model may not be capable of accurately capture the effects of

rapidly changing length scales associated with highly turbulent flow. Despite the

limitations imposed on massively separated flow, the Spalart-Allmaras equation is

suitable for the majority of flight conditions tested and a discrepancy between CFD and

wind tunnel results at high angles of attack may be expected. The Spalart-Allmaras model

is selected under the “Define-Models-Viscous” menu in FLUENT. 10

49
Once all the appropriate models are specified, the user must ensure that the fluid material

used by FLUENT is air. This is found under the “Define-Materials” menu. The

appropriate values for density and viscosity should be entered under the “Material

Properties” section of the menu. Unless otherwise stated, these values corresponded to

standard atmosphere sea level conditions for the CFD runs presented in this thesis.

The desired flow velocity and angle of attack for a given case are specified under the

“Define-Boundary Conditions” menu. For each velocity inlet boundary condition

specified in GAMBIT, the velocity magnitude and direction is specified in FLUENT. For

example, to run a FLUENT case for 30 meters per second flow at six degrees angle of

attack, the magnitude is specified as 30 meters per second and the components of the

flow are specified as cos(6◦) in the Z-direction and sin(6◦) in the Y-direction. Refer to

section 4.1 of this report for the axes previously specified in GAMBIT for the RV-7 case.

There are no additional variables to change within the “Boundary Conditions” menu once

the velocity magnitude and direction are specified for each velocity inlet.

4.4.2 Solution Initialization and Setup

Before a case can run in FLUENT, the solution must be initialized to provide an initial

guess for the solution flow field. To begin the initialization process, the governing

equations must first be converted to algebraic equations that can be solved numerically.

This is accomplished through a control volume technique that involves integrating the

governing equations about each control volume to yield discrete equations from which

each quantity is conserved on a control volume basis. A mathematical example of this

can be found in section 26.2 of the FLUENT Users Guide. For the RV-7 case, the

50
pressure, momentum, and modified turbulent viscosity equations must be discretized

under the “Solve-Controls-Solution” menu in FLUENT. “Pressure” should be left at the

default setting of “Standard” while both “Momentum” and “Modified Turbulent

Viscosity” should be set as “Third-Order MUSCL.” The MUSCL scheme is applicable to

arbitrary meshes and provides improved spatial accuracy over a second-order upwind

scheme. All other parameters under the “Solution Controls” menu should be left at the

default settings. 10

Once the solution controls have been specified, the solution is ready to be initialized.

Under the “Solve-Initialize-Initialize” menu in FLUENT, the user specifies where the

solution is to be initialized from. For the RV-7 cases, the solution is computed from the

inlet face. Once this is specified, the “Initial Values” field automatically updates to

include all previously specified values for Gauge Pressure, Velocity Components, and

Modified Turbulent Viscosity. Clicking the “Init” button initializes the solution.

Several steps remain before the case is ready to be run. Under the “Solve-Monitors-

Residual” menu, the convergence criterion for continuity and x, y, and z velocity are

specified as 0.0001. Next, under the “Solve-Monitors-Force” menu, the force vectors

corresponding to the angle of attack for the particular case are specified. For example, for

a six degree angle of attack case, the drag force monitors should be set as Y= 0.10453,

which is sin(6◦), and X= 0.99452, which is cos(6◦). Similarly, for lift the Y-value should

be set as Y=0.99452 and the X-value should be set as -0.10453. Because pitching

moment is also required, the moment center should be specified as well as the axis about

which the moment will be taken. Taking into account the previously specified coordinate

system, the pitching moment is about the x-axis. For each force monitor, the “Print” and

51
“Plot” options should be selected. In addition, the aircraft body should be selected under

“Wall Zones.” This informs FLUENT to calculate the forces from the aircraft body.

The final user inputs prior to running a case in FLUENT are the reference values from

which the forces will be calculated. Under the “Report-Reference Values” menu, the

wing area should be specified as the reference area. In addition, the reference length

should be the wing chord length (or other reference length chosen by the user) and the

reference velocity should be set to match the velocity that the current case will be run at.

Unless otherwise stated, the reference velocity for all RV-7 cases is 30 meters per second.

To begin iterations of the case, the “Solve-Iterate” menu is opened and the user specifies

the number of iterations to be run. A high number, such as 1,000 iterations, should be

specified at first to ensure enough iterations for solution convergence. Clicking on the

“Iterate” button begins the iterative process. FLUENT will automatically stop iterating

when either the previously specified convergence criterion are met or the maximum

number of specified iterations occur. By monitoring the force values both numerically

(by the “Print” command) and visually (by the “Plot” command) during the iterative

process, the user can determine if it is feasible to continue iterations prior to FLUENT

automatically stopping the process. For example, if none of the force values are changing

significantly between iterations and appear to have leveled out, the user may choose to

stop the iterative process prior to solution convergence to save computing time.

A summary of the actions taken to setup a case in FLUENT is shown in Table 4.1. Note

that once a case has been setup in FLUENT, the case may be saved so that only steps 6-7

and 9-12 need to be addressed for each new case to be ran. In addition, after each case is

52
run, the case and data files should be written in order to save all the information

pertaining to the case.

Table 4.1: FLUENT Setup Summary

4.5 Grid Independence Study

A grid independence study was conducted to ensure that the grid used for the RV-7

geometry was sufficient to accurately model the flow around the aircraft. The grid used

for data collection of the RV-7 contained 1,627,345 mesh elements, with a concentration

of elements in areas of rapidly changing flow conditions. Even though this is a relatively

high number of elements for the given case, an additional mesh was created in GAMBIT

to observe the effects of increased mesh density on the solution.

To begin the grid study, the original grid used for all RV-7 cases was opened in

GAMBIT. The growth rate of the “mesh” size function attached to the flow volume was

then adjusted to a growth rate of 1.05, or 5%. A small growth rate such as this ensures

53
that a very dense mesh is created in areas where the flow volume meets the aircraft

surface mesh. Therefore, an even greater amount of elements are placed in areas of

rapidly changing flow conditions than the original mesh. The element size limit of 150

was retained to maintain a minimum number of elements in flow areas far from the

aircraft. The re-meshing of the flow volume with the new “mesh” size function resulted

in a final mesh containing 2,608,255 elements, with a worst skew of only 0.829.

Therefore, by reducing the growth rate from 10% in the original grid to 5% for the new

case, an additional 980,910 mesh elements were created.

A FLUENT case was setup and ran to determine the effect of the increased mesh density

near the aircraft body on values of lift, drag, and moment coefficient. Because the effects

of mesh density become increasingly important at higher angles of attack when the flow

is changing rapidly over short distances, the test case was run at a 10 degree angle of

attack. Table 4.2 provides a summary of the grid study. It is clear from this study that the

original mesh used for the RV-7 cases at various angles of attack sufficiently captures the

various flow characteristics associated with turbulent flows.

Table 4.2: Grid Study Summary

54
4.6 FLUENT Results

A number of cases were run in FLUENT to determine the lift, drag, and moment

coefficients of the RV-7. Each case was run at the same speed as the wind tunnel: 30

meters per second. Angles of attack ranged from 0 to 17 degrees in two degree

increments from 0 to 10 degrees, and one degree increments from 11 to 17 degrees in

order to collect more data at higher angles of attack. In addition, a run was conducted at

six degrees angle of attack and 96.1136 meters per second, or 215 miles per hour. This is

the published maximum airspeed of the RV-7. The maximum airspeed case was run in

order to determine if the RV-7 flies at such a wide range of airspeeds as to significantly

alter the lift and drag coefficient values throughout the flight envelope. 2

Figure 4.7 shows the lift curve for the RV-7 cases run at 30 meters per second. The data

indicates the onset of stall at an angle of attack of approximately 14 degrees. Data

collected at pre-stall conditions follows a very linear trend. The data also indicates that at

a zero degree angle of attack, the RV-7 has a lift coefficient of approximately 0.18. This

makes sense because the RV-7 employs a NACA 23013.5 non-symmetric airfoil.

55
1.4

1.2

1.0

0.8
CL

0.6

0.4

0.2

0.0
0 2 4 6 8 10 12 14 16 18
α

Figure 4.7: Lift Curve from FLUENT Data

The drag coefficient is plotted against angle of attack in Figure 4.8. At zero angle of

attack, the drag coefficient is 0.027, or 270 drag counts. The value seems reasonable for

the RV-7 configuration because it is a relatively “clean” aerodynamic design and also

because the effects of the propeller and landing gear were not taken into account. The

drag coefficient continues to rise as angle of attack increases, which is expected. Because

the Spalart-Allmaras model used is a relatively simple turbulence model and thus may

not accurately model highly turbulent flows, the values of drag coefficient at stall and

post stall angles of attack may be lower than the actual values. This does not take away

from the main goal of populating the RV-7 flight simulator database since the majority of

flight simulator flights will take place at pre-stall angles of attack.

56
0.25

0.20

0.15
CD

Approximate
0.10 Stall Onset

0.05

0.00
0 2 4 6 8 10 12 14 16 18
α

Figure 4.8: Drag Curve from FLUENT Data

Figure 4.9 shows the moment coefficient plotted against angle of attack for the FLUENT

runs. The data indicates an increase in moment coefficient as angle of attack increases.

The data follows a fairly linear trend until around the stall angle of 14 degrees. At this

point the moment coefficient begins to decrease at a slightly lower rate than at pre-stall

angles of attack. The data indicates a moment of approximately -0.03 at a zero degree

angle of attack, which makes sense since the RV-7 uses a positive-camber NACA

23013.5 airfoil. The results of the 30 meters per second runs are summarized in Table

4.3.

57
0.2
0.1
0
-2 -0.1 0 2 4 6 8 10 12 14 16 18

-0.2
Cm,cg

-0.3
-0.4
-0.5
-0.6
-0.7 α

Figure 4.9: Moment Curve from FLUENT Data

Table 4.3: FLUENT Results Summary 30 Meters per Second

58
As previously mentioned, a case was run at 96.1136 meters per second and six degrees

angle of attack in order to determine the affects of airspeed on lift, drag, and moment

coefficients. Because the flight simulator tables developed by the Cal Poly Flight

Simulation Team do not take Reynolds Number into account, only sea level data needed

to be collected. The case was not run at an angle higher than six degrees because the

aircraft will be unable to fly at high angles of attack at the aircraft maximum airspeed due

to structural limitations. Therefore, analysis of higher angles of attack at maximum

airspeeds would be inappropriate. Table 4.4 shows a summary comparison of the

maximum airspeed case versus the wind tunnel case. A 30 meters per second airspeed

(~67 miles per hour) is only 15% higher than the published stall speed of 58 miles per

hour for the RV-7. Therefore, because the difference in both lift and drag coefficients are

minimal over the majority of the flight simulator envelope, the values for lift and drag

coefficient obtained from the 30 meters per second cases will be used to populate the

flight simulator tables. However, the percent difference between the data will be used as

part of the error band placed on the CFD data in the following chapter. 2

Table 4.4: Airspeed Effects Summary

59
Chapter 5: Wind Tunnel and CFD Data Comparison

The data collected from both wind tunnel and CFD testing is compared in order to

determine what data should be used to populate the flight simulator tables for the Cal

Poly Motion Simulator. The similarities and differences between the data sets, as well as

an estimate of the errors associated with the data, are examined.

5.1 Error Estimates

Prior to comparing the wind tunnel and CFD data, it is imperative that possible errors in

the data be recognized and accounted for. An error analysis was conducted on the wind

tunnel data. The most probable error method was employed to predict the maximum error

on lift, drag, and moment data. The error percentages for the CFD data were based on the

percent change in the data between the initial case and both the grid test case and the

velocity study case.

5.1.1 Wind Tunnel Error Calculation Procedure

The most probable error method used on the wind tunnel data is outlined in this section.

Due to the similarity of methods used, the most probable error method for lift coefficient

will be shown in its entirety while the drag coefficient error method will only be partially

shown. Recall that the error calculated from this method represents the maximum error

associated with each data point. The moment coefficient error method follows a similar

procedure. To begin, partial derivatives are performed on the subsequent equations for

coefficient of lift and drag and then divided by the equation to produce the results listed.

60
L
CL = [5.1]
q∞ S

D
CD = [5.2]
q∞ S

∆C L ∆L ∆q∞ ∆S
= − − [5.3]
CL L q∞ S

∆C D ∆D ∆q ∞ ∆S
= − − [5.4]
CD D q∞ S

The most probable error function is completed by squaring each term of the preceding

products and taking the square root of the sum of those elements. This is done in order to

locate the most likely error range for each equation due to the least scale readings of the

equipment. The delta terms are then replaced with the least scale readings to reflect the

finite value differences represented while the denominator is substituted for the measured

value at the given angle of attack.

2 2 2
∆C L  ∆L   ∆q∞   ∆S 
=   +   +   [5.5]
CL  L   q∞   S 

2 2 2
∆C D  ∆D   ∆q∞   ∆S 
=   +   +   [5.6]
CD  D   q∞   S 

61
Next, the equations for lift and drag used in their corresponding coefficient equations are

extracted. Because of their similarity in format, only the lift equation will be derived in

detail.

L = N cos α − A sin α [5.7]

D = N sin α + A cos α [5.8]

For both the normal and axial terms, a partial derivative is taken as shown using the chain

rule for the lift equation. The derivative contains two parts: one including a delta force

and the second containing a delta angle of attack.

L = N cos α − A sin α [5.9]

∆N = ( ∆N wing cos α − N wing sin α∆α ) [5.10]

∆A = ( ∆Awing sin α − Awing cos α∆α ) [5.11]

Applying equations 5.10 and 5.11 to equation 5.9 yields equation 5.12, shown below. The

final delta L terms for each angle of attack are found from equation 5.13. The delta N and

delta A terms found in the previous step are applied for each case.

∆L = ( ∆N wing cos α − N sin α∆α ) − ( ∆Awing sin α − Awing cos α∆α ) [5.12]

( ∆N cos α ) + ( N wing sin α∆α ) + ( ∆Awing sin α ) + ( Awing cos α∆α )


2 2 2 2
∆L = wing [5.13]

62
The final values needed to complete the error analysis of the wind tunnel data are found

using the equations below. These final equations take into account the effect of the strut

used to mount the model to the sting balance. The values from these equations are applied

to the above equations for delta L and delta D (not shown).

N wing = {N wing + strut @ 30 m / s − N wing + strut @ 0m / s }− {N strut @ 30 m / s − N strut @ 0m / s } [5.14]

Awing = {Awing + strut @ 30 m / s − Awing + strut @ 0 m / s }− {Astrut @ 30 m / s − Astrut @ 0 m / s } [5.15]

A summary of the error percentage for the wind tunnel data found using the most

probable error method is shown in Table 5.1. The error analysis showed error values of

less than 10% for both lift and drag coefficients for low to moderate angles of attack, and

higher error values for higher angles of attack. The error percentages for lift and drag

coefficients at each angle of attack are plotted in the comparison section of this chapter.

Table 5.1: Wind Tunnel Error Summary

63
5.1.2 CFD Error Analysis

The previous investigations involving grid independence and the effects of flow velocity

on the data solutions are combined to provide an overall error estimate of the CFD data

obtained from FLUENT. The total error for each coefficient was found by adding the

percent differences between coefficient values from each study. A very small error was

found for the lift and moment coefficient data, while the drag coefficient values yielded a

higher error percentage, though still reasonable. This may be due to the relatively simple

Spalart-Allmaras turbulence model used in FLUENT. Table 5.2 shows a summary of the

CFD error estimates.

Table 5.2: CFD Error Summary

5.2 Data Comparison

The lift, drag, and moment coefficients for wind tunnel and CFD tests are compared by

plotting both data sets on the same plot for each coefficient. Figure 5.1 shows the lift

curves for the wind tunnel and CFD data. The error bars are not visible on the CFD curve

because of the low error percentage associated with the CFD lift coefficient data. The

wind tunnel data indicates a wider range of error than CFD. The two lift curves compare

quite well on most key points. For example, both wind tunnel and CFD data suggest

nearly the same lift coefficient value for straight and level flight. In addition, the lift

curve slopes of both curves appear to be quite close. Finally, both data sets indicate the

64
onset of stall at approximately the same angle of attack. Based on this comparison, it was

concluded that the following data would be used to populate the flight simulator

database: For -6 to -2 degrees angles of attack, wind tunnel data will be used; for 0 to 12

degrees, CFD data will be used; and finally, wind tunnel data will be used for stall and

post-stall conditions from 13 to 20 degrees. CFD data is used for the moderate angles of

attack because it follows a more linear trend, while wind tunnel is used at higher angles

of attack because of the known decrease in accuracy of the CFD model for massively

separated flows.

1.4

1.2

1.0

0.8

0.6
CL

0.4

0.2

0.0
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
-0.2 α
Wind Tunnel
-0.4 CFD

-0.6

Figure 5.1: Wind Tunnel and CFD Lift Curve Comparison

The drag curves for both wind tunnel and CFD testing are shown in Figure 5.2. The error

bars for both wind tunnel and CFD data appear to overlap for each data point, suggesting

65
a close agreement between wind tunnel and CFD results. Both data sets indicate a nearly

identical drag coefficient value at zero degrees angle of attack and the data continues to

match quite well until stall and post-stall angles of attack. Once again, the CFD may not

be completely capturing the effects for the massively separated flow at high angles of

attack because a simple one-equation turbulence model was used. However, CFD

provides excellent drag coefficient data at moderate angles of attack. For this reason, the

same data ranges that were used to populate the lift coefficient simulator tables are used

for the drag coefficient simulator tables.

0.60
Wind Tunnel
CFD
0.50

0.40

0.30
CD

0.20

0.10

0.00
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20

-0.10
α

Figure 5.2: Wind Tunnel and CFD Drag Curve Comparison

66
The moment coefficients for both wind tunnel and CFD testing are shown in Figure 5.3.

Both data sets appear to agree quite well for most angles of attack, with most of the CFD

data falling within the error range found for the wind tunnel data. The CFD data is quite

linear at moderate angles of attack while the wind tunnel data is somewhat linear but

greater than the CFD data at these angles of attack. As with the lift and drag data, both

data sets indicate an almost identical value for zero angle of attack conditions. The

moment coefficient data used for the flight simulator table is the same as the lift and drag

coefficient cases.

0.2

0.1 Wind Tunnel


CFD
0.0
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
-0.1
α
-0.2
Cm,cg

-0.3

-0.4

-0.5

-0.6

-0.7

-0.8

Figure 5.3: Wind Tunnel and CFD Moment Coefficient Comparison

67
A summary of the data presented to the Cal Poly Motion Simulator Group for

implementation into the RV-7 simulator database is shown in Table 5.3. Any

interpolation schemes necessary for compliance with flight simulator database schemes

will be implemented by members of the flight simulator team at Cal Poly.

Table 5.3: Flight Simulator Data Summary

68
Chapter 6: Conclusions and Future Work

Current industry practices of utilizing both wind tunnel and CFD technology to obtain

accurate data for implementation into a flight simulator database was implemented on a

homebuilt RV-7 aircraft. A process was developed for using aircraft construction

drawings provided by the aircraft kit manufacturer to create an accurate solid model of

the aircraft using SolidWorks. Rapid prototype technology was used to construct a 1/15th

scale wind tunnel model of the aircraft from the SolidWorks model. Wind tunnel testing

was conducted at the Cal Poly three by four foot low-speed wind tunnel to obtain lift,

drag, and moment coefficients. The SolidWorks model was imported into a CFD

preprocessor, GAMBIT 2.2.30, to create a CFD grid of the aircraft and surrounding flow

volume. The CFD tool FLUENT 6.2 was used to predict lift, drag, and moment

coefficients of the RV-7. Error analyses were done on both the wind tunnel and CFD data

sets. The wind tunnel and CFD data were compared and found to be reasonably similar

for the majority of angles of attack. The resulting compilation of both wind tunnel and

CFD data into a single data set represents the foundation for completely populating the

RV-7 flight simulator database.

6.1 Future Work

The work presented in this thesis provides a method for using industry practices to

populate a flight simulator database with accurate lift, drag, and moment coefficients.

Both the wind tunnel and CFD models used for data collection in this thesis will be made

available to the Aerospace Engineering Department at Cal Poly, San Luis Obispo to

69
ensure that students have the available resources in the future to continue an increasingly

detailed investigation of the RV-7. Neither the propeller nor the landing gear effects were

taken into account in this thesis, thus providing an avenue for future investigations. In

addition, because the RV-7 is a homebuilt aircraft with many available options, an

extensive list of these options could be generated to determine the effect of items such as

different powerplants on the aerodynamics of the aircraft. None of the control surfaces

were modeled on the aircraft used for analysis in this thesis. Therefore, the dynamic

stability characteristics of the RV-7 could be investigated.

In addition to the various work pertaining to the aircraft itself, an extensive study could

be conducted on various CFD options available. For example, future studies could utilize

the programs used in this thesis, or additional CFD programs may be found to be more

appropriate or easier to use. The aerodynamic model used for CFD may also change to

include the propeller and landing gear.

Finally, the methods employed for data analysis and comparison between wind tunnel

and CFD could be investigated. Investigation of a more systematic approach for merging

the data sets would be useful. Also, alternate error methods as well as different wind

tunnel techniques could be explored.

The above list provides insight into the amount of work that remains to be done on the

RV-7. This thesis is meant to provide a reference for future work on not only the RV-7,

but homebuilt aircraft in general.

70
Bibliography

1.) Flight Factory Simulations Homepage. October 15, 2005.


<http://www.flightfactory-simulations.com/>.

2.) Van’s Aircraft RV-7 Section. October 15, 2005


<http://www.vansaircraft.com/public/rv-7int.htm>.

3.) Seeley, B., et al. “Aircraft Performance Report.” CAFE Foundation. 1993.

4.) Stephens, A. T. Stability and Controls Engineer, The Boeing Company. Personal
Conversation. April 20, 2006.

5.) Tinoco, E., et al. “Progress Toward CFD for Full Flight Envelope.” The
Aeronautical Journal Paper 2962.

6.) Johnson, F.T., Tinoco, E.T., Nu, N. J. “Thirty Years of Development and
Application of CFD at Boeing Commercial Airplanes, Seattle.” AIAA 2003-3439
Presented at the 21st Annual Applied Aerodynamics Conference. June 2003.

7.) Saephan, S., van Dam, C.P., Fremaux, C.M. “Forces and Moments on Generic
Aircraft Forebodies at High Angles of Attack Rotary Conditions.” AIAA 2003-5472.

8). Turley, Dale R. Longitudinal Stability and Control Derivatives. PP 27-38. California
Polytechnic State University Masters Thesis. 1981.

9.) Kubik, S. Cal Poly Flight Simulation Team Member. Personal Conversation April
29, 2006.

10.) FLUENT 6.2 User’s Guide. Fluent, Inc. 2005.

11.) GAMBIT 2.2. User’s Guide. Fluent, Inc. 2003.

12.) GAMBIT 2.2 Tutorial Guide. Fluent, Inc. 2003.

13.) GAMBIT Modeling Guide. Fluent, Inc. 2003.

14.) CFD-Online FLUENT Forums. March-April, 2006. < http://www.cfd-


online.com/Forum/fluent.cgi>.

71
Appendix A: Construction Drawings Used for Solid Model
Note: Drawings not to scale

Figure A.1: Firewall Bulkhead

Figure A.2: Canopy Frame

72
Figure A.3: F-706 Aft Fuselage Bulkhead

Figure A.4: F-708 Aft Fuselage Bulkhead

73
Figure A.5: F-712 Aft Fuselage Bulkhead

Figure A.6: Side View for Guide Curves

74
Figure A.7: Top View for Guide Curves

Figure A.8: Front View

75
Figure A.9: Front Fuselage for Locating Bulkheads

Figure A.10: Aft Fuselage for Locating Bulkheads and Measuring Empennage

76
Appendix B: Wind Tunnel Data Tables

Table B.1: General Test Information

Table B.2: Calculated Test Conditions

Table B.3: Calibration Data

77
Table B.4: Raw Data Model and Strut

Table B.5: Raw Data Strut Only

78
Table B.6: Final Aircraft Values

Table B.7: Final Calculated Coefficients

79
Appendix C: Wind Tunnel Sample Calculations

Sample calculations are shown below for obtaining final Cm,cg values from the data

recorded from the sting balance. Because the calculations for moment coefficient are

similar to the calculations for CL and CD, only the process for obtaining the final Cm,cg

values will be shown. Note that the sample calculations shown are for the α = 0◦ case.

To begin, the moment of the aircraft about the sting moment center is found by

converting the raw data from volts to ft-lbf. and subtracting the effects of the strut, as

shown by equation A.1. M mc ,a is the moment of the aircraft about the sting moment

center, aM is the conversion factor for going from voltage to ft-lbf, Emc 30 m / s is the

recorded pitching force at 30m/s in volts, and Emc 0 m / s is the recorded pitching force at

0m/s in volts.

{
M mc ,a = aM ( Emc 30 m / s − Emc 0 m / s )mod el + strut − ( Emc 30 m / s − Emc 0 m / s ) strut only } [A.1]

 ft − lbf 
M mc,a = 1.98   {(−0.186V − -0.3478V ) − (−0.234V − -0.2314V )} [A.2]
 V 

M mc,a = 0.326 ft − lbf [A.3]

The moment about the aircraft center of gravity is found by equation A.4, where N a is

the aircraft normal force, l N is the moment arm of the normal force, Aa is the aircraft

axial force, and l A is the moment arm of the axial force.

80
M cg = M mc, a − N a l N − Aa l A [A.4]

M cg = 0.326 ft − lbf − 0.925lbf (0.4166 ft ) − 0.1867lbf (0.23596 ft ) [A.5]

M cg = -0.104 ft − lbf [A.6]

Finally, the moment about the aircraft cg is written as the moment coefficient:

M cg
Cm = [A.7]
cg
q∞ Sc

Where q∞ is the dynamic pressure, S is the wing planform area, and c is the mean

aerodynamic chord.

−0.104 ft − lbf
Cm,cg = [A.8]
(11.335 ft / s)(0.537 ft 2 )(0.322 ft )

Cm,cg = −0.0531 [A.9]

81

You might also like