The Measurement of the Pressure Distribution Over the Wing of an Aircraft in Flight

Home , Pressure coefficient

The Measurement of the Pressure Distribution over the Wing of an Aircraft in Flight

Matthew McCarty

A thesis submitted in fulfilment of the requirements for the Degree of

Masters of Aerospace Engineering

School of Aerospace, Civil and Mechanical Engineering

University of New South Wales

Australian Defence Force Academy

August 2008 Statements

Originality Statement ‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institute, except where due acknowledgement is made in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that the assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.’ Signed ………………………………………………… Date ………………………………………………… Copyright Statement ‘I hereby grant to The University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the Universities libraries, in all forms of media, now or hereafter known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I have either used no substantial portions of copyright material in my thesis, or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.’ Signed ………………………………………………… Date ………………………………………………… Authenticity Statement ‘I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are minor variations in formatting, they are the result of conversion to digital format.’ Signed ………………………………………………… Date …………………………………………………

ii Abstract

A measurement system has been developed for use on a light aircraft to measure the pressure distribution over the wing surfaces. The measurement system was developed as a low-cost alternative to existing advanced measurement systems.

The system consisted of low profile, low cost pressure sensors that interfaced digitally with microcontrollers for data acquisition. The pressure sensors and microcontrollers were developed into self-contained sensor modules with all electronic components mounted on flexible circuit board that formed the base of the modules. Two types of module were developed; a module with a single pressure sensor and a module with a row of seven pressure sensors at fifteen millimetre spacing. The total cost of the sensor modules was approximately ninety dollars for a single sensor module and one hundred and forty dollars for the seven sensor module.

Studies were carried out using numerical methods to predict the pressure distribution over a NACA2412 airfoil. The numerical studies were used to evaluate the effect of adding the sensor modules to the wing, and the effect of the sensor distribution on measured force coefficients. Numerical predictions were made using the XFOIL software package. This software was validated using the Hess-Smith inviscid panel method.

Flight testing was carried out with the pressure distribution measurement system to confirm the operation of the system and to make preliminary measurements. The flight testing focused on the measurement of steady state pressure distributions for comparison with the numerical predictions. Good agreement was found between the measured pressure distributions and the XFOIL predictions. Integration of the pressure distributions enabled comparison of normal force, lift force and quarter chord moment coefficients. The measured force coefficients showed the expected trends with angle of attack although it was found that the limited number of sensor modules used caused large error in the quarter chord moment coefficient compared to the numerical predictions.

iii Acknowledgments

First and foremost I would like to thank my supervisor Dr Michael Harrap. The chance to undertake postgraduate research that involved flying was a rare one and I am extremely grateful for the opportunity. I would like to thank those from the ‘aviation research group’ at UNSW@ADFA; Sue Burdekin, Raymond Lewis, and Martin Copeland for their support. For the tremendous help and time devoted to teaching a mechanical engineer the intricacies of electrical and electronic engineering I would like to thank Andrew Roberts, Evan Hawke, Tony Peebles, and Geno Ewyk. Thanks needs to go to the other academic staff from the University, whose expertise in various areas was invaluable to bounce ideas off, especially John Milthorpe, Alan Fien, John Young, Krishna Shankar, and Andrew Neely. A big thanks needs to go to the workshop staff that helped with the project, especially Marcus DeAlmeida, Bob Bleakley, Dave Sharp, Jim Baxter and Wayne Jealous. I would like to acknowledge the support of the office staff at the School of ACME; Carol Obrien, Cheri Khalil, Gill Taylor, and Vira Berra. I would like to thank the postgrads from ACME; Chris Hang, Orio Kieboom, Laxmikant Kannapan, Vishwas Puttige, Arif Ashraff, Kartik ‘Ram’ Ramakrishnan, Steve Choi, Abhi Kallapur, Kamal Singh, and Sebastian Oberst. Thanks to my housemates; Ra, Beth, Dave and Sam. Finally thanks must go to my family, Mum, Dad, Jennifer and Thomas whose support from abroad for my endeavours ‘across the ditch’ was invaluable.

iv Table of Contents

Statements...... i Statements...... ii Abstract...... iii Acknowledgments ...... iv Table of Contents ...... v List of Figures...... vii List of Tables ...... xii List of Tables ...... xii 1 Introduction...... 1 1.1 Why measure pressure distribution? ...... 1 1.2 Air Pressure Distribution ...... 1 1.3 Measuring the pressure distribution in flight ...... 6 1.4 Objective ...... 6 1.5 Flight Labs...... 6 1.6 Project Origin...... 6 1.7 Thesis structure...... 7 2 A Review of In-Flight Pressure Measurement Techniques...... 9 2.1 Why measure aircraft pressure distributions?...... 9 2.2 Methods for measuring pressure distributions ...... 11 2.2.1 Tubed Pressure Belts ...... 11 2.2.2 Flush surface orifices...... 16 2.2.3 Pressure sensitive paint...... 17 2.2.4 Modern advancements ...... 19 2.3 Aerodynamic properties measured using pressure belts ...... 23 2.4 Comparisons between pressure distribution measurement methods...... 26 2.4.1 Comparison with wind tunnel testing...... 26 2.4.2 Comparison between flush orifice and pressure belt methods ...... 28 2.5 Conclusions...... 30 3 Development of the pressure measurement system ...... 31 3.1 Description of the system...... 31 3.2 Pressure Sensor...... 33 3.2.1 SCP1000 pressure sensor...... 35 3.3 Sensor modules ...... 37 3.4 Microcontrollers...... 38 3.4.1 Slave microcontroller...... 38 3.4.2 Master microcontroller ...... 39 3.5 Network topology...... 40 3.6 Circuit Board...... 41 3.7 Cost breakdown...... 42 3.8 Computer software...... 42 3.9 Test aircraft description ...... 44

v 3.9.1 Air data instrumentation ...... 46 3.10 Sensor Characterisation...... 46 3.10.1 Sensitivity check ...... 47 3.10.2 Temperature Stability...... 51 3.10.3 Frequency response...... 52 3.11 Error Analysis ...... 57 3.12 Certification of the pressure belt...... 58 3.13 Conclusions...... 59 4 Numerical Investigations...... 61 4.1 Background to numerical flow calculation methods...... 61 4.1.1 Hess-Smith panel method...... 63 4.1.2 XFOIL software ...... 65 4.1.3 Calculation of force and moment coefficients from pressure distributions...... 66 4.2 Cessna 182 airfoil measurement and analysis...... 67 4.3 XFOIL results...... 71 4.3.1 Effects of viscosity...... 71 4.3.2 Effect of estimating trailing edge pressure coefficient ...... 73 4.3.3 Effect of sensor distribution...... 75 4.3.4 Effect of pressure belt thickness...... 80 4.4 Lifting line theory ...... 81 4.5 Conclusions...... 84 5 Flight Investigations...... 86 5.1 Description of the test flights ...... 86 5.2 Initial attachment of sensor modules...... 88 5.2.1 Results...... 88 5.2.2 Discussion ...... 89 5.3 Steady state results...... 93 5.3.1 Pressure distributions...... 93 5.3.2 Steady state force coefficients ...... 96 5.4 Discussion of steady state force and moment coefficients...... 99 5.4.1 Flow separation...... 100 5.5 Short period pitch oscillations...... 106 5.5.1 Results...... 106 5.5.2 Discussion ...... 108 5.6 Conclusions...... 112 6 Conclusions and Recommendations...... 114 6.1 Introduction...... 114 6.2 Conclusions...... 114 6.3 Recommendations for future work ...... 117 References ...... 120 Appendix A. Hess-Smith panel method equations ...... A-1 Appendix B. CAR 35 approval...... B-1 Appendix C. Electrical circuit schematic ...... C-1 Appendix D. Software functions...... D-2

vi List of Figures

Figure 1-1: Forces on an aircraft in flight. Reproduced from [1]...... 2 Figure 1-2: Total forces acting on an airfoil section can be broken into pressure forces and shear forces. Reproduced from [2]...... 2 Figure 1-3: The resultant force acting on an airfoil (R) can be broken down into lift (L) and drag (D) components relative to the freestream air velocity, or normal (N) and axial (A) components relative to the airfoil chordline. Reproduced from [3]...... 3 Figure 1-4: Typical distribution of pressure over an airfoil. Red lines represent pressures lower than atmospheric while blue lines represent pressures higher than atmospheric...... 3 Figure 1-5: A pressure distribution curve for a NACA 2412 airfoil at 5° angle of attack...... 4 Figure 1-6: Variation in pressure distribution with increasing angle of attack...... 5 Figure 2-1: A typical pressure distribution for a NACA2412 airfoil. The surface air pressures are displayed in terms of non-dimensionalised pressure coefficients...... 9 Figure 2-2: Cross section of a typical tubed pressure belt. Reproduced from [14]...... 11 Figure 2-3: Figure (a) shows a comparison between pressure distribution measured by 3 tube sizes in flight at similar angles of attack. Figure (b) shows a comparison between a pressure distribution measured in-flight and a pressure distribution predicted using two-dimensional computational methods for the same angle of attack. Reproduced from [9]...... 13 Figure 2-4: Cross-section of the pressure belt used in [11]...... 15 Figure 2-5: Level flight pressure distributions on a PW-9 pursuit aircraft. Reproduced from [17]...... 17 Figure 2-6: Camera setup used to monitor a strip of pressure sensitive paint in-flight. The camera is set up to observe the paint on the top surface of the wing through an over-wing window. Reproduced from [18]...... 19 Figure 2-7: Schematic of the acquisition system architecture developed in [20]...... 20 Figure 2-8: Pictorial diagram of a Boeing aircraft wing with the pressure belt installed showing a breakdown of the components in the pressure belt...... 21 Figure 2-9: Schematics of the pressure belt segments developed by Boeing. The top schematic is a plan view of a single belt segment while the bottom schematic is a cross-section of a segment. The scales of the two schematics are different. Reproduced from [6]...... 22 Figure 2-10: Direct comparison between pressure measured using a tubed pressure belt and the advanced MEMs sensor module system. Reproduced from [22]...... 23 Figure 2-11: Pressure distribution measured on the upper surface of the F-16XL, showing shock wave locations. Reproduced from [11]...... 24 Figure 2-12: Section normal force coefficient (Cn) as a function of angle of attack. Reproduced from [10]...... 24 Figure 2-13: Section lift coefficients calculated using the measured pressure distributions from the HARV program. Reproduced from [10]...... 25

vii Figure 2-14: An example of the flow separation patterns recorded during the HARV research program. The flow separation patterns were obtained using wool tufting and confirmed with measurements from the pressure belts. Reproduced from [10]...... 26 Figure 2-15: Comparison between pressure coefficients measured in flight with wind tunnel data for an F- 100 aircraft. The unlabelled x-axis represents the normalised chord-length of the aircraft wing. Reproduced from [23]...... 27 Figure 2-16: Comparison between lift coefficients measured on the F-100 aircraft in-flight and from wind tunnel testing. Reproduced from [23]...... 28 Figure 2-17: Comparison between pressure coefficients measured using flush orifices and tubed pressure belts. The top graph is a high Mach number condition (M=0.9) and the bottom graph is a low Mach number condition (M=0.5). Reproduced from [14]...... 29 Figure 3-1: Diagram showing the pressure belt system configuration ...... 33 Figure 3-2: VTI Technologies SPC1000 series absolute pressure transducer. Reproduced from [26]...... 36 Figure 3-3: Single sensor module (left) and high resolution sensor module (right)...... 37 Figure 3-4: Schematic of the relationship between the master and slave microcontrollers (right), compared with a standard bus network topology (left). Left picture reproduced from [27]...... 39 Figure 3-5: Circuit layout of the single sensor circuit (left) and the high density circuit (right)...... 41 Figure 3-6: Screenshots showing key tab panels from interface software. Top left shows data from the aircraft instrumentation while top right shows traces of the pressure for each pressure sensor. Lower left shows the pressure coefficient plot in-flight while lower left shows the in-flight lift coefficient display. 43 Figure 3-7: Cessna 182RG three view drawing. Reproduced from [29]...... 44 Figure 3-8: Air data instrumentation consisted of two single sensor modules housed in sealed aluminium die-cast boxes with the pressure inputs from the air data boom connected...... 46 Figure 3-9: Sensor calibration at room temperature...... 48 Figure 3-10: Comparison between the pressures measured by the SCP1000 sensor module (black) and a Kestrel 4000 portable weather station (red)...... 49 Figure 3-11: Linearity comparison between the SCP1000 sensor module and the Kestrel 4000...... 50 Figure 3-12: Comparison between the Canberra airport weather station pressure (black) and the SCP1000 sensor module pressure (red)...... 51 Figure 3-13 ...... 52 Figure 3-14: Raw data gathered for establishing a noise base...... 54 Figure 3-15: Detrended pressure readings used to establish a noise base...... 54 Figure 3-16: Noise base spectrum for a typical sensor module...... 55 Figure 3-17: Power spectral density showing the sensor response when subjected to a 1Hz triangular waveform...... 56 Figure 3-18: Power spectral density showing the sensor response when subjected to a 2Hz triangular waveform...... 56 Figure 3-19: Variance plot indicating the precision of the SCP1000 sensor...... 58 Figure 3-20: Sensor module installation specification. Reproduced from [34]...... 59

viii Figure 4-1: Superposition of elementary flows. The stream functions of the elements are given in polar coordinates. Reproduced from [3]...... 62 Figure 4-2: Panel codes derive their name from the division of an airfoil into numerous short, straight line segments...... 62 Figure 4-3: Correct circulation over an airfoil corresponds with the flow smoothly leaving the trailing edge; the Kutta condition (right). The circulation of the left is an arbitrary circulation resulting in physically incorrect streamlines. Reproduced from [2]...... 63 Figure 4-4: A cosine distribution is commonly used to divide an airfoil into discrete panels. Reproduced from [31]...... 63 Figure 4-5: Drag coefficient convergence of the Hess-Smith panel method...... 65 Figure 4-6: The average pressure between two nodes was used when computing the force coefficients. . 67 Figure 4-7: Comparison between the measured Cessna 182RG airfoil and the standard NACA2412 section...... 68 Figure 4-8: Comparison between the measured NACA2412 modified airfoil and the standard NACA2412 airfoil, showing the large difference at the leading edge...... 69 Figure 4-9: A comparison between the measured NACA 2412 airfoil and NACA 2412 airfoils leading edge radius modifications...... 70 Figure 4-10: Comparison between pressure distributions for a standard NACA 2412 airfoil and NACA 2412 airfoils with modified nose radius...... 71 Figure 4-11: Comparison between inviscid and viscous lift coefficient variation with angle of attack..... 72 Figure 4-12: Pressure distributions predicted using XFOIL showing the effect of viscosity...... 72 Figure 4-13: Comparison between experimental lift coefficient values and inviscid values predicted by XFOIL for a NACA 2412 airfoil section. The experimental data is from Abbot and Von Doenhoff [39].73 Figure 4-14: Comparison between trailing edge pressure coefficient estimate methods at a single angle of attack...... 74 Figure 4-15: A comparison between the three sensor distribution methods that were evaluated...... 75 Figure 4-16: Pressure distributions for the three sensor distribution methods evaluated at 5° angle of attack...... 78 Figure 4-17: Comparison between the calculated lift coefficient using the three sensor distribution methods...... 79 Figure 4-18: The main effect of the addition of a pressure belt is a reduction in the peak leading edge suction pressure...... 81 Figure 4-19: The spanwise loading of a finite aircraft wing. The case shown is for a wing with elliptical loading that experiences constant downwash Reproduced from [2]...... 82 Figure 4-20: The effect of downwash from a finite wing is to reduce the effective angle of attack. Reproduced from [32]...... 82 Figure 4-21: Wing plan form (left) and wing twist (right) used in lifting line calculations...... 83 Figure 4-22: Spanwise distribution of local section angle of attack for aircraft angles of attack of 0°, 5°, 10° and 15°. The aircraft angle of attack is taken relative to the fuselage datum line...... 83

ix Figure 4-23: Relationship between the aircraft angle of attack and the section angle of attack for the section where the pressure belt is attached...... 84 Figure 5-1: Map showing flight test area...... 88 Figure 5-2: Diagram showing the initial attachment and fairing to the wing surface using aluminium tape...... 88 Figure 5-3: A typical pressure distribution measured during the first test flight, showing the effect that the tape fairing was having on the results...... 89 Figure 5-4: A wedge shape was used to simulate the flow over a faired sensor to determine the error created by the fairing...... 90 Figure 5-5: The effect of individually faired sensors on the pressure distribution for different fairing ratios at 6° angle of attack...... 91 Figure 5-6: Diagram showing the use of close cell foam to fill between the pressure sensors and create a constant thickness...... 91 Figure 5-7: The initial method of fairing the sensor modules (left) was unsatisfactory. The final attachment of the sensor modules to the wing (right) included foam that was used to provide a constant thickness fairing in between the sensor modules. In both cases the leading edge of the wing is to the bottom of the picture...... 92 Figure 5-8: Comparison between pressure distributions measured with and without the inter-sensor foam fairing at a single angle of attack. Also shown is the predicted pressure distribution for the particular angle of attack...... 93 Figure 5-9: Pressure distribution measured during steady state flight at an angle of attack of 2.4°. The measured distribution has been compared with a distribution predicted using XFOIL for the same section angle of attack...... 94 Figure 5-10: Pressure distribution measured during steady state flight at an angle of attack of 5°. The measured distribution has been compared with a distribution predicted using XFOIL for the same section angle of attack...... 95 Figure 5-11: Pressure distribution measured during steady state flight at an angle of attack of 12°. The measured distribution has been compared with a distribution predicted using XFOIL for the same section angle of attack...... 96 Figure 5-12: Time history showing the section angle of attack (top) and the calculated section lift coefficient (bottom) during a slow, quasi-steady increase in angle of attack up to stall...... 97 Figure 5-13: Comparison between the calculated section normal force coefficient and the predicted normal force coefficient using XFOIL...... 98 Figure 5-14: Comparison between the lift force coefficients calculated from flight data and numerical predictions using XFOIL...... 99 Figure 5-15: Time history of section angle of attack (top) and section lift coefficient showing flow separation occurring and the aircraft stall point...... 101 Figure 5-16: Flow separation progression as seen in the measured pressure distributions. Top left shows the flow to be attached prior to stall, while top right shows intermittent flow separation effects prior to

x complete flow separation. Bottom left shows the flow to be separated with bottom right showing the flow reattachment after recovery from the wing stall...... 103 Figure 5-17: Comparison between the lift force coefficients calculated from flight data and lift force coefficients measured for a two-dimensional NACA2412 airfoil in a wind tunnel...... 104 Figure 5-18: Lift force coefficient variation with angle of attack for different airfoils, showing the effect of different stall modes on the angle at which stall occurs. Reproduced from [40]...... 105 Figure 5-19: Time history of the section angle of attack showing a sequence of short period pitch manoeuvres...... 107 Figure 5-20: Time history of a single short period oscillation showing the range of section angle of attack and the highly damped nature of the motion...... 107 Figure 5-21: Lift force coefficient variation during short period pitch manoeuvre, showing the hysteresis loop...... 108 Figure 5-22: The rotational motion of the aircraft during the short period pitch manoeuvre caused the angle of attack vane to misread...... 109 Figure 5-23: Lift force coefficient variation during short period pitch manoeuvre...... 110 Figure 5-24: Unsteady lift force coefficient variation for a NACA0012 airfoil during pitching oscillation about the axis at x/c =0.25, with M∞=0.755 and α0=2.5°. Reproduced from [41]...... 111 Figure 5-25: Measured pressure distributions showing the effect of nose up pitching (red) and nose down pitching (blue) compared to steady state flight (black)...... 112

xi List of Tables

Table 1: Comparison between pressure sensor characteristics ...... 34 Table 2: Characteristics of the SCP1000 pressure sensor ...... 36 Table 3: Characteristic of the Microchip Microcontrollers...... 38 Table 4: Cost breakdown for sensor modules...... 42 Table 5: Existing aircraft sensors...... 45 Table 6: Error from room temperature sensor calibration on sensor 5 ...... 48 Table 7: Error from week long comparison with Kestrel 4000 portable weather station ...... 49 Table 8: Errors in sensor response at varying temperatures...... 52 Table 9: Summary of expected error due to sensor characteristics...... 58

Table 10: Errors in force and moment coefficients due to estimating the trailing edge CP and sensor distribution...... 77 Table 11: Description of the test flights...... 87

xii 1 Introduction

“To design a flying machine is nothing, to build it is not much, but to test it is everything” Otto Lilienthal (1848-1896) This thesis concerns the problem of measuring the distribution of pressure over the wing of an aircraft during flight. In this chapter the thesis project will be introduced and background information will be presented.

1.1 Why measure pressure distribution? The measurement of the pressure distribution across the chord of an aircraft wing in flight is often carried out during flight testing on new aircraft designs and during flight research programs. The importance of the measurement lies with the ability to verify numerical predictions of the pressure distribution and/or any scale wind tunnel testing that has been carried out. The measurement of the air pressure distribution also enables flight loads to be calculated and this provides a verification that the aircraft structure is operating within the design loading conditions.

1.2 Air Pressure Distribution The four main forces acting on an aircraft are lift, drag, thrust and weight. This work is focused on the lift force that is generated by distributions in air pressure. During steady-state flight, the weight force of an aircraft will be balanced by the lift generated, while for a powered aircraft the drag of the aircraft will be balanced by the thrust produced. The relationship between the forces is shown in Figure 1-1.

Figure 1-1: Forces on an aircraft in flight. Reproduced from [1].

The forces acting on an aircraft due to air can be divided into pressure forces and shear forces (Figure 1-2). Shear forces are due to the viscosity of the air while pressure forces are created by changes in air velocity as it passes over the aircraft.

Figure 1-2: Total forces acting on an airfoil section can be broken into pressure forces and shear forces. Reproduced from [2].

The distribution of pressure over an airfoil creates a net resultant force that is typically broken up into components; lift and drag forces relative to the free stream direction or normal and axial forces relative to the airfoil as shown in Figure 1-3. The resultant force is usually expressed as a lift force, drag force and a moment acting about a point located at a quarter of the total chord length from the leading edge of the airfoil.

Figure 1-3: The resultant force acting on an airfoil (R) can be broken down into lift (L) and drag (D) components relative to the freestream air velocity, or normal (N) and axial (A) components relative to the airfoil chordline. Reproduced from [3].

A typical pressure distribution over an airfoil is shown in Figure 1-4. The pressures acting on an airfoil at any point are non-dimensionalised into pressure coefficients according to equation 1.

Figure 1-4: Typical distribution of pressure over an airfoil. Red lines represent pressures lower than atmospheric while blue lines represent pressures higher than atmospheric.

p − p∞ CP = 1 ρ V 2 (1) 2 ∞ ∞

where p is the pressure at any point on the surface,

p∞ is the freestream static pressure

V∞ is the freestream fluid velocity

ρ∞ is the freestream fluid density

3 The distribution of pressure is characterised by pressures much lower than atmospheric pressure over the upper leading edge of the wing. This is associated with the area of largest surface curvature and as the air is accelerated around the leading edge the static pressure is lowered. As the air travels towards the trailing edge over the upper surface the pressure recovers toward atmospheric pressure. The airflow over the lower surface is characterised by a stagnation point on the lower leading edge surface, where the kinetic energy of the air is converted completely to static pressure. This point is the point of highest absolute pressure. The air pressure over the lower surface is typically higher than atmospheric pressure, with the pressure reducing towards the trailing edge. A typical non-dimensionalised pressure distribution is shown in Figure 1-5.

Figure 1-5: A pressure distribution curve for a NACA 2412 airfoil at 5° angle of attack.

Negative pressure coefficients represent local air pressures lower than atmospheric pressure, with positive pressure coefficients representing local air pressures higher than atmospheric pressure. The maximum pressure coefficient attainable is a coefficient of one, representing the stagnation pressure. A pressure coefficient of zero represents atmospheric pressure. The y-axis of a pressure coefficient graph is reversed so that the upper line in the graph shows the pressure coefficients over the upper airfoil surface.

The distribution of pressure coefficients over an airfoil varies with the angle of attack of the airfoil, and to a lesser extent the Reynolds number of the airflow. Pressure

4 distributions can be estimated using a number of well developed methods including empirical methodology, wind tunnel testing and numerical computational methods. A typical variation of pressure distribution with angle of attack is shown in Figure 1-6 for three angles of attack. The peak pressure coefficients over the upper leading edge surface increase with increasing angle of attack while the stagnation point on the lower leading edge surface tends to move rearward with increasing angle of attack.

Figure 1-6: Variation in pressure distribution with increasing angle of attack.

The area enclosed by a pressure coefficient distribution represents the resultant force coefficient acting on the airfoil. The normal and axial force coefficients and moment coefficients about any chord position can be calculated by integrating the pressure distribution. The lift force coefficient can also be calculated provided the angle of attack of the airfoil is known.

When designing an aircraft, it is important to have good estimates of the pressure distribution over the lifting surfaces so that the aircraft behaviour can be predicted and the airframe structure can be designed to support the flight loads. Once the aircraft has been designed and built it is important to verify the actual pressure distribution over the aircraft wing during flight to confirm that the estimated distributions used to design the aircraft were accurate. Measuring the pressure distribution over an airfoil surface can

5 provide this verification of the flight loads and also verifies any wind tunnel testing or numerical simulations that have been carried out.

1.3 Measuring the pressure distribution in flight Various methods have been developed for measuring the pressure over an airfoil in flight. The most common method involves wrapping small diameter tubes around the airfoil section. The tubes are used to pipe the air pressure from discrete points on the airfoil section to a pressure sensor. This method and others for measuring the pressure distribution will be discussed in detail in Chapter 2.

1.4 Objective The objective of this research project is to design and build a measurement system for measuring the pressure distribution over the wing of a light aircraft in flight. The system will be tested and used on a Cessna 182RG aircraft.

The purpose of the pressure distribution measurement system will be to add an aerodynamic research capability to the aircraft owned and operated by the School of Aerospace, Civil and Mechanical Engineering (ACME) of the Australian Defence Force Academy campus (University College) of the University of New South Wales (UNSW@ADFA).

1.5 Flight Labs The Cessna 182RG (C182RG) aircraft owned and operated by the School of ACME at UNSW@ADFA is currently used to carry out flight laboratories for an undergraduate course. The aircraft is already equipped with sensors and measurement systems for making measurements pertaining to the flight characteristics of the aircraft. The flight labs currently focus on aircraft control and stability and it is hoped that the pressure distribution measurement system will enable aerodynamic concepts to be demonstrated to the undergraduate students.

1.6 Project Origin This project originated from a final year thesis project by an undergraduate student, Christopher Clyde in 2004 [4]. The project investigated the potential for measuring the pressure distribution over the tail surface of the C182RG aircraft. The work focused on

6 developing a tubed pressure belt that would attach to the tailplane with the instrumentation for the system being housed in the tailcone of the aircraft. The project progressed as far as building a prototype of a switching manifold for cycling through a bank of tubes in order to multiplex the tubes into one pressure sensor [4].

1.7 Thesis structure The thesis will be structured as follows. The first chapter has outlined the background to the research project. The origin and objectives of the thesis have also been outlined.

The second chapter will present a review of literature relevant to the problem of measuring the pressure distribution on an airfoil section in flight. The different methods used to make the measurements will be discussed, along with comparisons between the different methods. The aerodynamic properties that can be discerned from pressure distribution methods will also be discussed.

The third chapter will outline the development of the pressure distribution measurement system. The details of the system will be described, as well as the initial bench testing that was carried out with the system.

The fourth chapter will describe the numerical methods that were used to predict the pressure distributions. The predicted pressure distributions were important as they enabled a comparison to be made with the pressure distributions gathered in flight. The numerical methods were also used to investigate several aspects of the measurement system such as the effect of the distribution of points where the pressure would be measured.

The fifth chapter describes the flight testing that was carried out and the experimental results that were gathered. The experimental results focused on making measurements during quasi steady-state flight but the results from dynamic flight manoeuvres are also presented.

The sixth chapter will present conclusions from the project and recommendations for future research work.

7 In this chapter an overview of the thesis project has been outlined and the background to the project has been presented. The next chapter will present a review and discussion of pertinent literature.

8 2 A Review of In-Flight Pressure Measurement Techniques

In this chapter a review of relevant literature will be presented. The literature review begins with a summary of the reasons for measuring the pressure distribution over the wing of an aircraft. The methods for measuring pressure distribution will then be discussed and the advantages and disadvantages of the methods will be outlined. Recently developed pressure distribution methods will then be discussed with a focus on the system developed for use by the Boeing Aircraft Company. The aerodynamic properties that can be measured by pressure belts will then be outlined followed by some comparisons that have been made between the measurement methods.

2.1 Why measure aircraft pressure distributions? The main reason for making measurements of the air pressure distribution in flight is to determine flight loads. As discussed in Chapter 1, by measuring the distributed pressure acting on a particular surface of an aircraft, the forces acting on the aircraft can be calculated. A typical pressure distribution is shown in Figure 2-1.

Figure 2-1: A typical pressure distribution for a NACA2412 airfoil. The surface air pressures are displayed in terms of non-dimensionalised pressure coefficients.

9 During the certification process for an aircraft, flight loads are measured in order to confirm the design flight loads. A flight load survey normally involves mapping the pressures acting on the aircraft surfaces during steady and manoeuvring flight throughout the aircraft’s design envelope [5]. The total loads acting on the aircraft structures can then be calculated from the measured pressures. An aircraft manufacturer will also use the data from a flight load survey to build up a database of loading conditions supported by wind tunnel results and Computational Fluid Dynamics (CFD) results [5, 6]. This database may be used to justify future modifications or design refinements to the aircraft. Third party companies specialising in modifying existing aircraft will often perform their own independent flight load surveys in order to build a database that can be used to justify their own modifications [5].

In-flight pressure distribution measurements can also be used to validate any CFD predictions. Most aircraft designers and manufacturers use CFD to some extent during the development of an aircraft design or modification, as CFD provides a cheap alternative to wind tunnel testing and can provide design data quickly. However CFD results require validation to show that the model, the meshing and the flow conditions used are consistent with reality. Measuring the pressure distributions over aircraft surfaces in flight provides data that can validate CFD results, or highlight any areas where the CFD results do not reflect reality [5]. One of the main problems with CFD is gaining accuracy in predicting the behaviour of turbulent flow. Even a relatively simple problem such as predicting the transition from laminar flow to turbulent flow across the surface of a flat plate can require millions of grid points and excessive amounts of computer execution time to make an accurate prediction [7].

Validating wind tunnel data is another function that in-flight distribution measurement can perform. As part of the development of an aircraft design, scale wind tunnel testing is normally carried out. Scale testing of an aircraft design in a wind tunnel is a well refined procedure that has been used since before human powered flight was achieved [3]. The disadvantage of scaled wind tunnel testing is that it is difficult to match scaling factors such as Reynolds number, surface roughness, and geometry [8]. This means that wind tunnel results also require validation. One way to validate wind tunnel results is by making pressure distribution measurements in flight.

10 Although wind tunnel testing and CFD are useful design tools, it is difficult to acquire useful data about flight at high angles of attack, flow separation phenomenon such as stall and spin, and transonic flight regimes. These are particular areas where in- flight measurements can be used to gather design and performance data.

One area where the measurement of in-flight pressure distributions is employed extensively is research flight testing. Organisations such as the National Aeronautics and Space Administration (NASA) conduct research flight testing to gather information about future aircraft designs and technologies and frequently investigate[9] high angle of attack flight [10] and boundary layer control [11].

Another field where pressure distribution measurements are often made is in automobile aerodynamics. However, most automobile aerodynamic testing is carried out in wind tunnels, including work completed in Australia into the pressure fluctuations on the mirror surface of a car side mirror [12]. This is one field where the development of a cheap, versatile pressure distribution measurement system would enable precise road testing to be carried out.

2.2 Methods for measuring pressure distributions Several methods exist for measuring pressure distributions over a surface. The more popular methods are tubed pressure belts and flush surface orifices.

2.2.1 Tubed Pressure Belts One of the most common methods for measuring pressure distribution across the chord of a wing in flight is the tubed pressure belt method. Pressure belts are made up of a bank of small diameter plastic tubes laid side by side, as shown in Figure 2-2. The tubes are wrapped around the wing in the chordwise direction and each tube is used to transmit the pressure from an orifice at a single chordwise location to a pressure sensor located off-wing. Each tube in the bank is used to measure the pressure at a single chordwise location. The tubes are normally multiplexed to a single pressure sensor using a mechanical pressure scanner that cycles discretely between up to 128 tube inputs [13].

Figure 2-2: Cross section of a typical tubed pressure belt. Reproduced from [14].

11 One advantage of the pressure belt system is its ease of application to the wing surface. The banks of tubes are normally attached to the wing using adhesive or tape and can be installed and removed quite easily.

The main disadvantage of a tubed pressure belt is that the pressures measured are not the true surface pressures on the wing. The addition of the tubing to the wing surface inflates the local size of the airfoil and also affects the local airflow to some extent, resulting in pressure measurements that differ from the true surface pressures.

It has been shown that by using small diameter tubes, the effect of increasing the local airfoil size is negligible [9]. One study carried out by NASA used a light aircraft similar in performance to UNSW@ADFA’s C182RG aircraft to evaluate the effect of various pressure belt tube sizes [9]. The study evaluated 3 tube diameters; 1.58mm, 3.175mm and 4.76mm. There was generally good agreement between the pressure distributions measured using each of the 3 tube sizes, as shown in Figure 2-3 (a). The most notable discrepancies were found to occur at the trailing edge region, where the larger tube sizes had a noticeable effect. The distributions measured using the thinnest tubing most closely reflected pressure distributions predicted using two-dimensional computational methods as shown in Figure 2-3 (b), highlighting the need to keep a pressure belt as thin as possible [9].

(a)

(b) Figure 2-3: Figure (a) shows a comparison between pressure distribution measured by 3 tube sizes in flight at similar angles of attack. Figure (b) shows a comparison between a pressure distribution measured in-flight and a pressure distribution predicted using two-dimensional computational methods for the same angle of attack. Reproduced from [9].

A specification for a pressure belt is often that the thickness of the tubes is less than the boundary layer thickness, as is the case with tubed pressure belts used by Boeing [6]. The property that is attempted to be exploited is that the pressure variation through

13 the thickness of a boundary layer is very small, and the pressure at the surface can be considered to be the same as at the displacement thickness of the boundary layer [3].

Theoretically, if a pressure belt was thinner than the displacement thickness, then the pressure on the surface of the belt should be roughly the same as the original surface pressure (without the pressure belt attached). This assumption may not be correct because the surface of a smooth pressure belt attached to the surface of an airfoil section will become the airfoil surface that the airflow travels over, and the corresponding boundary layer should set up from the surface of the pressure belt. This means for even the thinnest practical tube size, the pressure belt will always record pressures that are incorrect and the main error should be due to the pressure belt modifying the airfoil section geometry. This fact is not widely acknowledged, with many pressure belt systems claiming that the pressures recorded must be close to the true pressures because the pressure belt is thin relative to the boundary layer thickness. An estimate of the error caused by different pressure belt thicknesses is presented in section 4.3.4.

Furthermore, the goal of keeping a pressure belt thinner than the boundary layer thickness is impossible to achieve in areas of the airfoil around the leading edge where the boundary layer is still laminar and has not developed fully and the displacement thickness is still very small. However, on an aircraft like the Cessna 182RG (that is used by UNSW@ADFA), the boundary layer thickness grows rapidly and the region where the boundary layer is laminar is confined to at most 10% of the chord length.

Another disadvantage of pressure belts is that the tubing causes acoustic and pneumatic lag. Any pressure variations transmitted through the tubes propagate as waves, which are damped by wall friction causing a magnitude attenuation and a phase lag [15]. The phase lag created limits the frequency at which the pressure can be sampled by the pressure sensor, particularly when using a Scanivalve device to cycle between numerous tubes. In this case, after switching between tubes a time delay is needed to allow the pressure to equalise before a measurement is made. This assumes that the flight measurements are being made under quasi-steady state conditions. The phase lag and magnitude attenuation associated with a tubed pressure belt often precludes investigations of dynamic flight conditions.

14 A third disadvantage of pressure belts is the time required to connect the tubing. Although the tubes are relatively easy to install on the wing surface, plumbing the numerous tubes to a pressure sensor located off-wing can be time-consuming. The difficult part of the installation is routing the tubing to the pressure sensor, which is normally housed in an instrumentation bay located off-wing. Boeing estimate that for an aircraft undergoing a flight load survey it can take up to 8 months to design, fabricate and install an entire tubed measurement system [16].

The final disadvantage of a tubed pressure belt is that during flight, a tube or orifice can become blocked by debris such as insects or water from flight through rain. This makes the tube unusable until the aircraft has landed and the tube or orifice cleared, or it necessitates a complicated system for clearing tube blockages in-flight.

Various NASA research programs have used tubed pressure belts to measure pressure distributions. The F-16XL [11] research program used pressure belts made up of 3.3mm outer diameter plastic tubing with a maximum tubing length of 7.6 meters leading to the off-wing pressure sensor. This length of tubing created a maximum acoustic and pneumatic lag of nearly one second at the test altitude, although this was considered acceptable in this case because each test point was maintained for ten seconds while measurements were made [11]. A typical cross-section of the belt setup is shown in Figure 2-4. The fairing on the side of the belt of 8 to 10 tube diameters was important in this application because of the large amount of cross-flow present due to the sweep of the wing on which it was mounted.

Figure 2-4: Cross-section of the pressure belt used in [11].

Another NASA research program that used tubed pressure belts was the F-18 High Alpha Research Vehicle (HARV) [10]. This program focused on flight at very high angles of attack. The pressure belts attached to the test aircraft had an outer diameter of 3.175mm and had a maximum pneumatic lag of 0.1 seconds, allowing a sampling

15 frequency of 10 Hertz to be used [10]. The results from these two research programs will be discussed further in section 2.3.

2.2.2 Flush surface orifices Another method used to measure pressure distribution is the flush surface orifice method. This method has been used successfully in flight testing but is more commonly used in wind tunnel testing, due to the modifications required to the test aircraft.

Flush surface orifices are created by drilling small holes through the surface over which the pressure distribution will be measured. Air pressure is transmitted from the orifice to a pressure sensor using tubing, similar to the tubed pressure belt method except that the tubing is internal to the aircraft wing.

The advantage of the flush orifice method over a tubed pressure belt is that the air pressures measured are the true pressures over the actual wing surface. The orifices are flush to the surface of the wing and unlike the pressure belt method, there is no protrusion that affects the local airflow characteristics.

The obvious disadvantage of the method is the permanent modifications that are required to be made to the aircraft. The orifices require the wing skin of the aircraft to be drilled and the holes can compromise the structural integrity of the airframe. This can create problems for aircraft of monocoque (stressed skin) or semi-monocoque construction, requiring structural reinforcement to be installed. However, because most aircraft used for flight testing programs are solely dedicated to that purpose, installation of permanent test instrumentation can be tolerated.

Another disadvantage of the flush surface orifice method is that the surface orifices still need to be connected to a pressure sensor which is normally located off-wing in an instrumentation bay. Due to the tubing being internal to the wing, the task of connecting the orifices normally requires the wing skin to be removed, which is a major engineering task.

Like a tubed pressure belt method, the flush surface orifice method also suffers from pneumatic and acoustic lag. The length of tubing used to transmit the pressure from the surface orifices to the pressure sensor creates a lag in the same way that the tubing used in a pressure belt does. This can be alleviated by using pressure sensors that

16 are mounted at the orifice locations on the underside of the wing skin but this requires the use of many sensors and is still difficult to install. In addition, if a problem develops with one of the sensors it is very difficult to fix.

Some of the earliest research flight testing used flush surface orifices to measure pressure distribution. Flight testing carried out in 1927-28 on a Boeing PW-9 Pursuit aircraft used flush through-surface orifices to measure pressures on the right upper and left lower wings [17]. The pressures were recorded using two photographic manometers that were carried in the aircraft [17]. The pressure distributions measured over the wing surfaces for a single flight condition are shown in Figure 2-5.

Figure 2-5: Level flight pressure distributions on a PW-9 pursuit aircraft. Reproduced from [17].

2.2.3 Pressure sensitive paint Pressure sensitive paint (PSP) is a paint that is commonly used in wind tunnel testing for measuring continuous pressure distributions. PSPs are made by dissolving a luminescent dye and a polymer binder in a solvent [18]. The resultant coating is one that changes photoluminescence (fluorescence and phosphorescence) depending on the oxygen content of the air to which it is exposed [19]. Because the oxygen content of air is inversely proportional to the absolute pressure, the photoluminescence (colour) of the paint changes predictably with pressure.

17 The advantage of pressure sensitive paints is that rather than measure the air pressure at a number of discrete points, pressure sensitive paints allow a continuous distribution to be measured.

Although pressure sensitive paints are often used in wind tunnel testing, they have not been used extensively in flight testing due to the complex setup required. The need to illuminate the coated surface with ultra-violet light and to photographically capture an image of the surface limits the areas of an aircraft where this setup can be employed.

One problem associated with using pressure sensitive paint in low speed flow applications is that small changes in pressure only cause small changes in luminescence that are difficult to detect because the sensitivity of the paint to changes in pressure is low [19]. A PSP system that was installed on an aircraft for use during flight would need to provide constant illumination of the surface and the cameras used to capture the distribution image would need to compensate for changing background light conditions.

A few examples exist documenting the use of pressure sensitive paints in flight testing applications. One example used a Beechjet 400A aircraft with a laser scanning system used to excite the paint [18], the setup of which is shown in Figure 2-6. Six test flights were made at night to provide contrast between the backlight conditions and the painted surface. It proved difficult to gather data due to the large temperature variation across the chord of the wing (due to the relatively cool integral fuel tanks) that modified the colour response of the paint, requiring correction [18].

Figure 2-6: Camera setup used to monitor a strip of pressure sensitive paint in-flight. The camera is set up to observe the paint on the top surface of the wing through an over-wing window. Reproduced from [18].

2.2.4 Modern advancements Modern implementations of pressure distribution systems have focused on using modern sensor technology and creating networks of smart sensors.

Work completed in Italy used a networked array of capacitive strips to measure pressure distributions [20]. The strips were quite thin which enabled them to be mounted directly on the wing surface, neglecting the need to transmit the pressures to a bulky pressure sensor located off-wing. An acquisition system was developed to collect data from the capacitive strips using a master-slave digital communications protocol, as shown in Figure 2-7. The system used a single wire, serial bidirectional communication protocol to transmit data between the sensor nodes and the central master device at a rate of 16.3 kbit/s [20].

Figure 2-7: Schematic of the acquisition system architecture developed in [20].

The capacitive strips were calibrated by mounting the strips on a wing profile in a wind tunnel, alongside an array of flush surface orifices that had been connected to a differential pressure sensor. The wing was tested at different angles of attack and free stream velocities to generate a variety of pressure profiles over the wing. The outputs of the strips were then related to the pressure recorded by the orifice to convert the measured capacitance values from the strips to pressure values. The danger with this approach is that the pressures recorded using the orifice method are the true surface pressures, whereas the capacitive strips are experiencing a slightly different pressure due to the effect that the strips themselves have on the airflow. The capacitive sensor strips were found to have a pressure resolution of approximately 5 Pascals [20].

Boeing/Endevco system A recent advanced pressure belt system was developed by Boeing, Endevco (a measurement systems company) and Georgia Institute of Technology for use by Boeing during certification flight testing. The pressure belt was unique in that it used pressure sensors mounted on the wing at each measurement location. The pressure sensors used were piezo-resistive Micro-Electro Mechanical (MEMs) sensors, with the data from the sensors transmitted digitally using a common data bus. The system was developed to replace tubed pressure belt systems that had been used by Boeing on aircraft undergoing flight load tests with a system that improved measurement performance and reduced installation time [6]. A breakdown of the system shown mounted on an aircraft wing can be seen in Figure 2-8.

Figure 2-8: Pictorial diagram of a Boeing aircraft wing with the pressure belt installed showing a breakdown of the components in the pressure belt.

The target accuracy for the system was 0.1% of full scale range over a temperature range of -55°C to 80°C, which would be one order of magnitude better than the accuracy of a tubed belt [21]. The pressure belt segments were designed with an overall height restriction of 0.1˝ (2.54 mm).

Sensor modules were developed that contained the pressure sensor and the integrated circuits required for acquisition and signal processing. The components were mounted on a polymeric tape that also provided the required electrical connections. The segments that formed the pressure belt were taped to the wing surfaces using 3M aluminium tape. The components were faired to the surface using silicon coatings. Perforated metal covers were placed over the pressure sensors to protect the sensors from destruction due to impact from water droplets with very large momentum [21]. A diagram of a belt segment is shown in Figure 2-9, showing the electronic components in a single module.

Figure 2-9: Schematics of the pressure belt segments developed by Boeing. The top schematic is a plan view of a single belt segment while the bottom schematic is a cross-section of a segment. The scales of the two schematics are different. Reproduced from [6].

Initial flight testing of the system was carried out using a Boeing 757-300 undergoing certification and a Boeing 737BBJ that was conducting an air pressure survey [21]. During the testing on the 757, the belts were mounted near a reference pressure port that was connected to a Honeywell PPT pressure transducer. During the test the pressure belt sensors were sampled at 5Hz (The sensors have a resonant frequency of 180kHz but the maximum sampling frequency was not published) [21]. The measurements from the pressure belt were found to be accurate within 0.1% of the full scale of the Honeywell sensor. The testing on the 737 aircraft used traditional tubed pressure belts and the new pressure belt segments for comparison, however direct comparison could not be made between the two methods because the sensor modules were not adjacent to the surface locations being measured by the tubed belt, although both sets of data showed the same trends. A direct comparison between the pressure measured using a tube connected to a pressure sensor, and the MEMs sensor module is shown in Figure 2-10. This data is from an initial flight test of a prototype MEMs sensor module [22] and it shows that the data from the MEMs module contains less variance than the tube data, although it would be expected that the pressures measured using the tube should contain less variance as a length of tube acts as a low-pass filter. Full details of the test could not be found.

Figure 2-10: Direct comparison between pressure measured using a tubed pressure belt and the advanced MEMs sensor module system. Reproduced from [22]. 2.3 Aerodynamic properties measured using pressure belts The NASA F-16XL program previously mentioned used tubed pressure belts to conduct investigations into shock wave positions on an aircraft that had been modified to resemble a proposed supersonic transport aircraft design [11]. These tests were carried out to assess the feasibility of testing a laminar flow control device on the aircraft. Any local shockwaves present could compromise the measurements made with the laminar flow control device installed and so a tubed pressure belt was used to determine the location of any shock waves. An example of the pressure distribution measured over the upper wing surface is shown in Figure 2-11. The two peaks that are marked were found to be due to shock waves from a gun trough on the airframe and the canopy closure joint. The position of these shock waves was evaluated at various angles of attack and Mach numbers to determine the areas of the wing that would be affected by the shock waves. It was found that the installation of a fairing over the gun trough negated the formation of the gun trough shock, but it was found that the canopy closure shock meant that only the forward 65 percent of the wing chord could be used for the laminar flow control experiment. The conclusions of the study were that the accuracy of the pressure distribution measurements may have been affected by localised lifting of the pressure belt tubing during initial flights and the strong spanwise flow that was experienced due to the sweep of the wing. This study highlighted the need for adequate

23 cross-fairing of a pressure belt if accuracy is to be obtained on a surface where spanwise flow is considerable, such as a swept wing.

Figure 2-11: Pressure distribution measured on the upper surface of the F-16XL, showing shock wave locations. Reproduced from [11].

Another NASA research program that used pressure belts to investigate flight at high angles of attack was the HARV program. This program used a modified F/A-18 aircraft that could fly at angles of attack up to 70°. Pressure belts were mounted at three wing stations to measure pressure distributions and from these distributions, calculate the forces acting on the wing sections. Figure 2-12 shows the wing section normal force

(Cn) measured using pressure belts mounted at an inboard wing span location.

Figure 2-12: Section normal force coefficient (Cn) as a function of angle of attack. Reproduced from [10].

24 Lift coefficients were also computed from the measured pressure distributions. The lift coefficient curve shown in Figure 2-13 was measured at an inboard wing station and the two distinct peaks in the lift coefficient were thought to be caused by a strong vortex that was generated by the leading edge extension on the F/A-18 aircraft.

Figure 2-13: Section lift coefficients calculated using the measured pressure distributions from the HARV program. Reproduced from [10].

The effects of sideslip and Mach number on the measured pressure distribution were evaluated. Good agreement was found between data from the left and right wing pressure belts with no sideslip, and the pressure belts were then used to investigate the effects of different sideslip angles on the measured pressure distributions. It was found that any sideslip tended to increase the suction pressures on the upper surface of the windward wing. Any influence the Mach number had on the pressure distribution was found to decrease as the angle of attack was increased beyond 20°.

Another phenomenon that was investigated using the pressure belt measurements on the HARV aircraft was the flow separation patterns over the wing. Flow separation was initially detected using the pressure belts and follow up flights were made using wool tufting to confirm the data from the pressure belts. Separation was detected using the pressure belts when the pressures measured became consistent with no significant variations in the chordwise distribution, although photographs of the wool tufting were used to accurately map the separation patterns. An example of a typical separation pattern is shown in Figure 2-14.

Figure 2-14: An example of the flow separation patterns recorded during the HARV research program. The flow separation patterns were obtained using wool tufting and confirmed with measurements from the pressure belts. Reproduced from [10]. 2.4 Comparisons between pressure distribution measurement methods The main disadvantage of pressure distribution measurement methods that are mounted on the wing surface (such as a tubed pressure belt or Boeing’s advanced sensor module system) is that the true surface pressures are not being measured. The addition of the measurement system has some effect on the airflow over the wing surface.

The investigation by NASA into the effect of a tubed pressure belt thickness [9] that has been discussed found that in the low speed regimes of a light aircraft, the thickness effect was negligible apart from the trailing edge region and this was confirmed by numerical predictions that are discussed in section 4.3.4.

2.4.1 Comparison with wind tunnel testing Comparisons have been made between tubed pressure belts and wind tunnel testing. As part of the certification of the Fokker 100 aircraft, pressure measurements were made in flight and compared with wind tunnel testing [23]. One of the main reasons for making pressure distribution measurements in flight is to provide data for design validation. Often the flight data is compared with wind tunnel data as the latter has been used to design the aircraft structure. The wind tunnel data for the F100 program had been collected using a 1:20 scale model. The comparison between the flight data and the wind tunnel data found good agreement between the measured pressure distributions, as shown in Figure 2-15. In this case, the pressure distributions were measured at the same spanwise wing location on the aircraft as they had been during the wind tunnel testing. The angle of attack that was used to compare the flight data to the wind tunnel data was the aircraft angle of attack rather than the local section angle of attack.

Figure 2-15: Comparison between pressure coefficients measured in flight with wind tunnel data for an F-100 aircraft. The unlabelled x-axis represents the normalised chord-length of the aircraft wing. Reproduced from [23].

From the measured pressure distributions, lift force coefficients were also calculated. In order to calculate accurate lift coefficients from the flight data, the aircraft angle of attack needed to be calibrated accurately. This was computed using the variation in pressure altitude and aircraft pitch angle during turning flight [23]. A regression analysis was used to establish an angle of attack equation which was then used to extract the angle of attack from the recorded flight data. A similar method was used to calibrate the angle of attack vane on the UNSW@ADFA C182RG aircraft, using the measured aircraft pitch angle during steady level flight to establish a relationship between the angle of attack vane reading and the true aircraft angle of attack.

Lift coefficients were calculated from the measured pressure distributions using the measured aircraft angle of attack. In this case, the sectional lift coefficient data from the full scale flight was compared with lift coefficients computed from wind tunnel testing, where the aircraft angle of attack in the wind tunnel was the reference angle of attack. If the calculated lift coefficients were to be compared with sectional lift coefficient data, the angle of attack of the section where the pressure belt was located would need to be computed or measured. In order to compute the sectional angle of attack from the aircraft angle of attack, the distribution of local angle of attack due to the spanwise loading needs to be computed. Lifting line theory can be used to compute the spanwise variations but lifting line theory is based on inviscid theory and is quite dependent on wing geometry. For the F100 study, lifting line theory was only used to correct the flight results for production tolerances, wing deformations due to the flight conditions and the normal loading induced by bank angle, and deformations due to fuel in the wing

27 tanks. A comparison between the lift coefficients measured in-flight and from wind tunnel testing is shown in Figure 2-16, showing excellent agreement between the lift coefficients measured in flight and those measured from wind tunnel testing. Note in Figure 2-16 that the x-axis displaying the angle of attack is shown to be the aircraft angle of attack, αc, which is done so that the lift coefficients from the flight data can be compared directly with the wind tunnel lift coefficients. The reason for not including any units on the x-axis is that the absolute angles are the aircraft angles of attack and are only used to compare the wind tunnel and flight data directly. The reason for not including any units on the y-axis could be that the results are commercially sensitive as they were gathered during testing of a production passenger aircraft.

Figure 2-16: Comparison between lift coefficients measured on the F-100 aircraft in-flight and from wind tunnel testing. Reproduced from [23].

2.4.2 Comparison between flush orifice and pressure belt methods One study carried out by NASA compared the data gathered from a tubed pressure belt and flush orifices [14]. The study was carried out to evaluate the effects that the tubing of a pressure belt has on the local flow characteristics. The largest differences between the two methods was found at quite high Mach numbers (M=0.9) and high angles of attack. At lower Mach numbers and angles of attack, the difference between the two methods was less pronounced, as shown in Figure 2-17 The pressures measured using the external tubing tended to be lower on the upper airfoil surface due to the

28 increased wing thickness from the tubing [14], however the tubing size that was used (4.7mm) was quite large and the study concluded that better agreement would have occurred by using smaller tubing. The discrepancies caused by the tubing resulted in section normal coefficients calculated from the pressure belt data being 10% higher than those calculated using the flush orifices [14].

Figure 2-17: Comparison between pressure coefficients measured using flush orifices and tubed pressure belts. The top graph is a high Mach number condition (M=0.9) and the bottom graph is a low Mach number condition (M=0.5). Reproduced from [14].

29 2.5 Conclusions The current literature relating to the use of tubed pressure belts highlights the inability to make dynamic measurements due to the effects the tubing has on the pressure propagation. The effects of thickness due to a measurement device being added to the wing has shown that for light aircraft, the differences created are small but can be minimised by ensuring that the increase in thickness is also small [9]. It has also been shown that at high speeds (M=0.9) for a reasonably thick pressure belt (4.7mm) the additional thickness can result in normal force coefficients up to 10% higher compared to the flush orifice method. Comparisons between flight measurements and wind tunnel measurements highlight the importance of taking into account three-dimensional effects such as the spanwise distribution of local angle of attack and wing deformations. Recent advancements have been the use of distributed sensor networks and new sensor technologies such as MEMs sensors to measure the surface pressure directly. The next chapter will outline the technical development of the pressure belt system that was developed for use on the UNSW@ADFA aircraft, and describe the bench testing that was carried out to verify the performance of the system.

30 3 Development of the pressure measurement system

In this chapter, the technical development of the pressure distribution measurement system (hereafter referred to as the pressure belt) will be described. The development work encompassed the design of the pressure belt system and its components, the testing of the components and the system, and the manufacture of the system.

The system was intended for use on the UNSW@ADFA Cessna 182RG (C182RG) aircraft; however it was expected that the system could be applied to other aircraft as well as applications in areas such as automobile aerodynamics. For this reason, the pressure belt system was designed so that it was not a customised solution for the C182RG aircraft but rather a system that could be applied easily to any surface that required the distribution of air pressure to be measured.

3.1 Description of the system The most important requirement of the pressure belt system was that it measured the air pressure at a number of discrete locations across the chord of an aircraft wing. Detailed specifications were not made to begin with as the measurement system was developed from scratch, although consideration was given to several aspects of the design. It is also a requirement that the system be able to provide a derived measurement of pressure coefficient (CP) which necessitates measurement of free stream pitot and static pressure (It is intended to monitor existing pitot and static pressure sensors on the test aircraft).

One of the important parameters to be specified was the size of the pressure belt. Tubed pressure belts used in previous work had thicknesses up to 3.3mm [10, 11] and it had been shown by Rivers et al. [9] that the pressure belts up to a thickness of 4.7mm have a negligible effect on the pressure distribution measured, particularly in the low airspeed regimes experienced by a light aircraft [9]. It was decided that the pressure belt should be made as thin as possible within the scope of the budget for the project. The project budget was the restriction on the profile thickness because thinner pressure

31 sensors are generally more expensive. The overall width of the belt was not constrained as this dimension was not restricted by any performance requirements. If the belt was used on a swept aircraft wing where spanwise flow would be present, the minimum width would have to be at least 100mm wide because it had been shown that the wider the associated fairing to the wing surface, the less the spanwise flow is disturbed and the more accurate the pressure measurements become [11].

The pressure belt system design was based on the advanced pressure belt developed by Boeing, Endevco and Georgia Tech [6, 21], outlined in section 2.2.4. The advantages that system provided were very appealing. By using discrete pressure sensors at the points of measurement, a higher sampling rate could be achieved than that of a tubed pressure belt and the difficulties associated with installing and plumbing the numerous tubes would be avoided. By using digital data transmission a higher level of data integrity would be achieved as analog data transmission could be susceptible to electronic noise interference from the airframe, aircraft electrical systems and radios. By connecting the discrete pressure sensors to microcontrollers, the data could be acquired and processed by the microcontrollers at the point of measurement. The appeal of this feature was that the data being output from the system would be the actual pressure readings and would not require any further processing.

Before commencing the detailed design of the pressure belt, a basic system design was finalised (Figure 3-1). The pressure belt system would consist of modules with small discrete pressure sensors that would be connected to and sampled by microcontrollers. The microcontrollers would be connected to a digital data bus that would transmit all the data from the sensor modules to a computer in the aircraft digitally.

Figure 3-1: Diagram showing the pressure belt system configuration 3.2 Pressure Sensor The main requirement of the system was to measure pressures at various points over the surface of a wing. This meant that discrete pressure sensors would be used, as opposed to a distributed measurement medium such as pressure sensitive paint. The specifications for the system also required that the components mounted on the wing maintained a small profile thickness. This provided the two main requirements for the pressure sensors. The sensors had to be small and had to either output digital data or be capable of being sampled by the microcontroller’s analog to digital converter. The other consideration was that the cost of the pressure sensors fell within the project budget.

The pressure sensor used in the system was selected after considering several possible sensors. The sensors that were considered are listed in Table 1, along with the sensor price as quoted by suppliers.

33 Table 1: Comparison between pressure sensor characteristics Sensor size in Price per mm (width x Interface Sensor sensor width x thickness Sensor type Range method (AUD) or diameter x thickness)

Honeywell 0-34.5 kPa $51.95 9.9 x 9.9 x 2.7 Analog Differential CPX05GF (0-5 psi) Honeywell 0-103.4 kPa $33.00 9.9 x 9.9 x 2.7 Analog Absolute CPX15A (0-15 psi) VTC $25.00 7.1ø x 2.1 Digital Absolute 30-120 kPa SCP1000 Endevco 0-103.4 kPa $1780.00 6.3ø x 0.76 Digital Absolute 8515C (0-15 psi) Kulite 0-172.4 kPa $1735.00 9.6 x 4.1 x 1.5 Digital Absolute LL072 (0-25 psi)

The Endevco sensor was considered as that was the sensor used in the advanced pressure belt developed by Boeing in conjunction with Endevco and Georgia Tech [21]. Both the Endevco sensor and the Kulite sensor are very accurate and small pressure sensors, however the cost ruled them out of consideration.

Of the remaining sensors, the Honeywell sensors were initially selected. At the time, the SCP1000 sensor was not available and the Honeywell CPX series offered a pressure sensor in a small package at a reasonable price. The Honeywell CPX sensors are piezo-resistive sensors that contain thin silicon diaphragms that expand and contract as the air pressures acting on either side of the diaphragm change. A Wheatstone bridge circuit is used to give a voltage output that varies linearly with applied pressure.

Of the two Honeywell CPX sensors, it was decided to use the differential pressure sensor as opposed to the absolute pressure sensor. The differential sensor had a higher sensitivity (voltage output per unit of applied pressure) than the absolute sensor which would ensure a high signal to noise ratio when used with an analog to digital converter, and allow small changes in pressure to be resolved.

It was originally envisaged that one port of the sensor would be sealed with a known reference pressure. It was thought that this would enable the range of the sensor

34 to be customised for its application by providing an initial pressure offset if required. The reference pressure selected to be applied to the sealed side of the sensor was a sea level reference pressure.

This plan to seal one side of the sensor with a known reference pressure proved to be unfeasible. The reason for this was that the pressure of the air that was sealed in the reference side of the sensor at sea level pressure would not remain constant. The ideal gas law (equation 1) shows that for a fluid that remains at a constant density (incompressible) the pressure is inversely proportional to the temperature. This meant that the small volume of air trapped when the sensor was sealed caused the pressure output of the sensor to be extremely sensitive to changes in temperature.

PRT= ρ ∆=PRT(ρ ). ∆ (assuming the trapped air in incompressible) for a 1° change in temperature (1) ∆=P (1.225 × 287.0528) × (1) ∆=P 351.6 Pa

As equation 1 shows, a 1° change in temperature would cause a 350 Pa change in pressure output. Measuring the temperature would enable the pressure readings to be corrected but the temperature measurements would need to be accurate to 0.02° in order to get a pressure accuracy of 10 Pa. This would require an extremely accurate temperature sensor with high resolution in addition to the pressure sensor.

At this point, a considerable amount of time had been spent developing the measurement system around the CPX05 sensor. When another sensor was sought to replace the CPX05, the SCP1000 pressure sensor was discovered.

3.2.1 SCP1000 pressure sensor The SCP1000 absolute pressure sensor is manufactured by VTI Technologies, a Finish company. The pressure sensor is a Micro-Electro Mechanical (MEMs) capacitive element sensor. A capacitive element pressure sensor uses a thin diaphragm that acts as one plate of a capacitor, with deflections in the plate due to pressure changes causing changes in capacitance that can be measured [24]. The SCP1000 sensor uses a thinned silicon wafer to form the diaphragm that flexes due to the difference in the exterior pressure and an internal vacuum reference [25].

Figure 3-2: VTI Technologies SPC1000 series absolute pressure transducer. Reproduced from [26].

This sensor had several advantages over the Honeywell CPX sensors. The digital output of the sensor reduced the complexity associated with converting a small analog sensor signal to digital level using the microcontroller. The SCP1000 sensors were also cheaper than the Honeywell CPX sensors and had internal temperature compensation as well as providing a temperature measurement output. The characteristics of the SCP1000 sensor are shown in Table 2

Table 2: Characteristics of the SCP1000 pressure sensor Feature Value

Supply Voltage 2.4-3.3V Pressure measurement range 30kPa-120kPa (absolute) Temperature measurement range -20°C - 70°C 9 Hz (High speed mode) Pressure data output refresh rate 1.8 Hz (High resolution mode) Pressure resolution 3 Pa (High speed mode) Relative pressure accuracy ±50 Pa Proof pressure 2.0 MPa Digital pressure word length 19 bits Digital temperature word length 14 bits

The SCP1000 sensor provided high accuracy at selectable data output rates. As Table 2 shows, the sensor is capable of operating in two modes, high speed and high resolution. The high speed mode offers a lower resolution but provides a sampling rate of 9 Hertz which would be needed if the sensors were used to investigate dynamic changes in pressure. The sensors also have the ability to operate in a triggered mode, enabling multiple sensors to make synchronised measurements.

36 3.3 Sensor modules In order to make the pressure belt system adaptable, the pressure sensors were developed into sensor modules. Each module contains all the components associated with operation of a single sensor. These components include the microcontrollers, voltage regulators, smoothing capacitors and the digital bus interface chips.

Two different modules were developed; a single sensor module and a high spatial resolution module, both of which are shown in Figure 3-3. The high resolution module consisted of seven pressure sensors connected to a single PIC microcontroller. This module was developed for use in areas where large changes in pressure occur over small areas, such as around the leading edge of an airfoil. The spacing between the sensors on the high resolution module was fifteen millimetres.

Figure 3-3: Single sensor module (left) and high resolution sensor module (right).

In turn, two versions of the high resolution module were designed; a left handed and a right handed module. This was done so that when the boards were mounted on the upper and lower leading edge surfaces of a wing, the sensor rows could be lined up while keeping the portion of the module with the microcontroller and other circuitry positioned back from the leading edge away from the area of greatest curvature.

Each module was designed with two four-way connectors for power and data connection and carry-through. These connectors enable the modules to be quickly and easily connected in any order or configuration. Strips of ribbon cable were used to make the connection cables for joining individual modules to form a sensor array.

37 3.4 Microcontrollers Microcontrollers are contained on each sensor module to control the pressure sensor and acquire its data, and to interface with the digital data bus. The microcontrollers selected were Programmable Intelligent Computer (PIC) microcontrollers, manufactured by Microchip. These microcontrollers are widely used in relatively simple applications that require an integrated circuit processor but where the application does not require particularly advanced functions.

Two different microcontrollers are used in the system, a master and a slave. The master is used to control the data flow between the PC gathering and displaying the collated data and the digital data bus. Each sensor module uses a slave microcontroller to interface with the pressure sensor. The characteristics of the two microcontroller models are shown in Table 3.

Table 3: Characteristic of the Microchip Microcontrollers Operating Package size in Microcontroller Model Pin count voltage mm (h x w x t)

Slave PIC18F24J10 2.0-3.6V 28 10.2 x 7.8 x 2 Master PIC18F6720 2.0-5.5V 64 12 x 10 x 1.1

The software for the microcontrollers was developed using MPLab, a programming environment developed by Microchip for use with the PIC Microcontrollers. The programming environment uses the C programming language as a high level language to write the software that will execute on the microcontroller. The MPLab software then compiles the C code into assembly code, the register based code used in the microcontrollers. It is possible to write software using the assembly language, but the use of a high level language such as C enables complex functions to be written easily and quickly. A full breakdown of the microcontroller software and the internal functions can be found in Appendix D. A brief description of the functions performed by the master and the slave microcontrollers follows.

3.4.1 Slave microcontroller The main function of the slave microcontroller is to retrieve data from the SCP1000 pressure sensor. The microcontroller communicates with the sensor using the

38 Serial Peripheral Interface (SPI) protocol. This protocol is a serial data transmission protocol that uses two data lines and a third line to clock the data transmission.

The slave microcontroller also controls the measurement mode of the pressure sensor (High Speed or High Resolution) and, if the modules are operating in triggered mode, the slave microcontroller will provide the physical trigger signals to the pressure sensor.

The other main function of the slave microcontroller is to send the data that was retrieved from the pressure sensor to the PC to be recorded and displayed. The data is transmitted over the digital data bus at a speed of 62,500 baud. The data transmission is performed in a daisy chain fashion, with each slave microcontroller sending data packets immediately after a previous sensor module has sent its data packets. When multiple pressure sensors are used together, the sensors take measurements synchronously but the data is sent from all the slave microcontrollers in a serial manner at high speed. This negates the need for the master microcontroller to individually prompt each module for its data and speeds up the transmission process.

3.4.2 Master microcontroller The main function of the master microcontroller is to relay data and commands between the PC and the sensor modules. Along with the PC, the master microcontroller forms the master node of the master-slave communications framework. The reason for having a master microcontroller as well as the PC is to convert between the RS232 data transmission protocol used by the PC and the RS485 data transmission protocol that is used by the sensor modules. A PIC18F6720 microcontroller was used as it has two serial communications ports that enable it to perform this function.

Figure 3-4: Schematic of the relationship between the master and slave microcontrollers (right), compared with a standard bus network topology (left). Left picture reproduced from [27].

39 The circuit board that the master microcontroller is mounted on also contains a voltage regulator to supply voltage for all of the sensor modules. Power is supplied at 6 volts from a Nickel-Cadmium battery which is regulated to 5 volts for the master circuitry. The master regulator supplies all the sensor modules with a 5 volt supply which is in turn regulated to 3.3 volts using regulator chips on each sensor module.

3.5 Network topology The digital data communications between the slave and master nodes is based around the RS485 protocol. This protocol is similar to the RS232 protocol that is widely used for PC serial communications. RS485 was developed to ensure robust error free communications over large distances and in electrically noisy environments such as factories [28]. This makes the RS485 protocol ideal for use as the communications protocol in the flight environment where it was expected that the data bus would be subjected to high levels of electronic noise. The sources for this noise would be the aircraft radios and the airframe which is used as the electrical ground plane for all the electrical systems in the aircraft.

The RS485 protocol uses two differential signal wires to achieve a high level of noise tolerance. By using differential signal lines, any noise signal on the signal lines should be common to both lines and will be cancelled out.

Another feature of the RS485 protocol is 9-bit data transmission which enables address detection to be implemented by the slave microcontrollers. Normally digital data is transmitted in blocks that are 8 digital bits in length. The 9th bit is used by the RS485 protocol to indicate whether the previous 8 bit data block is an address. The microcontrollers have the ability to ignore any data where the 9th bit is a zero, but when a command is issued to a particular microcontroller, the 8 bit block will contain the address of a particular slave microcontroller and the 9th bit will be set to a digital one. All the microcontrollers will read in the address but only the slave microcontroller whose address matches the 8-bit data will disable its address detection ability to read in subsequent commands. The other microcontrollers will continue to disregard any data without the 9th bit set to a digital level of one.

The network was configured using a half-duplex configuration. This means that each node in the network communicates using one set of signal lines (2 differential

40 wires). In a full-duplex configuration, two sets of signal lines (4 wires) are used, with one set of lines used to send data from the master node to the slave nodes and the other set of lines used to send data back from the slaves to the master. The disadvantage of the half-duplex configuration is that the data bus can only be used by one node at a time to send data, whereas the full duplex configuration allows the master to send data while one of the slave nodes is also sending data. Another disadvantage of a half-duplex configuration is that the data transmission rate is reduced compared to the full-duplex configuration, but the advantages are that the half-duplex configuration reduces complexity, components and power consumption (the full-duplex configuration requires interface chips that draw more power).

3.6 Circuit Board The components required for each sensor module were mounted on a circuit board to provide the required electrical connections. It was decided that the circuit board would form the base of the sensor modules in a similar manner to the Boeing pressure belt system. The circuit board design and layout was done using Altium Designer. Examples of the circuit board layout are shown in Figure 3-5, and the circuit board schematics for the sensor modules can be found in Appendix C.

Figure 3-5: Circuit layout of the single sensor circuit (left) and the high density circuit (right).

Because the circuit board was to form the base of each sensor module, it was decided to manufacture the circuit boards from a flexible circuit board material. Unlike the Boeing pressure belt segments, the circuit board material used was 0.2mm fibreglass

41 instead of polyimide as the fibreglass is much tougher while still being flexible enough to be mounted around the leading edge of the aircraft wing.

The circuit boards for the modules were manufactured by Lintek, a local company. A large panel containing 6 high density module boards (3 left-handed, 3 right-handed) and 51 single module boards was manufactured. The copper tracks on the circuit boards were coated with a silver finish and soldermask was applied on both sides of the circuit boards.

3.7 Cost breakdown The total cost of the sensor modules, including outsourced labour was calculated to be $87.40 for a single sensor module and $260.70 for a high resolution module. The bulk of the cost comes from the pressure sensors themselves ($25.00) with the mounting costs also forming a large part of the overall cost due to the need to mount the pressure sensors using optical placement and a reflow oven for the soldering process. The breakdown of cost is shown in Table 4. The cost of the modules is very cheap compared to the Boeing/Endevco system where the sensors alone cost $1780.

Table 4: Cost breakdown for sensor modules. Component Single sensor module High resolution module

Pressure sensors $25.00 $175.00 Electronic components $12.95 $12.95 Circuit board $17.45 $17.45 Component mounting $32.00 $55.30 (Labour) Total cost $87.40 $260.70

3.8 Computer software In addition to the microcontroller software, software was also written to be run on a PC to display and record the data from the sensor modules. Displaying processed data from the sensors as the data is gathered was one of the aims of the project as this makes the pressure belt a useful tool for the flight laboratory program, allowing students to see results from the pressure belt as they fly.

42 The software package used to develop the software was Labwindows/CVI. The software is a development package used for creating data gathering and displaying software with graphical user interfaces. The software code for the graphical user interface was written using the C programming language.

The user interface was broken into several panels as can be seen in Figure 3-6. Each panel was used to display information about a particular aspect of the pressure belt system operation, such as the panels for configuring the belt at the start, the ground initialisation and the panels displaying the data in flight.

Figure 3-6: Screenshots showing key tab panels from interface software. Top left shows data from the aircraft instrumentation while top right shows traces of the pressure for each pressure sensor. Lower left shows the pressure coefficient plot in-flight while lower left shows the in-flight lift coefficient display.

One of the displays that was useful during the flight testing was a plot showing the calculated lift force coefficient against section angle of attack that updated in real time. This plot allowed the lift curve to be observed as the data was gathered and immediately showed areas where more data was needed. It also allowed the dynamic effects from various manoeuvres to be observed. The other displays that were used during flight testing were the displays of the aircraft auxiliary instrumentation and the real-time pressure coefficient display.

43 3.9 Test aircraft description The testing of the pressure belt system was carried out using a light aircraft owned and operated by the School of Aerospace, Civil and Mechanical Engineering at UNSW@ADFA. The aircraft is a Cessna Skylane 182 RG aircraft, of which a 3 view diagram is shown in Figure 3-7.

Figure 3-7: Cessna 182RG three view drawing. Reproduced from [29].

The aircraft was purchased in 1998 and is used by the school as a flying laboratory to support courses in aircraft performance, stability and control [30]. The aircraft has several items of auxiliary instrumentation installed that provide data such as flight angles, control positions, accelerations and angular rates. Table 5 contains a list of the available auxiliary instrumentation.

44 Table 5: Existing aircraft sensors Flight Parameter Sensor used to measure parameter

Port on air data boom, Borgelt pressure Dynamic pressure sensor Port on air data boom, Borgelt pressure Static pressure sensor Aircraft angle of attack Alpha vane on air data boom Aircraft sideslip angle Beta vane on air data boom Aileron position Control position rheostat on right aileron Elevator position Control position rheostat on elevator Rudder position Control position rheostat on rudder Inclinometer mounted in baggage Aircraft pitch angle compartment

The signals from the aircraft sensors are acquired, amplified and conditioned using a central data box that is carried in the cabin of the aircraft. The data box also contains batteries for powering the sensors that can be charged using the aircraft’s electrical system via a cigarette lighter port.

The output from the data box is an amplified and conditioned analog voltage signal for each of the sensor channels. These signals were acquired by the PC using an eight channel National Instruments USBDAQ data acquisition module. Because only eight channels could be measured, the parameters recorded were limited to the flight angles (alpha, beta, aircraft pitch) and the control positions (elevator, aileron, and rudder). The remaining two channels were used to monitor the voltages of the batteries supplying the pressure belt system and the data box.

It was originally intended to monitor the two channels from the data box supplying dynamic pressure and static pressure information from the Borgelt unit. However, during testing of the sensor modules with the data box it was found that the Borgelt unit in the data box that provided the pressure data was not temperature compensated and as a result the pressure signals drifted with changing temperature. It was decided to replace the Borgelt unit with two sealed pressure sensor modules.

45 3.9.1 Air data instrumentation The air data instrumentation used to replace the Borgelt unit consisted of the two prototype single sensor modules that were built to test the circuit design and the microcontroller software. These prototype sensor modules were not mounted on flexible circuit boards and had only been used for developing and testing the microcontroller software.

The modules were mounted in sealed boxes made from cast aluminium, as shown in Figure 3-8. These boxes formed a chamber into which the pressures from the air data boom pitot and static ports were fed. A nipple fitting was attached to each of the boxes to connect the plastic tubing from the air data boom static and pitot ports. The prototype sensor modules were connected to the data bus in the same manner as the rest of the pressure belt sensor modules. From the data gathering software perspective this setup was simpler than if the Borgelt unit had been used because it simply meant that two channels of data from the pressure belt data stream contained the reference pressures used to compute the pressure coefficients.

Figure 3-8: Air data instrumentation consisted of two single sensor modules housed in sealed aluminium die-cast boxes with the pressure inputs from the air data boom connected. 3.10 Sensor Characterisation Following manufacture of the sensor modules and prior to integration into the pressure belt, tests were carried out to verify the sensor characteristics against the

46 manufacturer’s specifications. Calibrations were carried out at different temperatures, in addition to long term comparisons against local weather station pressures. Finally, the frequency response of the sensors was investigated. All of the sensor characterisation testing was carried out using a single pressure sensor.

3.10.1 Sensitivity check The sensitivity and accuracy of the sensors was verified by using a pressure calibrator. A long term comparison was made with a portable weather station and the Canberra Airport Bureau of Meteorology weather station.

The first testing that was carried out used a Druck DPI610 pressure calibrator with a ±1 bar (±100 kPa) range to calibrate the SCP1000 sensor over its full range of 30kPa to 120kPa absolute pressure. The accuracy of the pressure calibrator was 0.025% of full scale (25 Pa) [31]. A comparison between the applied pressure and the SCP1000 pressure is shown in Figure 3-9. The linearity of the best fit line is good, with an R2 value of 1 indicating a near perfect linear regression fit. The slope of the linear regression line shown in Figure 3-9 is 0.998, indicating that the sensitivity of the sensor may be in error. However the error is not large; with an applied vacuum pressure of 60kPa the SCP1000 sensor would register a reading of 59.922kPa. This error of 78Pa is well within the manufacturers quoted absolute accuracy of ±150Pa for the temperature range [26].

47 Sensor 5 calibration (24ºc)

20000

10000

Applied pressure (Pa) 0 -70000 -60000 -50000 -40000 -30000 -20000 -10000 0 10000 20000

-10000

-20000 y = 0.9987x 2 R = 1 -30000

-40000 Sensor pressure difference (Pa) difference pressure Sensor -50000

-60000

-70000 Figure 3-9: Sensor calibration at room temperature.

The error at each point during the test was also calculated and the average error, maximum error and standard deviation on the error shown in Table 6. The average error is low at 0.40%, with the maximum error during the test being 1.80%. The standard deviation for the error was calculated to be 0.43%. The error was calculated as a proportion of the reference pressure at each point, not the full scale range as is often the practise of sensor manufacturers.

Table 6: Error from room temperature sensor calibration on sensor 5 Average Error 0.40% Maximum Error 1.80% Standard deviation 0.43% The long term comparisons were made using a Kestrel 4000 hand-held weather tracker. The specified accuracy of the Kestrel is ±150Pa (similar to the SCP1000) and the unit has a pressure resolution of 10Pa [32]. The test was carried out over one week with the sensor module and the Kestrel being placed in a covered shed that was open to outside air pressure and temperature. Figure 3-10 shows the SCP100 pressure compared with the pressure measured by the Kestrel over the week.

48 Comparison between Kestrel and SCP1000 over a week 95400

SCP1000 Pressure 95200 Kestrel corrected for altitude

95000

94800

94600 Pressure (Pa) Pressure

94400

94200

94000 5/05/2008 0:00 6/05/2008 0:00 7/05/2008 0:00 8/05/2008 0:00 9/05/2008 0:00 10/05/2008 0:00 11/05/2008 0:00 12/05/2008 0:00 13/05/2008 0:00 Time Figure 3-10: Comparison between the pressures measured by the SCP1000 sensor module (black) and a Kestrel 4000 portable weather station (red).

The average error was calculated to be 0.023%, with a maximum error of 0.09% and a standard deviation of 0.017% as shown in Table 7. Due to its function as a weather station, the Kestrel converts local static pressure readings to estimates of mean sea level (MSL) pressure, which were corrected for the altitude at which the test was conducted.

Table 7: Error from week long comparison with Kestrel 4000 portable weather station Average error 0.023 % Maximum error 0.09 % Standard deviation 0.017 %

Figure 3-11 shows the relationship between the corrected Kestrel pressure and the SCP1000 pressure. The graph confirms the linear performance of the SCP1000, assuming the Kestrel is performing as specified.

49 Kestrel Pressure versus SCP1000 Pressure 95400

95200 y = 1.0001x R2 = 0.9938

95000

94800

SCP1000 Pressure (Pa) Pressure SCP1000 94600

94400

94200 94000 94200 94400 94600 94800 95000 95200 95400 Kestrel Pressure (Pa) Figure 3-11: Linearity comparison between the SCP1000 sensor module and the Kestrel 4000.

Figure 3-12 shows the comparison between the SCP1000 sensor and the Bureau of Meteorology (BOM) Canberra Airport weather station over 48 hour period. The constant offset of approximately 280Pa shown in the graph is due to the elevation difference between the airport weather station and the shed where the SCP1000 was located.

50 BOM pressure versus SCP100 pressure 95700

BOM station pressure 95600 SCP1000 Pressure

95500

95400

95300

95200 Pressure (Pa) 95100

95000

94900

94800 10/05/2008 0:00 10/05/2008 4:48 10/05/2008 9:36 10/05/2008 0:00 11/05/2008 4:48 11/05/2008 9:36 11/05/2008 0:00 12/05/2008 10/05/2008 14:24 10/05/2008 19:12 10/05/2008 14:24 11/05/2008 19:12 11/05/2008 Time Figure 3-12: Comparison between the Canberra airport weather station pressure (black) and the SCP1000 sensor module pressure (red).

3.10.2 Temperature Stability The SCP1000 sensors are temperature compensated to correct for changes in response of the capacitive diaphragm element with temperature. The factory calibration process that the sensors undergo involves testing the sensors at several temperatures and fitting a calibration curve to the results. Some initial testing of the prototype sensor modules indicated that the sensors may have had erroneous responses at temperatures other than room temperature. These initial tests involved subjecting the prototype sensor modules to varying temperatures at constant pressure although the pressure was not monitored independently during the test so it was unclear as to whether the changes in pressure indicated by the sensors were a genuine erroneous response. Follow up tests were conducted using the Druck pressure calibrator and an environmental chamber to vary the temperature. The sensor was soaked at four temperatures; -17°C, -6°C, 3°C and 9°C representing a range of temperatures that the sensors would be subjected to in operation. At each temperature point, a calibration process identical to that used for the initial room temperature calibration was used. Figure 3-13 shows a typical calibration with a linear best fit line, in this case for a temperature of -17°C. The linear best fit curves are very similar in slope to the room temperature calibration (Figure 3-9). The

51 slope of the best fit lines relative to the Druck pressure varied from 0.9972 at -17°C to 0.9983 at 9°C. This means that if the sensor was to be subjected to a true absolute pressure of 50,000Pa (a pressure altitude of 18350ft), the sensor would record pressures ranging from 50,077Pa (9°C) to 50,127Pa (-17°C).

Sensor 5 calibration (-17ºc) Applied Pressure (Pa) 0 -60000 -50000 -40000 -30000 -20000 -10000 0 y = 0.9972x R2 = 1 -10000

-20000

-30000

-40000 Sensor pressure difference (Pa) difference pressure Sensor

-50000

-60000 Figure 3-13

Table 8 shows the average error, maximum error and the standard deviation in error for each of the temperatures at which the sensor was tested. The average errors recorded are all under 1% and are comparable to the average error recorded during the room temperature test.

Table 8: Errors in sensor response at varying temperatures. Temperature -17°C -6°C 3°C 9°C

Average error 0.39 % 0.45 % 0.33 % 0.29 % Maximum 1.12 % 1.59 % 1.36 % 1.31 % error Standard 0.24 % 0.42 % 0.34 % 0.30 % deviation

3.10.3 Frequency response Investigations were carried out to look at the frequency response of the sensors. The main reason for doing this was to establish whether the quality of data from the sensors could be improved by using low-pass filtering, and to determine the suitability

52 of the sensors for making dynamic pressure measurements. From the initial testing of the sensor modules, it could be seen that scatter was present in the data. It was thought that if there was high frequency noise in the sensor signal then an appropriately designed low-pass filter could be used to remove these components of the signal.

The first step was to establish a typical noise base for the sensors. The noise base represents the frequency response of the sensor when the sensor is subjected to a fixed pressure (a DC input). In the context of the sensors application, the noise base represents what the sensors measure when the aircraft is motionless. When the aircraft is flying, the frequency response will show magnitude and frequency components above the noise base representing the pressure signal applied to the sensor. If the noise base had high frequency components then application of a low pass filter would be justified.

The noise base data was gathered over a one hour period during which the sensor was stationary in an office. The data was gathered using the triggered acquisition mode with a triggering frequency of 5Hz used. Although the maximum sampling frequency of the pressure sensors is 9Hz, this frequency depends on the operating conditions the sensor is subjected to and from experimentation, it was found that 5Hz was the highest frequency at which all the pressure sensors could be triggered synchronously. During the hour, the air pressure varied due to atmospheric conditions resulting in the pressure time-series as shown in Figure 3-14. The data was detrended using Matlab to remove these low frequency effects and to remove the DC components. The result of the detrending is shown in Figure 3-15.

Figure 3-14: Raw data gathered for establishing a noise base.

Figure 3-15: Detrended pressure readings used to establish a noise base.

The measured noise base spectrum is shown in Figure 3-16. The noise base is almost constant at 5 decibels up to the Nyquist frequency of 2.5 Hz for the triggered mode. The peak shown at very low frequencies is due to the inability of the detrending process to remove all of the dc components of the signal.

Figure 3-16: Noise base spectrum for a typical sensor module.

After establishing a noise base, tests were made to measure the frequency response of the sensor when subjected to signals of different frequency. This was done by using a syringe fitted to a linear actuator to cycle the pressure at controlled frequencies. The resultant pressure variation that the sensor was subjected to resembled a triangular wave form. The mean pressure applied during the tests increased over time and detrending was again used to remove the DC components.

The frequency spectrum recorded by the sensor when subjected to a 1Hz waveform is shown in Figure 3-17. The sampling frequency of the sensor was 5Hz. The spectrum shows clear peaks at 1Hz and 2Hz as expected, with the 1Hz fundamental component containing more power than the 2Hz harmonic component. The results for a 2Hz waveform (Figure 3-18) show the reverse, with the 2Hz peak dominating the 1Hz peak. In this case the peak at 1Hz is an alias. The peak at 1Hz occurs because the spectrum is symmetrical about the Nyquist frequency of 2.5Hz and is due to the harmonic component at 4Hz being folded back into the lower half of the spectrum.

Figure 3-17: Power spectral density showing the sensor response when subjected to a 1Hz triangular waveform

Figure 3-18: Power spectral density showing the sensor response when subjected to a 2Hz triangular waveform

The frequency response from the sensor when subjected to a waveform showed that the noise base in between the signal peaks was only slightly higher at 10dB than the noise base of 5dB. It was decided from this testing that the use of a low pass filter

56 would not be advantageous as the pressure signals from the sensors contained no noticeable noise components and the increase in the noise base was negligible.

3.11 Error Analysis An error analysis was carried out to estimate the error that could be expected in the pressure coefficient measurements due to the error from the pressure sensors. The method used to estimate the propagation of uncertainty into the pressure coefficient calculation was the same method used to estimate the error in pressure coefficient for the tubed pressure belt on the NASA HARV aircraft [10]. For the pressure belt on the HARV aircraft, it was calculated that the 95 percent confidence interval uncertainty for pressure coefficient (Cp) was ±0.05. A description of the error propagation calculation method can be found in in Appendix B of [33].

For this estimate of the error propagation, two cases were considered; a worst case error, as well as a likely error. The worst case error was based on the absolute accuracy of the pressure sensors, as specified by the sensor manufacturer [26]. The likely error case was calculated using the relative accuracy of the pressure sensors. The relative accuracy would be the more correct error to assume because the pressure coefficient is calculated using the differences between the wing sensors and the static and pitot pressure sensors. This means that absolute pressure inaccuracies do not affect the calculated pressure coefficients as strongly as any relative pressure inaccuracies would.

The absolute accuracy of the SCP1000 sensors specified by the manufacturer is ±200 Pa over a temperature range of -20°C to +70°C while the relative accuracy is ±50 Pa over a temperature range of +10°C to +40°C. The precision of the sensors was assumed to be ±10 Pa based on the variance that had been measured using the detrended data for the frequency response tests (Figure 3-19). The error was calculated assuming flight conditions of cruise velocity (77 m/s) and a pressure altitude of 2300m (approximately 7500 feet).

Figure 3-19: Variance plot indicating the precision of the SCP1000 sensor.

As Table 9 shows, the 95% confidence error in measured pressure coefficient would be between ±0.04 and ±0.072. These values are comparable with an error of ±0.05 calculated for the tubed pressure belt used by NASA on the HARV aircraft [10].

Table 9: Summary of expected error due to sensor characteristics. Precision 95% confidence Case Bias error error Cp error Worst case (absolute accuracy) ±200 Pa ±10 Pa ±0.072 Best case (relative accuracy) ±50 Pa ±10 Pa ±0.040

3.12 Certification of the pressure belt The attachment of the pressure belt system to the aircraft was covered by a Civil Air Regulations part 35 (CAR 35) approval. CAR 35 covers the approval of design of modification or repair, under which the attachment of the sensor modules falls as they become a modification to the aircraft. CAR 35 specifies that any modification must be approved by either the Civil Aviation Safety Authority (CASA), or an authorised person who is referred to as a CAR 35 delegate. The authorised delegate approves a modification application once they are satisfied that the modification will not affect the safety of the aircraft.

58 When the CAR 35 application was made, it was agreed with the CAR 35 delegate, Auto Avia Design, that the initial placement of the sensor modules would be on the constant chord (flapped) section of the wing and the sensors would not be placed on the flap surfaces. For future work, the CAR 35 approval could be extended to include other areas of the wing including control surfaces although this approval would be more difficult to attain and would probably require testing of the sensor placements on the aircraft in flight to check that the safety of the aircraft had not been compromised.

The sensor modules were approved for placement between the wing root and the end of the constant chord section. The sensor modules were approved to be taped to the wing surface using 3M 425 aluminium backed tape in the manner shown in Figure 3-20.

Figure 3-20: Sensor module installation specification. Reproduced from [34].

The CAR35 engineering instruction also contained a procedure for installing the sensors, including requirements on cleaning the surface before the sensor modules were attached. Other requirements such as before flight checks and ongoing maintenance checks were also specified. The maintenance release for the aircraft was annotated to note the installation of the pressure belt system when it was installed in accordance with the engineering instruction. The supporting documentation for CAR 35 approval and the subsequent engineering instruction sheet for attachment of the sensor modules to the wing can be found in Appendix B.

3.13 Conclusions The development of the pressure distribution measurement system has been outlined in this chapter. The framework for the system was based on an advanced

59 pressure belt developed for use by Boeing. The system developed for this work cost significantly less to develop than the cost of acquiring the Boeing/Endevco system.

The system consists of low-profile pressure sensors interfacing with microcontrollers. The components for each sensor module are mounted on flexible circuit board material, with two types of sensor modules developed, a single sensor module and a high density module with seven sensors.

The characteristics of the sensor modules were confirmed with tests using a pressure calibrator and long-term comparisons with weather stations. The frequency response of the sensors was also tested. An error analysis was carried out to evaluate the propagation of errors from the sensors through to the calculated pressure coefficients. It was found that the 95 percent confidence error in the pressure coefficient would be between ±0.040 and ±0.072 which is comparable to the error in a typical tubed pressure belt.

The next chapter describes the numerical predictions of pressure distribution. The numerical methods were used to provide a comparison with the results gathered during flight testing and to investigate aspects of the pressure belt system configuration.

60 4 Numerical Investigations

The main aim of this work was to develop a system for measuring the pressure distribution over an aircraft wing in flight. The system was tested on a light aircraft owned by UNSW@ADFA. In order to evaluate the data measured experimentally, numerical predictions of the pressure distribution were made for comparison. This chapter describes the numerical methods used to calculate the pressure distributions.

The chapter begins with a description of the two numerical methods used; the Hess-Smith method and the XFOIL software. The Hess-Smith method was implemented using Matlab and the results from XFOIL were analysed using Matlab. The results from investigations made with the numerical methods are then discussed. The final part of this chapter will consider the corrections needed to compare two- dimensional predicted pressure distributions with experimental data from a finite three- dimensional aircraft wing.

4.1 Background to numerical flow calculation methods Panel methods are a numerical method used to solve the flow conditions over an airfoil. Under incompressible, inviscid, irrotational flow conditions, the Euler equations describing compressible inviscid flow, simplify to Laplace’s equation [35]. Laplace’s equation describes the fluid flow in terms of a velocity potential,ϕ . The Laplace equation is contained in Appendix A.

The Laplace equation describes an incompressible, irrotational flowfield and can be solved to satisfy certain boundary conditions using either finite-difference or finite- volume methods, or by using a superposition of flows that individually satisfy the Laplace equation (panel methods). Of the two solution methods, panel methods are the more efficient and practical to implement as they take advantage of the linearity of the Laplace equation and they avoid the need to generate a grid within the flowfield (which is required to solve finite-difference or finite-volume problems).

Panel methods use the principal of superposition of flows to model the flow over an airfoil. The approach normally used is to express the flow field in terms of the

61 velocity potential based on two or more elemental flows (source, sink or vortex) in the presence of a uniform onset flow. The elemental flows and the uniform onset flow both satisfy the Laplace equation individually, and because the Laplace equation is linear, a combination of the flows also satisfies the equation. Figure 4-1 illustrates the principle of superposition of flows.

Figure 4-1: Superposition of elementary flows. The stream functions of the elements are given in polar coordinates. Reproduced from [3].

Panel methods are so named because the principle is to divide the airfoil into a number of short straight line segments; the panels. Each panel has either a single fluid element or a combination of fluid elements positioned at its midpoint. A common combination of elements is the source-vortex combination. Each of the individual elements is given a particular strength that determines its effect on the overall flow.

Figure 4-2: Panel codes derive their name from the division of an airfoil into numerous short, straight line segments.

Boundary conditions are applied to obtain a unique solution of the Laplace equation. The boundary conditions commonly used are that the flow must only be tangential to the panels that make up the airfoil surface (no flow normal to the airfoil surface) and that the Kutta condition is satisfied. The Kutta condition is a boundary

62 condition used to determine the overall circulation for an airfoil [36]. The correct circulation at any particular angle of attack corresponds with the flow smoothly leaving the top and bottom surfaces at the trailing edge. The Kutta condition is illustrated in Figure 4-3.

Figure 4-3: Correct circulation over an airfoil corresponds with the flow smoothly leaving the trailing edge; the Kutta condition (right). The circulation of the left is an arbitrary circulation resulting in physically incorrect streamlines. Reproduced from [2].

Panel methods reduce the problem to determining the strength of the fluid elements positioned on the panels so that Laplace’s equation is satisfied and the boundary conditions are met.

4.1.1 Hess-Smith panel method The Hess-Smith method is a popular panel method that uses source elements and vortex elements to satisfy Laplace’s equation [36]. As with all panel methods, the airfoil is divided into panels. A cosine distribution of panels, as shown in Figure 4-4 is commonly used as this generates a higher density of panels at the leading and trailing edges of the airfoil.

Figure 4-4: A cosine distribution is commonly used to divide an airfoil into discrete panels. Reproduced from [31].

63 The Hess-Smith panel method is an inviscid method and was only used to validate the XFOIL software [37]. XFOIL was the main numerical method used to predict pressure distribution as it incorporates viscosity effects. A full listing of equations and description of the Hess-Smith method is included in Appendix A.

Sensitivity of the Hess-Smith panel method to panel size The panel size sensitivity describes the tendency of a numerical solution to approach an asymptotic value as the mesh size or panel size is decreased. Coarse panelling will reduce computation times but may also provide incorrect results. A study was carried out to determine the minimum number of panels the airfoil should be divided into to ensure correct results.

The sensitivity of the Hess-Smith panel method was determined by evaluating the changes in section lift coefficient, drag coefficient and moment coefficient with an increasing number of panels. The coefficients were evaluated at a single angle of attack (10°) with the number of panels increasing from 10 panels to 300 panels in 10 panel intervals.

The lift force and leading edge moment coefficients did not show much variation in value with the increase in panelling. The most telling indicator of panel size sensitivity was the variation in drag coefficient, shown in Figure 4-5. The Hess-Smith method is based on potential flow and the major assumption is that the fluid is inviscid. This means that the only forces that are modelled are pressure forces and shear forces are neglected. It was discovered by D’Alembert in 1744 that the drag predicted by any inviscid solution would be zero due to the fact that shear forces are responsible for drag [3]. As well as neglecting viscous drag, inviscid models are unable to predict pressure drag that results from any flow separation, as the flow separation is due to viscosity.

The variation in drag coefficient (Figure 4-5) showed that if fewer than 50 panels were used, drag forces are present and hence the solution is not a feasible potential flow solution. The drag coefficient did converge to nearly zero above 50 panels, showing that for the Hess-Smith panel method, correct results could be expected provided at least 50 panels were used. For the XFOIL software, the default number of panels used is 160 and similar panel size sensitivity would be expected. The validity of the boundary layer

64 viscous modelling in XFOIL is checked separately by the software as each solution is computed.

Figure 4-5: Drag coefficient convergence of the Hess-Smith panel method.

4.1.2 XFOIL software XFOIL was the main panel method used to provide predictions of the pressure distribution. XFOIL is a widely used software package that uses a panel method based on vortex elements. XFOIL was developed by Professor Mark Drela of MIT.

The advantage of XFOIL is that it includes viscous formulations that are used to calculate the boundary layer and wake flows. The boundary layers and the wakes are described using a two-equation lagged dissipation boundary layer formulation [38] XFOIL also includes compressibility corrections although the corrections were not used in this case due to the fact that the aircraft considered in this study flies at speeds lower than Mach 0.3.

The inclusion of boundary layers in a panel code method is accomplished by first estimating the pressure distribution without any boundary layer modelling. The pressure distribution is then used to calculate the boundary layer development. The surface pressure distribution is then recalculated using the displacement thickness as the new airfoil surface. This method is based on the assumption that the change in pressure through the boundary layer thickness is so small that it can be neglected, and the

65 pressure at the edge of the boundary layer is effectively the surface pressure. This calculation process is repeated until a level of convergence has been reached.

4.1.3 Calculation of force and moment coefficients from pressure distributions The importance of measuring the pressure distribution lies in the ability to calculate the section forces and moments by integrating the measured pressures. This section describes the method that was used to calculate the force and moment coefficients. The surface integration was implemented using Matlab and was used to calculate force and moment coefficients from XFOIL distributions and from distributions measured in flight.

The force due to pressure acting on a surface can be determined using equation 2. The total force was determined by integrating the pressure acting on all surfaces of the airfoil.

dF%%= Pds (2) dC%%FP= C ds

where F%% is the force acting on a surface of length ds due to a pressure P.

The force coefficient, CF is given by

F% C% F = 1 ρVS2 (3) 2

The pressure coefficients at any two adjacent points were used to calculate an average pressure that was assumed to act on a surface between the two points. This is shown in Figure 4-6.

Figure 4-6: The average pressure between two nodes was used when computing the force coefficients.

The two-dimensional normal and axial force coefficients, Cn and Ca were computed in Matlab using numerical approximations to equation 4 and equation 5 respectively.

The moment coefficient about the leading edge (Cm) was calculated using a numerical approximation to equation 6, while the moment coefficient about the quarter chord point

(Cm,c/4) was calculated using a numerical approximation to equation 7. The two- dimensional lift force coefficient (Cl) was calculated using equation 8, using the normal and axial force coefficients, and the local section angle of attack. ~ ~ = dxCC n ∫ P (4) dx airfoilthetorelativeiswhere chord

CCdy%%= aP∫ (5) where dy is relative to the airfoil chord

CCxdxydy%%=−(. + . ) mP∫ (6)

CCxdxydy%%=−(( − 0.25). + . ) mc,/4 ∫ P (7)

CC%%=−cosα C % sinα ln a (8)

4.2 Cessna 182 airfoil measurement and analysis Most of the numerical evaluation was based on a standard NACA2412 airfoil from the NACA 4-digit series. The numbering indicates that the airfoil has a maximum

67 camber height of 2% and the position of maximum camber occurs at a location of 40% chord. The thickness of the airfoil is 12% as indicated by the last two digits.

Early model Cessna aircraft used this exact airfoil section but it was known that most latter model Cessna aircraft (including the C182RG model owned by UNSW@ADFA) used a modified NACA2412 airfoil section [29]. The modifications to the airfoil shape were known to include changes to the leading edge shape but the extent of the modifications is not publicly available.

For this reason, the airfoil section of the Cessna 182RG was measured for comparison with the NACA2412 section. The airfoil section was measured on the inboard constant-chord section of the wing by trimming two mating sheets of corrugated (sign writing) plastic to a close fit of the upper and lower surfaces of the wing. These female profiles were then traced onto paper and the airfoil coordinates were measured from an estimated mean chord line at 5mm chord intervals. The same process was used to measure the flap profile. A comparison between the standard NACA2412 section and the measured wing and flap coordinates is shown in Figure 4-7.

Figure 4-7: Comparison between the measured Cessna 182RG airfoil and the standard NACA2412 section.

A close-up view of the airfoil leading edge is shown in Figure 4-8. The lower leading edge surface of the measured airfoil has a definite bulge with a slightly reflexed surface that blends into the standard NACA2412 airfoil.

Figure 4-8: Comparison between the measured NACA2412 modified airfoil and the standard NACA2412 airfoil, showing the large difference at the leading edge.

It was intended that the measured airfoil geometry would be input into the panel codes to predict the pressure distribution. However the panel codes were found to be very sensitive to surface smoothness, and the measured profile could not be smoothed enough to give reasonable results. In order to predict the effect that the leading edge change may have on the pressure distribution, known modifications to the NACA2412 airfoil were investigated.

The known quantified modifications to the standard NACA2412 airfoil did involve changes to the leading edge radius. A comparison between the leading edge of the measured airfoil section and two modified sections; NACA2412-63 and NACA2412-93 is shown in Figure 4-9. The first digit after the standard 4 digit designation indicates a roundness factor and the second digit specifies a maximum thickness location factor. As Figure 4-9 shows, the lower surface of the measured airfoil matches the NACA2412-93 section back to approximately 2% chord but the upper surface more closely matches the NACA2412-63 section.

Figure 4-9: A comparison between the measured NACA 2412 airfoil and NACA 2412 airfoils leading edge radius modifications.

The effect of the modified nose sections was evaluated using XFOIL. Figure 4-10 shows a comparison between the pressure distributions at 5° angle of attack for the standard NACA2412 airfoil and the NACA2412-63 and NACA2412-93 airfoils. The geometry of the NACA2412-63 airfoil is very similar to the geometry of the standard NACA2412 airfoil and the pressure distributions reflect this similarity. The NACA2412-93 has a much larger nose radius and the effect of this appears to be to increase the peak suction pressures over the upper leading edge surface and to modify the flow over the lower leading edge surface, as can be seen in Figure 4-10.

Figure 4-10: Comparison between pressure distributions for a standard NACA 2412 airfoil and NACA 2412 airfoils with modified nose radius.

From these results, it could be expected that the pressure distribution on the lower leading edge surface of the C182 airfoil would be similar to the NACA2412-93 lower leading edge distribution.

4.3 XFOIL results XFOIL was used to investigate several factors of the pressure belt setup. The factors were the sensor module placement, the effect of not measuring the trailing edge pressure and the effect of pressure belt thickness.

4.3.1 Effects of viscosity A comparison between the lift coefficients predicted using the Hess-Smith method and XFOIL for a NACA2412 airfoil shows the difference between an inviscid solution and solutions incorporating viscous effects (Figure 4-11). The lift curves for the viscous solutions taper as the angle of attack is increased. This is due to the boundary layer growing as the angle of attack is increased.

Figure 4-11: Comparison between inviscid and viscous lift coefficient variation with angle of attack.

A comparison between the pressure distributions for an inviscid and a viscous solution is shown in Figure 4-12 for a NACA2412 airfoil at 5° angle of attack. The general effect of viscosity is to slightly lower pressure coefficients over the upper surface of the airfoil.

Figure 4-12: Pressure distributions predicted using XFOIL showing the effect of viscosity.

72 A comparison between the viscous lift coefficients predicted by XFOIL and published experimental data is shown in Figure 4-13. The published lift coefficient data for the NACA2412 airfoil was taken from the well-known investigations into two- dimensional airfoil flow carried out by Abbott and Von Doenhoff [39].

Figure 4-13: Comparison between experimental lift coefficient values and inviscid values predicted by XFOIL for a NACA 2412 airfoil section. The experimental data is from Abbot and Von Doenhoff [39].

Figure 4-13 shows good agreement between the published data and XFOIL although the published data shows a sharp stall at 16° angle of attack. This is not present in the XFOIL predictions because XFOIL can not model flow separation. It is very difficult to correctly model boundary layer growth, transition and separation points using coupled panel methods and this is the reason for the disagreement at higher angles of attack. Complex correction methods have been developed for panel methods to include some of these effects but are outside the scope of this thesis. Through a range of -10° to +10° angle of attack, agreement between XFOIL and the experimental data is very good.

4.3.2 Effect of estimating trailing edge pressure coefficient The integration of pressure distribution to calculate the lift coefficient is a surface integration and would need to include an estimate of the pressure distribution over the full airfoil chord to avoid gross inaccuracies. Because the pressure at the trailing edge

73 could not be measured in flight, an estimate of the trailing edge pressure coefficient was required.

XFOIL was used to compare the error from three different estimates of the pressure coefficient at the trailing edge. The three methods considered were taking the average of the rearmost upper surface and lower surface sensor pressures; taking a weighted average of these two sensor pressures (2/3 of the lower surface value, 1/3 of the upper surface value) and assuming that the pressure coefficient at the trailing edge is zero for all angles of attack. The last case is analogous to saying that the velocity of the air on the upper and lower surface of the airfoil at the trailing edge has returned to the freestream velocity. Figure 4-14 shows a comparison between pressure distribution at the trailing edge for each of the three trailing edge assumptions, compared to the actual trailing edge pressure computed using XFOIL.

Figure 4-14: Comparison between trailing edge pressure coefficient estimate methods at a single angle of attack.

Table 10 shows the errors in the force coefficient and moment coefficient for each assumption. The errors were calculated as a percentage of the coefficient value at 15 º angle of attack. Table 10 shows the errors in force coefficients (Cl, Cn, Ca) are not large for any of the methods (~1%). However the error in the quarter chord moment coefficient is significant for all three methods. This is because the value of the quarter chord moment coefficient is relatively small and is sensitive to any error. The weighted

74 average method produced the lowest average errors. However the assumption that the pressure coefficient at the trailing edge is zero produces the lowest maximum errors, and this method will be used when calculating the sectional force and moment coefficients from any flight data. The other advantage of this method is that it is computationally simple to implement.

4.3.3 Effect of sensor distribution XFOIL was also used to evaluate the error in measurement that would be created by only measuring the pressure at a number of discrete locations. Three sensor placement methods were evaluated; a linear distribution, where the spacing between the sensors was constant; a cosine distribution, where the chordwise spacing of the sensors varied as the cosine of a fixed angular spacing (giving less spacing toward the leading and trailing edges); and a custom spacing that used two high resolution sensor modules around the leading edge and a cosine spacing of single sensor modules over the rest of the airfoil section. Each distribution method used a quantity of 18 pressure sensors as this was the number of pressure sensors available. A comparison between the three distribution methods is shown in Figure 4-15.

Figure 4-15: A comparison between the three sensor distribution methods that were evaluated.

The sensor distribution methods were evaluated qualitatively by examining the pressure distributions and quantitatively by calculating the error in force and moment

75 coefficients for each method. When calculating the coefficients, it was assumed that the pressure coefficient at the trailing edge was zero.

Figure 4-16 shows the pressure distributions that would be measured by each of the sensor distributions at an angle of attack of 5°. Figure 4-16 (a) shows that the linear sensor placement fails to capture the rapid changes in pressure that occur around the leading edge of the airfoil but the linear spacing does provide good coverage over the rest of the airfoil.

Figure 4-16 (b) shows the pressure distribution that would be expected if a cosine distribution of sensors was used. The cosine distribution captures the rapid spatial changes in pressure that occur around the leading edge; however at this angle of attack the sensor positioning did not capture the peak suction pressure coefficient or the stagnation point.

Figure 4-16 (c) shows the pressure distribution that would be measured using the custom placement of sensors. As expected, the high resolution sensor modules around the leading edge capture the rapid changes in pressure in this region well, but the coarse spacing of sensors over the rest of the airfoil causes some of the pressure changes on the upper airfoil surface to be incorrectly approximated by straight lines, particularly those changes occurring around 0.1% chord.

76 Table 10: Errors in force and moment coefficients due to estimating the trailing edge CP and sensor distribution. All errors are given as a percentage of the coefficient value at 15º angle of attack.

Error

Cl Cm,c/4 Cn Ca

Parameter Method Average Maximum Average Maximum Average Maximum Average Maximum Trailing Average 0.8% 1.2% 19.2% 39.4% 0.8% 1.2% 1.3% 1.6% edge Weighted 0.5% 1.1% 15.6% 36.2% 0.5% 1.1% 0.8% 1.4% pressure average coefficient Zero 0.7% 0.8% 15.6% 27.7% 0.7% 0.8% 1.1% 1.3% Linear 9.3% 14.7% 94.6% 189.8% 9.9% 16.5% 5.0% 12.7% Sensor Cosine 1.0% 1.2% 36.3% 92.6% 0.9% 1.2% 2.6% 4.7% distribution Custom 2.8% 7.3% 107.5% 235.3% 2.9% 7.5% 2.0% 4.1% 1mm 2.1% 4.5% 33.5% 71.5% 2.1% 4.6% 1.1% 4.1% Pressure belt 3mm 2.2% 4.4% 42.7% 89.6% 2.2% 4.3% 1.5% 5.8% thickness 5mm 2.4% 5.1% 63.7% 116.4% 2.4% 4.9% 1.9% 7.8%

(a)

(b)

78 The distribution methods were also evaluated quantitatively by examining the error in the calculated force and quarter chord moment coefficients. Figure 4-17 shows the lift coefficient variation with angle of attack for each of the sensor placement methods compared to the full pressure distribution. This graph, along with the tabulated errors in force and moment coefficients in Table 10 show that the linear placement of sensors would be unsatisfactory and would lead to large errors due to the pressure changes around the leading edge not being fully captured.

Figure 4-17: Comparison between the calculated lift coefficient using the three sensor distribution methods.

The best distribution method for minimising the error in calculated lift coefficient would be the cosine distribution method, giving an average error of 1.0%, showing why it is a good method for panelling an airfoil for numerical calculations. Although the average error in lift force coefficient for the custom sensor placement is 2.8%, it was decided that this method would be used as the high spatial resolution sensor modules would be able to capture the rapid changes in pressure distribution around the leading edge which would be of particular interest when comparing flight testing results to predicted pressure distributions. The ideal solution would have been to remove some sensors from the high resolution modules for use as single sensor modules but this was unfeasible as once the sensors had been mounted to the circuit board they could not be removed without damaging the sensors and the circuit boards.

79 Table 10 shows that the errors in the quarter chord moment coefficient are very large for all three sensor placement methods. This is due to the pressure coefficient variations toward the trailing edge being incorrectly approximated and this creates a relatively large error because the absolute value of the moment coefficient is small to begin with.

4.3.4 Effect of pressure belt thickness The final aspect that was investigated using XFOIL was the effect of the thickness of a pressure belt. The effect of constant thickness tubed pressure belts had been evaluated in flight by NASA [9]. The study found that there was no significant change caused by the addition of pressure belts up to thicknesses of 5mm. XFOIL was used to confirm that the changes due to addition of a constant thickness pressure belt were not significant. The addition of a pressure belt was modelled by inflating the airfoil panelling by the thickness of a pressure belt. Three thicknesses were evaluated; 1mm, 3mm, and 5mm. The actual thickness of the sensor modules when attached to the wing was just over 3mm.

Figure 4-18 shows a comparison between the pressure distributions over the upper surface of the leading edge for the 3 pressure belt thicknesses compared to the original pressure distribution. Figure 4-18 shows that the predicted effect of a pressure belt is to lower the peak suction pressure coefficients slightly. The effect is not significant and the remainder of the pressure distribution is very similar to the original distribution.

The errors in force and moment coefficient in Table 10 show that the theoretical effect of a pressure belt is small but significant. The variation between the three thicknesses is not large, but all thicknesses create an error of approximately 2% in the lift force, normal force and axial force coefficients.

Figure 4-18: The main effect of the addition of a pressure belt is a reduction in the peak leading edge suction pressure. 4.4 Lifting line theory Classic lifting line theory was used to develop an estimate of the local angle of attack for the point on the aircraft wing where the pressure belt was attached. The local angle of attack differs from the aircraft angle of attack due to the spanwise distribution of lift that a three-dimensional wing generates. This section describes the use of lifting line theory to develop a correction that would allow the measured pressure distributions to be compared with the theoretical distributions.

Lifting line theory was used to provide the link between the two-dimensional panel methods and the three-dimensional aircraft wing. The flow over a two-dimensional airfoil section can be considered to be part of a wing of infinite span. A wing of finite span sheds vortices from the wing tips and these shed vortices create spanwise variations in wing loading and local angle of attack [2]. Figure 4-19 shows a typical spanwise variation in lift across a finite wing, along with the downwash created by the shed vortices which causes the local section angle of attack to vary.

Figure 4-19: The spanwise loading of a finite aircraft wing. The case shown is for a wing with elliptical loading that experiences constant downwash Reproduced from [2].

Lifting line theory is a method for relating two-dimensional airfoil theory to three- dimensional wing behaviour. Lifting line theory was developed by the aerodynamicist Prandtl between 1911 and 1918 [3]. The spanwise loading effects are created by assuming that the wing contains infinite horseshoe vortices that form a trailing vortex sheet. The strength of these vortices is varied, creating the spanwise lift variation. The vortices create a varying downwash from the wing that in turn causes the local section angle of attack to vary along the span of the wing. This means that at any given spanwise wing location the section angle of attack will be modified by an induced angle due to the downwash created by the trailing vortices, as shown in Figure 4-20.

Figure 4-20: The effect of downwash from a finite wing is to reduce the effective angle of attack. Reproduced from [32].

The inputs to the lifting line theory were the wing geometry and the two- dimensional airfoil characteristics. The wing characteristics consisted of the chord dimensions and the variation of chord with span, and the wing twist parameters including the root incidence angle and the variation of wing twist with span. The wing twist data was obtained from Janes All the Worlds Aircraft [29] and is geometric twist that is built into the wing to create a desired spanwise loading. The effects of aero- elasticity were not considered and the aerodynamic effects of the fuselage and wing

82 strut presence were assumed to be negligable. Figure 4-21 shows the wing planform and twist variation that was used to model the Cessna 182RG wing.

Figure 4-21: Wing plan form (left) and wing twist (right) used in lifting line calculations.

Figure 4-22 shows the distribution of local angle of attack computed for four aircraft angles of attack. The spanwise distribution was used to compute the local angle of attack for the location where the pressure sensor modules were going to be mounted. For this work, the pressure belt was mounted at a point 1.74m from the aircraft centreline.

Figure 4-22: Spanwise distribution of local section angle of attack for aircraft angles of attack of 0°, 5°, 10° and 15°. The aircraft angle of attack is taken relative to the fuselage datum line.

83 Figure 4-23 shows the relationship between the local section angle of attack for the wing station where the pressure sensor modules were to be mounted and aircraft angle of attack. The aircraft angle of attack is generally higher than the local section angle of attack. The equation that was used to compute the local section angle of attack from the measured aircraft angle of attack is shown in equation 9.

α = α + 47.1824.0 AC (9)

Figure 4-23: Relationship between the aircraft angle of attack and the section angle of attack for the section where the pressure belt is attached 4.5 Conclusions This chapter has presented a summary of the numerical methods that were used to predict the pressure distributions over a NACA2412 airfoil. The pressure distributions were used to evaluate the errors that could be expected due to sensor module placement, the thickness of the sensor modules and the errors in force and moment coefficients due to estimating the trailing edge pressure coefficient.

The predicted effect of estimating the trailing edge pressure coefficient on the force coefficients was small. However the estimate methods did create a significant error in the quarter chord moment coefficient. Because the difference between the estimate methods was small, it was decided that to assume that the pressure coefficient at the trailing edge was zero as this was easy to implement computationally.

84 It was found that a cosine patterning of pressure sensor modules would produce the smallest errors in force and moment coefficients. However it was decided that a custom placement using high resolution sensor modules around the upper and lower leading edge surfaces would be used. The average error in lift force coefficient due to this patterning method was 2.8%.

It was found that the addition of a constant thickness pressure belt caused average errors in the force coefficients of approximately 2%. However there was no significant difference in the errors for three thicknesses up to a thickness of 5mm.

The next chapter describes the flight testing that was carried out and the results from the flight testing are presented.

85 5 Flight Investigations

This chapter describes the flight testing that was carried out with the pressure belt system using the UNSW@ADFA Cessna 182RG aircraft. The chapter begins with a description of the test flights that were carried out. The initial flights showed that the sensor fairing was causing erroneous measurements and the solution to this problem is discussed. The main focus of the test flights was the measurement of steady state pressure distributions. Results of these tests are presented and this is followed by a discussion of the steady state force and moment coefficients that were calculated from the measured pressure distributions. The measurement of flow separation is then discussed. The final section discusses the measurements made during a dynamic flight manoeuvre. These measurements showed significant differences to the measurements made during steady state flight.

5.1 Description of the test flights Five test flights were carried out with the pressure belt system. The main focus of the flight testing was to make measurements of the pressure distribution at various angles of attack. The first three flights were used to check the operation of the pressure belt system, with the final two flights being used to gather data from specific flight manoeuvres. A summary of the flights is contained in Table 11.

The first three test flights focused on testing and refining the operation of the pressure belt system. After the inter-sensor fairings had been installed and tested (see section 5.2), three further flights were made to make measurements of the pressure distributions during various flight conditions.

86 Table 11: Description of the test flights Flight description Flight time Tests carried out

Preliminary flight, No measurements made, the attachment 70 minutes no data recorded of the sensors was checked. Measurements made during cruise flight 1st test flight, un- 30 minutes and a slow increase in angle of attack to faired attachment stall 2nd test flight, faired 60 minutes Test of the inter-sensor fairing attachment - Steady flight at cruise angles of attack - Slow increase in angle of attack up to stall, 0° flap, undercarriage raised - Slow increase in angle of attack up to 3rd test flight, faired 80 minutes stall, 10° flap, undercarriage raised attachment - Steady flight at various engine power settings, undercarriage raised - Short period pitch oscillations, undercarriage raised - Slow increase in angle of attack up to stall, 20° flap, undercarriage raised - Slow increase in angle of attack up to 4th test flight 60 minutes stall, 30° flap, undercarriage raised - Short period pitch oscillations, undercarriage raised

The purpose of the slow increase in angle of attack manoeuvre up to the stall point, (flights 3 and 4) was to gather data from the full range of angles of attack. The manoeuvre was repeated for flap deflections of 10°, 20°, and 30° to investigate the influence of the flaps on the measured pressure distribution. The flap results will not be presented because the data gathered was not conclusive due to pressure sensor modules not being installed on the flap surfaces. The purpose of the short period pitch manoeuvre (flights 3 and 4) was to gather data using the pressure belt system during a dynamic manoeuvre.

The data gathering test flights were carried out in an area outside of the Canberra Airport airspace, on a track between Canberra Airport and the Rugby NDB, as shown in Figure 5-1. This area was used as it enabled flight at altitudes up to 9500 feet while remaining outside of the Canberra Airport controlled airspace.

Figure 5-1: Map showing flight test area 5.2 Initial attachment of sensor modules The sensor modules were initially attached to the wing by taping directly over them with strips of aluminium tape that had small holes cut into it to expose the pressure sensor port. The tape was used to fair the sensors to the wing, as shown diagrammatically in Figure 5-2.

Figure 5-2: Diagram showing the initial attachment and fairing to the wing surface using aluminium tape.

5.2.1 Results The results from the first test flight showed significant differences when compared with the predicted pressure distributions from XFOIL. A typical pressure distribution is shown in Figure 5-3, along with the pressure distribution predicted by XFOIL for that

88 section angle of attack. The difference in the pressure distribution was thought to be due to the tape fairing creating a local flow effect. The fairing between the sensors around the leading edge was of near constant thickness because the sensors were close to each other but further back, discrete fairing of each sensor was used. It was thought that the constant fairing caused the leading edge pressure coefficients to correspond well with the predicted coefficient, while the discrete fairing is thought to cause the pressure coefficients measured by the single sensor modules to differ so greatly from the predicted coefficients.

Figure 5-3: A typical pressure distribution measured during the first test flight, showing the effect that the tape fairing was having on the results.

5.2.2 Discussion To investigate the effect of the tape fairing directly to the surface of the wing, the Hess-Smith panel method was used to model the fairing. The effect of the tape fairing was estimated by calculating the pressure distribution over modified wedge shapes of different half-length to height ratios at 0° angle of attack. At 0° angle of attack, the flow over a wedge is symmetrical about the chord line and can be used to approximate the flow over a faired sensor on a flat surface as shown in Figure 5-4.

Figure 5-4: A wedge shape was used to simulate the flow over a faired sensor to determine the error created by the fairing.

Three ratios of wedge half-length to wedge height were evaluated; 1:5, 1:10 and 1:20 which for a sensor height of 3mm represent fairing lengths of 15mm, 30mm and 60mm respectively. For each shape, the pressure coefficient at the middle of the wedge (where the pressure port would be located for a faired sensor) was estimated using the Hess-Smith panel code. This value of pressure coefficient was used to find a ratio of upstream velocity to the velocity at the point of the sensor. This velocity ratio was then used to evaluate the change in pressure coefficient on the airfoil that could be expected due to the fairing. The basic effect of the fairing is to accelerate the flow velocity to a higher value as it passes over the sensor.

The predicted pressure coefficients at the sensors are more negative than the original surface pressure coefficients because the fairing is accelerating the local airflow. This effect is shown in Figure 5-5 for a NACA2412 airfoil at an angle of attack of 6°, and is consistent with the pressure distributions measured during the first test flight.

Figure 5-5: The effect of individually faired sensors on the pressure distribution for different fairing ratios at 6° angle of attack.

The lower fairing ratios of 1:10 and 1:5 create significant errors in the pressure coefficients measured by faired pressure sensors, which were not acceptable and a solution was needed to ensure accurate measurements could be made.

In the previous chapter, the effect of pressure belts of various thicknesses was evaluated by numerically predicting the pressure distributions for airfoils that had been inflated in thickness. This study showed that the effect of adding a constant thickness pressure belt to a wing was negligible. The problem with the initial sensor attachment was that the fairing created an undulating surface that caused the surface airflow to accelerate and decelerate. Clearly the solution to the problem was to provide a filling between the sensor modules that created a constant thickness addition to the wing surface. This is shown diagrammatically in Figure 5-6.

Figure 5-6: Diagram showing the use of close cell foam to fill between the pressure sensors and create a constant thickness.

91 Closed cell foam of 3mm thickness was used to achieve the inter-sensor fairing. The foam was cut to a width of 90mm and 8mm circular openings for the pressure sensors were made in the foam using a punch. The sensor modules were taped to the wing surface using single strips of aluminium tape. The strips of foam were then laid down over the sensor modules and trimmed to fit. Aluminium tape was laid over the foam and small diameter openings were made using a punch to expose the pressure sensor ports to the surface airflow. The foam was taped down to the wing surface using aluminium tape, with the result shown in Figure 5-7.

Figure 5-7: The initial method of fairing the sensor modules (left) was unsatisfactory. The final attachment of the sensor modules to the wing (right) included foam that was used to provide a constant thickness fairing in between the sensor modules. In both cases the leading edge of the wing is to the bottom of the picture.

The third test flight was used to evaluate the effect of the inter-sensor fairing. The foam filling was found to work well. A comparison between pressure distributions measured with and with-out the inter sensor fairing shown in Figure 5-8. It can be seen that the measured pressure distribution with the inter-sensor foam fairing is a closer approximation to the pressure distribution predicted by the Hess-Smith panel method1 than the pressure distribution measured without the inter-sensor fairing.

1 The Hess-Smith method was used instead of XFOIL as it was easily adapted to model this flow situation

Figure 5-8: Comparison between pressure distributions measured with and without the inter- sensor foam fairing at a single angle of attack. Also shown is the predicted pressure distribution for the particular angle of attack. 5.3 Steady state results The steady state pressure distributions and force and moment coefficients were measured during the third test flight. The manoeuvre that was used to gather the data was a steady increase in angle of attack from cruise flight to stall. The Reynolds number during the manoeuvre ranged from 2.9x106 to 7.2x106. It was shown in section 4.3.1 that the effect of Reynolds number variation should not be significant.

5.3.1 Pressure distributions A comparison between the measured pressure distribution and a pressure distribution predicted using XFOIL is shown in Figure 5-9 for a low angle of attack of 2.4°. As the comparison shows, the overall agreement between the measured pressure coefficients and the predicted pressure coefficients is quite good, although the shape of the measured leading edge suction peak is slightly different to the predicted shape. This difference is thought to be due to the increase in the leading edge radius with the faired pressure sensors installed. Also included on the plot are the error bands for the measured pressure coefficients. These error bands are the uncertainty in the pressure sensor readings, as calculated in section 3.11.

Figure 5-9: Pressure distribution measured during steady state flight at an angle of attack of 2.4°. The measured distribution has been compared with a distribution predicted using XFOIL for the same section angle of attack.

As Figure 5-9 shows, for this angle of attack the stagnation point on the airfoil (where the pressure coefficient is one) is well captured due to the use of a high resolution sensor module on the lower surface of the leading edge. The high resolution sensor module on the upper surface at the leading edge also captures the peak suction pressures that occur in this region.

A second comparison between the measured pressure distribution and predicted pressure distribution is shown in Figure 5-10 for a section angle of attack of 5°. Again good agreement was found between the measured and predicted pressure distributions, with the leading edge suction peak and stagnation points being well captured. Good agreement can be seen between the measured suction peak around the leading edge and the suction peak predicted using XFOIL for a Reynolds number of 5.7x106.

Figure 5-10: Pressure distribution measured during steady state flight at an angle of attack of 5°. The measured distribution has been compared with a distribution predicted using XFOIL for the same section angle of attack.

Finally, a comparison between the measured pressure distribution and the predicted pressure distribution for a high section angle of attack of 12° is shown in Figure 5-11. Again the suction peak around the leading edge upper surface is well captured, with the peak suction value being lower at CP=-6 than the peak suction value predicted using

XFOIL of CP=-7.

Figure 5-11: Pressure distribution measured during steady state flight at an angle of attack of 12°. The measured distribution has been compared with a distribution predicted using XFOIL for the same section angle of attack.

5.3.2 Steady state force coefficients The measured pressure distributions from steady state flight were used to calculate section force and moment coefficients. These were compared with force and moment coefficients predicted using XFOIL. The force and moment coefficients were calculated by integrating the pressure distributions, as described in section 4.1.3.

The manoeuvre consisted of a slow increase in the angle of attack of the aircraft from cruise angles of attack (~0°) up to the point of stall (~14°). The manoeuvre can be considered to be quasi-steady state as the rate at which the angle of attack was increased was typically between 0.1° per second and 0.2° per second with a peak rate of 0.5° per second. The section angle of attack was calculated from the measured aircraft angle of attack using the correction developed from the lifting line theory. The data gathered during the stall will be presented separately in Section 5.4.1 as it is not considered to be a steady state effect.

Figure 5-12 shows a time history of the measured section angle of attack and calculated section lift coefficient during the manoeuvre that was used to gather the steady state pressure distributions and steady state force and moment coefficients.

Figure 5-12: Time history showing the section angle of attack (top) and the calculated section lift coefficient (bottom) during a slow, quasi-steady increase in angle of attack up to stall.

The normal force coefficient calculated from the measured pressure distributions versus section angle of attack is shown in Figure 5-13, along with the normal force coefficients predicted with XFOIL using a Reynolds number of 5.7x106. The agreement between the measured coefficients in flight and the predicted coefficients will be discussed in the next section.

Figure 5-13: Comparison between the calculated section normal force coefficient and the predicted normal force coefficient using XFOIL.

The measured pressure distributions were also used to calculate section lift force coefficients. Unlike the calculation of the normal force coefficient, which is a direct integration of the pressure distribution (due to the orientation of the force relative to the airfoil), the calculation of the lift coefficient required the local angle of attack to be known. Section 4.4 described the use of lifting line theory to develop a relationship between the aircraft angle of attack (measured with the alpha vane on the air data boom) and the local airfoil section angle of attack. The lift force coefficients calculated from flight are shown in Figure 5-14.

Figure 5-14: Comparison between the lift force coefficients calculated from flight data and numerical predictions using XFOIL. 5.4 Discussion of steady state force and moment coefficients There is generally good agreement between the calculated normal force coefficient and the predicted values. As the angle of attack increases, the growth of the boundary layer causes the normal force coefficient curve to flatten, which is shown in the XFOIL predictions and the measured flight data.

The measured normal force coefficient is smaller than the values predicted by XFOIL below 3° angle of attack and larger above 3°. From the measured pressure distributions it was noted that in general the measured leading edge suction peak was lower than predicted peak which prompts the question why is the normal force coefficient higher above 3° if the peak pressures measured were lower?

The answer to this question is that the cause is the distribution of the sensor modules. The limited number of sensor modules meant that the leading edge area of the airfoil section had good coverage with the high resolution modules, but only two single sensor modules were used on the upper and lower airfoil surface back to a normalised chord length of 0.5. Straight line interpolation was used between the measured sensor points and this interpolation, along with the assumption that the pressure coefficient at the trailing edge was zero causes the area enclosed by the pressure distribution curve to

99 be higher than it would have been if more sensors were used. The error caused by the sensor placement and the trailing edge pressure coefficient estimate was estimated in section 4.3.

In a similar manner to the normal force coefficient curve, the lift force coefficients are slightly higher than the predicted lift force coefficients above 3° angle of attack. Again this is thought to be due to a lack of sensor modules located toward the rear of the airfoil section, and also the estimate of the trailing edge pressure coefficient. It was shown in the previous chapter when analysing the effect of the sensor distributions that the custom distribution of sensors used would cause the calculated lift coefficient to be higher than the actual lift coefficient at higher angles of attack, which is what was observed in flight.

Another contribution to the discrepancies between the flight data and the experimental wind tunnel data is the lifting line theory. The main drawback of the basic lifting line theory is that it is an inviscid theory and is heavily dependent on wing geometry. Lifting line theory assumes that the lift curve slope for the airfoil sections that make up the wing is linear and infinite, which in reality is not the case. This means that the correction for section angle of attack obtained using lifting line theory is a good approximation at relatively low angles of attack (<10°) where the slope of the lift curve is constant, but large errors may be present in the estimated section angle of attack above 10° due to the non-linearity in the airfoil lift curve slope not being taken into account.

5.4.1 Flow separation In addition to the quasi-steady state manoeuvres that were used to build up the force coefficient data, airfoil flow separation was also investigated. A time history of the angle of attack and section lift coefficient during the stall occurrence is shown in Figure 5-15.

100

Figure 5-15: Time history of section angle of attack (top) and section lift coefficient showing flow separation occurring and the aircraft stall point.

The time history of the angle of attack shows that the section angle of attack increased to 14° and was held at this angle for approximately 10 seconds before the angle of attack dropped rapidly down to 5° and then recovered to 12° before being reduced back to cruise angles of attack lower than 5°. This sharp drop in angle of attack marks the point at which the aircraft stalled and control was lost. The stall recovery was achieved when the angle of attack recovered to 12° followed by the pilot continuing the recovery back to a cruise condition.

The bottom graph in Figure 5-15 shows the time history of the calculated section lift force coefficient. It can be seen that while the section angle of attack was being held at 14°, the section was experiencing the effects of flow separation indicated by the lift coefficient experiencing large fluctuations. Complete flow separation can be seen to occur at an elapsed time of 269 seconds when the lift coefficient dropped down to a value of approximately one. It should be noted that the aircraft stalled at an elapsed time of approximately 271 seconds, while the initial stages of stall had been felt as early as 260 seconds elapsed time. This was due to the spanwise separation pattern that most aircraft experience. As was shown in the previous chapter, the spanwise distribution of lift and angle of attack means that the wing sections closest to the wing root experience the highest lift coefficients and local angles of attack for any given aircraft angle of

101 attack. The progression of flow separation on a typical wing therefore starts at the root and progresses toward the tip. This is an inherent safety feature as it allows the outboard wing sections where the ailerons are located to stall last, allowing roll control to be maintained during a normal stall.

The trends displayed by the measured lift force coefficients compare well with those predicted using XFOIL, with the slope of the lift force coefficient graph decreasing as the angle of attack is increased up to the stall angle, in this case 14° at which point the flow separates and the lift force coefficient values drop significantly. However the uncertainty in section angle of attack due to the inviscid lifting line theory is unclear. The study carried out comparing pressure distribution measurements from flight with wind tunnel data for a Fokker 100 aircraft noted the difficulty in using lifting line theory to calculate a local section angle of attack [23]. In that case, lifting line theory was only used to correct the results from flight for production tolerances and wing deformations. Because the comparison was being made between flight results and wind tunnel results, the aircraft angle of attack was used when calculating the local lift coefficient. This means that the absolute value of the calculated section lift coefficients would be wrong, but the wing section where the pressure distribution had been measured was the same for the flight tests and the wind tunnel tests, and the results from both cases could be compared directly against aircraft angle of attack.

As well as the calculated lift coefficient changes, the measured pressure distributions also showed the point of flow separation well. Figure 5-16 shows the progression of flow separation as seen in the measured pressure distributions. The point of complete flow separation is marked by a distinct flattening of the pressure over the upper surface of the airfoil section, consistent with flow separation observed previously with pressure belts [10].

102

Figure 5-16: Flow separation progression as seen in the measured pressure distributions. Top left shows the flow to be attached prior to stall, while top right shows intermittent flow separation effects prior to complete flow separation. Bottom left shows the flow to be separated with bottom right showing the flow reattachment after recovery from the wing stall.

103 A comparison between the lift force coefficient measured in flight and experimental lift force coefficients for a two-dimensional NACA2412 airfoil is shown in Figure 5-17. The experimental data was taken from Abbot and Von Doenhoff [39]. Two sets of experimental wind tunnel data are shown; lift coefficients measured using a smooth airfoil section and lift coefficients measured using an airfoil section with standard roughness. The experimental data was measured using a two-dimensional wing section in a wind tunnel. As Figure 5-17 shows, the effect of roughness is to lower the lift force generated at higher angles of attack and to cause flow separation to occur at a lower angle.

Figure 5-17: Comparison between the lift force coefficients calculated from flight data and lift force coefficients measured for a two-dimensional NACA2412 airfoil in a wind tunnel.

The section angle of attack at which flow separation was recorded during flight was approximately 14°, which does not agree with the wind tunnel testing carried out by Abbot and Von Doenhoff for a smooth airfoil section which suggests that the NACA2412 airfoil should stall at a higher angle of attack of approximately 17°. The data from the airfoil with simulated roughness (Figure 5-17) indicates that any roughness should cause the stall angle of attack to be lowered, which is the case in the flight data and the stall angle for the airfoil with simulated roughness matches the stall angle measured in flight well. The reason for the retardation in the stall angle is most likely that a different stall mode was experienced by the wing section in flight,

104 compared to a smooth airfoil. Figure 5-18 shows the effect that different stall modes have on the angle at which stall occurs. The plot was taken from Hoerner [40] in which the three stall modes; trailing edge stall; leading edge stall; and leading edge separation are summarised. Trailing edge stall was most likely experienced by the smooth airfoil tested by Abbot and Von Doenhoff as this stall mode is experienced at angles of attack above 15°. The stall mode is characterised by a steady progression of separation from the trailing edge forward.

Figure 5-18: Lift force coefficient variation with angle of attack for different airfoils, showing the effect of different stall modes on the angle at which stall occurs. Reproduced from [40].

The stall mode that was experienced by the airfoil with standard roughness tested by Abbot and Von Doenhoff, and by the airfoil section during flight was most likely leading edge stall. The stall angle for this stall mode is lower, closer to 13° as shown in Figure 5-18. The stall mode is characterised by the flow suddenly separating with the leading edge marking the point of separation and is common on airfoils with round nose sections and low camber [40]. This sudden separation was observed in the measured pressure distributions and can be seen in the lower left graph contained in Figure 5-16.

105 Rather than observing flow separation progressing from the trailing edge forward as had been observed during the NASA HARV program [10] (which was shown in the measured pressure distributions by a flattening of the pressure distribution over the upper surface toward the trailing edge) the measured pressure distributions showed a sudden drop and flattening of the pressure coefficients over the entire upper surface of the airfoil section.

5.5 Short period pitch oscillations One manoeuvre that was carried out during the first data gathering flight was a short period pitch oscillation. This manoeuvre is normally carried out during the flight laboratory flights to demonstrate the longitudinal stability and damping characteristics of the aircraft. The manoeuvre was carried out during the data gathering flights as a test of the measurement system under dynamic conditions and provided an opportunity to use the system in a situation where a traditional tubed pressure belt could not be used due to pneumatic lag.

5.5.1 Results A time history of the section angle of attack during a series of manoeuvres is shown in Figure 5-19. The manoeuvre consisted of the pilot providing an elevator doublet control input; pitching the aircraft nose down and then quickly pitching the aircraft nose up. The pilot then released the controls to observe the stability and damping of the aircraft motion.

106

Figure 5-19: Time history of the section angle of attack showing a sequence of short period pitch manoeuvres.

A close up of a single short period oscillation is shown in Figure 5-20. It shows the angle of attack change in a sinusoidal manner, with very little motion continuing after the control inputs and the aircraft returning rapidly to the trim point at which the manoeuvre was begun.

Figure 5-20: Time history of a single short period oscillation showing the range of section angle of attack and the highly damped nature of the motion.

107 When the data was being analysed, it was noted that the force coefficients exhibited significant hysteresis, as shown by the plot of lift force coefficient shown in Figure 5-21.

Figure 5-21: Lift force coefficient variation during short period pitch manoeuvre, showing the hysteresis loop. The trim dynamic pressure was approximately 1860Pa and the static pressure was approximately 80000Pa. The aircraft weight was approximately 1040kg.

5.5.2 Discussion During the short period manoeuvre, the aircraft translated vertically (plunging) and rotated about the centre of gravity (pitching). The effect of both motions is to modify the oncoming flow. The vertical translation would cause a vertical velocity component to be added to the uniform oncoming flow. This combination would result in a uniform oncoming flow that would be at a modified angle to the original flow. For an airfoil plunging downwards, the airfoil would experience an increase in the angle of attack while the upwards motion would decrease the angle of attack. Although the vertical translation would modify the angle of attack, the oncoming flow would still be uniform and should not cause hysteresis.

The effect of the pitch rotation is to create an upstream flowfield that is oscillatory and hence unsteady. This will cause the flow over the airfoil to be unsteady. These phenomena are the likely cause of the hysteresis. The dynamic response of the angle of

108 attack vane itself was assumed to be much quicker than the frequency of the induced motion and therefore unlikely to be a source of error.

The rotational motion of the aircraft would affect the measured angle of attack. The effect of aircraft rotation on the angle of attack vane is shown in Figure 5-22. Possible error in the angle of attack measurements was estimated using the recorded aircraft pitch angles. The pitch angle data was used to estimate the pitch rate of the aircraft by differentiating it with respect to time. From the pitch rate data, it was noted that the maximum pitch rate experienced was approximately 20° per second and this was only achieved momentarily during the manoeuvre. The pitch rate was used to estimate the vertical velocity induced at the position of the angle of attack vane according to equation 10.

Figure 5-22: The rotational motion of the aircraft during the short period pitch manoeuvre caused the angle of attack vane to misread.

The velocity todue the theofrotation aircraft, v is bygiven ω ×= rv the angular theoferror vane is bygiven (10) v sinθ = U ⎛ v ⎞ θ = sin −1 ⎜ ⎟ ⎝U ⎠

Assuming that the centre of gravity was located near the quarter chord point, the distance from the centre of rotation to the angle of attack vane was approximately 0.7m. Assuming a worst case pitch rate of 30° per second, a vertical velocity of 0.35m/s would be induced at the position of the angle of attack vane. At an aircraft velocity of 60m/s, this would cause an angular error of approximately 0.33° due to the rotation. In principle it would be possible to correct the angle of attack measurements for this effect. The correction was not applied to the measurements reported here due to the noisiness of the pitch rate signal, although the value of the peak pitch rates could be discerned. If

109 the error in the angle of attack due to the pitch rates was to be corrected for, a pitch rate gyro would provide a better record of the pitch rates. In this case the pitch rates were not significant so the correction to the angle of attack measurements was neglected.

Lift coefficient data from a second series of short period pitch oscillations is shown in Figure 5-23. The difference between this data and the previous series of oscillations (Figure 5-21) is that this series was carried out with the aircraft landing gear lowered and both the trim angle of attack and the trim lift coefficient were higher. Overall the pattern looks similar to the hysteresis recorded during the first test (Figure 5-21) and given that the data shown in both plots are the result of a sequence of oscillation motions, the hysteresis appears to be very repeatable. The effects of different test conditions such as aircraft weight, centre of gravity position and possible inertial effects due to the landing gear position were not considered when these tests were made.

Figure 5-23: Lift force coefficient variation during short period pitch manoeuvre. The trim dynamic pressure was approximately 1900Pa and the static pressure was approximately 80000Pa. The aircraft weight was approximately 1040kg and the undercarriage was lowered.

A search of available literature found work that had been carried out at the Northwestern Polytechnical University in China [41], looking at the hysteresis properties of oscillating wings. The study used Euler equations to numerically simulate the unsteady flow over an airfoil while it was oscillating in pitch about different axis positions, with different frequencies and amplitudes. This study found that hysteresis

110 was present in the force coefficients due to the motion of the airfoil. An example of the hysteresis found in the lift force coefficient is shown in Figure 5-24.

Figure 5-24: Unsteady lift force coefficient variation for a NACA0012 airfoil during pitching oscillation about the axis at x/c =0.25, with M∞=0.755 and α0=2.5°. Reproduced from [41].

The numerical investigation into the hysteresis due to pitching oscillation found that the lift coefficient during the nose pitch up motion was less than the lift coefficient during the nose down pitch motion. The author concluded that the hysteresis was due to the unsteady motion of the wing creating a relative airflow different to the angle of attack. This is thought to be the cause of the hysteresis that was measured during the flight testing.

The study found that hysteresis was dependent on the chord position about which the airfoil oscillated and the frequency of the oscillation. A non-dimensional (reduced) frequency was used to characterise the oscillation frequency. The reduced frequency was computed by multiplying the actual frequency by the chord length and dividing the result by the freestream air velocity. For the short period pitch oscillations that were carried during the flight testing, the frequency of the oscillations was consistent at approximately 0.7 Hz. This corresponds to a reduced frequency of 0.714 for a freestream velocity of approximately 60 m/s. With the pitching motion occurring about the approximate quarter chord position, the pitching hysteresis shown in Figure 5-24 most closely matches the conditions during flight.

111 Figure 5-25 shows three measured pressure distributions showing the effect of pitch rate; positive pitch rate in red, no pitch rate in black and a negative pitch rate in blue. It can be seen that the leading edge suction peak for the positive pitch rate distribution is much lower than for no pitch rate. This shows the effect of the non- uniform flowfield causing the airfoil to ‘see’ a lower local apparent angle of attack during the unsteady pitching manoeuvre.

Figure 5-25: Measured pressure distributions showing the effect of nose up pitching (red) and nose down pitching (blue) compared to steady state flight (black).

The opposite effect can be seen from the nose-down pitch distribution. The blue distribution shows that the pressure coefficients around the upper leading edge surface are slightly higher compared to the steady-state distribution due to the downward rotation of the airfoil. This indicates that the apparent angle of attack that the airfoil is experiencing during this motion is higher due to the rotation.

5.6 Conclusions The flight testing of the pressure belt system has been described, along with analysis of the data that was recorded from the test flights. The main focus of the test flights was to measure the pressure distributions for comparison with predicted pressure

112 distributions. There was good agreement found between the measured pressure distributions and the predicted pressure distributions. Integration of the measured pressure distributions to calculate section force and moment coefficients highlighted the need for more sensor modules distributed over the airfoil section in order to avoid errors due to interpolation between the measured pressure points. The pressure distributions measured during airfoil stall showed that leading edge separation was probably the stall mode that the wing was undergoing. Short period pitch oscillations were carried out in flight and the lift force coefficients calculated from the pressure distribution measurements showed hysteresis, thought to be due to the airfoil experiencing a non- uniform onset flow due to the aircraft rotation. The next chapter will present conclusions from the thesis project and recommendations for future work.

113 6 Conclusions and Recommendations

6.1 Introduction This focus of this thesis was the development of an advanced pressure distribution measurement system for use on a light aircraft. One of the major requirements of the system was that it was affordable to procure and develop. This chapter summarises the conclusions of the work and presents recommendations for future research work that could be carried out using the pressure distribution measurement system.

6.2 Conclusions A system for measuring the air pressure at numerous points over an aircraft wing in flight has been developed. The system was developed to provide an aerodynamic research capability to the UNSW@ADFA aircraft that is currently used for undergraduate flight laboratories.

Of the methods that currently exist for measuring pressure distribution in flight, the pressure belt method is considered to be versatile and easy to install but suffers from pneumatic and acoustic lag due to the tubing that is used to pipe surface pressures to off-wing pressure sensors. The advanced system developed by Boeing provides a novel solution to the problem by using pressure sensors mounted directly on the wing. The main disadvantage of the Boeing system is the extremely high cost.

Sensor modules were developed consisting of small pressure sensors interfaced digitally with microcontrollers. The microcontrollers were used to facilitate the transmission of the data over a digital data bus to eliminate errors due to electromagnetic noise. The pressure sensors, microcontrollers and associated electronic components were mounted on flexible circuit board material. The modular packaging of the sensors was done to enable versatility. Two forms of sensor module were developed; a single pressure sensor module and a high resolution sensor module containing seven pressure sensors with a spacing of 15 millimetres between sensors. An RS485 network protocol was used for the digital transmission of data. This particular protocol was used due to its inherent tolerance to electromagnetic noise. The total cost

114 (including parts and labour) of manufacturing a single sensor module was $87.40 and the cost for a high resolution module was $260.70. This was acceptable within the scope of the project which was to develop an affordable pressure distribution measurement system and is at least one order of magnitude cheaper than the Boeing/Endevco pressure belt based on the cost of the pressure sensors alone.

The pressure belt system was granted CAR35 approval for attachment to the test aircraft. The approval stipulated that the sensor modules could not be placed on any control surfaces including the flaps.

The pressure sensors were tested and it was found that the absolute accuracy of the pressure sensors compared well with the manufacturers specifications. The accuracy was checked using a pressure calibrator as well as by comparison with a portable weather station and local meteorological observations. The temperature compensation of the sensor modules was found to be working fine, after some initial doubts. The final aspect of the sensor performance that was investigated was the frequency response. This was done to evaluate whether a low-pass filter implemented in the microcontroller software would be beneficial. The noise base of the pressure sensors was measured and the frequency response to a triangular waveform was measured for comparison. It was found that the noise base level of 5dB compared well with the apparent noise level of 10dB from the frequency response tests and it was decided that there would be little benefit from the implementation of a low pass filter.

An error propagation analysis was carried out using the absolute sensor accuracy specified by the sensor manufacturer. The 95% confidence error in the pressure coefficient was calculated to be ±0.072 for a worst case situation. This compared well with the error calculated for a tubed pressure belt system used by NASA of ±0.05.

Numerical methods were used to predict the distribution of air pressure over a NACA 2412 airfoil. Two methods were used; the Hess-Smith panel method and XFOIL software. The numerical predictions were used for comparison with flight data as well as to estimate the effect of the configuration of the pressure belt system.

Three sensor module distribution patterns were analysed; a linear spacing of sensor modules, a cosine spacing of sensor modules and a custom spacing that used high resolution modules at the leading edge of the airfoil section. The cosine distribution was

115 found to the best method for reducing the errors in calculated force and moment coefficients, however it was decided that the custom distribution would be used as the error was not much larger but the use of the high resolution sensor modules at the leading edge would capture the rapid changes in pressure in this region well.

The effect of estimating the pressure coefficient at the trailing edge was also analysed. The approval to attach the sensor modules to the wing surface stipulated that modules could not be placed on the flap surface. This meant that in order to calculate force and moment coefficients from the measured pressure distributions, the trailing edge pressure coefficient at any angle of attack would need to be estimated. It was found that a trailing edge pressure coefficient estimate of zero for all angles of attack produced the least error in force and moment coefficients and this was the estimate that was used when integrating the distributions measured in flight.

Lifting line theory was used to estimate a relationship between the local angle of attack of the Cessna 182 wing section where the pressure belt system was to be attached and the aircraft angle of attack. Prandtl’s classic lifting line theory was used and the relationship was calculated for a wing station 1.74 metres from the aircraft centreline.

The results from the initial flight testing showed that the sensor attachment method was causing considerable error in the measured pressure coefficients. The exact cause of this error was thought to be the tape fairing of the pressure sensors to the wing surface. This was confirmed by using numerical panel methods to predict the effect of different fairing ratios. The solution was found to be a constant thickness filling in between the sensor modules. This was achieved using closed cell foam that was cut to fit over the sensor modules.

The steady state pressure distribution measurements showed good agreement between the flight results and pressure distributions predicted using numerical methods. The results did highlight the need for more single sensor modules to be distributed over the airfoil section, although the use of the high resolution sensor modules around the leading edge showed that reasonable results could be obtained with a minimal number of sensors.

The steady state force coefficients that were calculated from the measured pressure distributions also showed good agreement with force coefficients predicted using the

116 Hess-Smith method and XFOIL. Common trends were present between the flight data and the viscous predictions of XFOIL, although it is thought that better agreement could be achieved if more sensor modules were placed on the airfoil, particularly toward the rear of the airfoil section. There were large discrepancies between the moment coefficients measured in flight and the predicted moment coefficients, although this was expected and was due to the distribution of sensor modules and the estimate of the trailing edge pressure coefficient.

Comparisons between lift coefficients measured in flight and experimental wind tunnel data showed the lift coefficients from flight to be higher than those achieved in the wind tunnel. It is thought that this is due to the limited number sensor modules used.

An analysis of the flow separation that occurred during a stall manoeuvre highlighted the usefulness of the system for exploring actual dynamic flight conditions that are difficult to simulate in a wind tunnel or with CFD analysis. The path to complete flow separation showed that the flow partially separated and then reattached before complete flow separation was achieved at the wing position where the pressure belt was mounted. The recorded pressure distributions indicate that the stall mode was most likely to be leading edge stall which explained the difference in the stall angles between the measured flight data and the experimental wind tunnel data.

Short period pitch oscillations were carried out during the flight testing and the resultant force coefficients showed hysteresis to be present. The hysteresis was probably due to the pitching motion of the airfoil and although the motion contained both pitching and plunging motions, it is thought that the effect of the plunging motion did not cause any hysteresis.

6.3 Recommendations for future work The pressure belt system that was developed is versatile and could easily be applied to any surface over which the distribution of air pressure is to be measured. One area where the system could be particularly useful is in automobile aerodynamics. The ability to make pressure measurements on an actual car as it is driving would be very interesting. Some preliminary measurements were made with the sensor modules on a Ford Falcon station wagon showing some interesting results, including the effect of preceding and oncoming vehicle wakes. The main hurdle to overcome with measuring

117 the pressure distribution in this particular application is obtaining accurate dynamic and static pressure information in order to normalise the pressure measurements.

As far as future work with the system on the UNSW@ADFA aircraft, the first step would be to manufacture more sensor modules to increase the resolution of the distribution measurements. Repeating the steady state force coefficient measurements with more sensor modules, particularly sensor modules mounted on the flap section of the wing should improve the comparison with the numerically predicted force coefficients.

The CAR35 approval could be extended to allow sensor modules to be mounted on the flap section with a provision that the flap system be disabled for these flights. This has been done before for aircraft that have had tubed pressure belts mounted over the flap section of a wing. The preferable alternative to this would be to modify the system so the sensor modules could be mounted on the flap section and still allow the flaps to be deployed. A simple way of doing this would be to use radio transceivers to form a wireless link between the main airfoil and flap sections of the digital data bus. The wireless transmission of data would give the system another level of versatility as connecting strips of ribbon cable would not be required. However, individual power supplies would be required for each sensor module if they were to be made completely stand alone. With the low power consumption of the sensor modules, small watch batteries could be used to power each sensor module.

One area of further research could involve a comparison with wind tunnel testing. The advantage of comparing results from flight testing directly with pressure distributions measured using a full aircraft model in a wind tunnel is that lifting line corrections would not be needed to calculate the local section angle of attack. Both methods could be compared directly using the aircraft angle of attack. It would be difficult to carry out this testing using the UNSW@ADFA low-speed wind tunnel as the operating speed is approximately 20 m/s and with a scale model the Reynolds number of the C182 aircraft in flight could not be matched.

The hysteresis that was found during the short period pitch manoeuvre requires further investigation. Further flight testing with more sensor modules on the airfoil section is required. In order to gather more insightful data, the test aircraft should be

118 equipped with a pitch rate gyro to measure the aircraft pitch rates during the manoeuvre. When the test is repeated, rather than using a traditional short period pitch manoeuvre where the controls are released and the aircraft is allowed to return to the trim point, the pitch oscillation should be forced using control inputs to give more than a three quarter cycle of motion.

The pressure belt system has proven to be extremely useful for gathering aerodynamic data in-flight. Exactly how the system is used in the flight laboratory program needs to be determined.

119 References

1. Barnard, R.H. and Philpott, D.R. (1995). Aircraft Flight: a description of the physical principles of aircraft flight. 2nd Ed. Pearson Education, Harlow. 2. Bertin, J.J. and Smith, M.L. (1989). Aerodynamics for Engineers. 2nd Ed. Prentice-Hall, Englewood Cliffs, NJ. 3. Anderson, J.D. (2007). Fundamentals of Aerodynamics. 4th Ed. McGraw-Hill, New York. 4. Clyde, C. (2004). Tailplane inflight loads measurement for a light aircraft. Final year thesis, UNSW@ADFA. 5. Leth, O., Leth, G., Strash, D.J., and Leth, N. (2008) Engineering Solutions in Support of Supplementary Type Certificate to a Transport-Category Aircraft. Journal of Aircraft, Vol. 45, No. 1, 16-22. 6. Holland, M., Eccles, L., and Paradis, l. (2001) A Pressure Belt System for an Airborne Pressure Survey. Sensors for Industry Conference, Rosemount, Illinois. 7. Anderson, J.D. (1995). Computational Fluid Dynamics; The Basics with Applications. 1st Ed. McGraw-Hill, Singapore. 8. Bushnell, D.M. (2006) Scaling: Wind Tunnel to Flight. Annual Review of Fluid Mechanics. 9. Rivers, N.A., van Dam, C.P., Brown, P.W., and Rivers, R.A. (2001). Flight Investigation of the Effects of Pressure-Belt Tubing Size on Measured Pressure Distributions. Nasa Langley Research Center TM-2001-209857. 10. Davis, M.C. and Saltzman, J.A. (2000). In-Flight Wing Pressure Distributions for the NASA F/A-18A High Alpha Research Vehicle. NASA TP-2000-209018, Dryden Flight Research Center. 11. Landers, S., Saltzman, J., and Bjarke, L. (1997). F-16XL Wing Pressure Distributions and Shock Fence Results from Mach 1.4 to Mach 2.0. NASA TM- 97-206219, Dryden Flight Research Center. 12. Alam, F., Jaitlee, R., and Watkins, S. (2007) Aerodynamic Effects on an Automotive Rear Side View Mirror. Australasian Fluid Mechanics Conference, Gold Coast, Australia. 13. Scannivalve Corporation. Miniature Pressure Scanners. [Online]. Available: http://www.scanivalve.com/doc/product_pressure_miniature.htm [2008, July 16] 14. Montoya, L.C. and Lux, D.P. (1975). Comparisons of wing pressure distribution from flight tests of flush and external orifices for Mach numbers from 0.50 to 0.97. NASA Flight Research Center TM X-56032. 15. Whitmore, S.A. (1988). Formulation of a General Technique for Predicting Pneumatic Attenuation Errors in Airborne Pressure Sensing Devices. NASA TM100430, Dryden Flight Research Center.

120 16. Karolys, A. and Swanson, B. (1999) The Pressure Belt: A Smart Sensor Network. NASA Tech Briefs, Vol. 23, No. 10. 17. Rhode, R.V. (1930). The Pressure Distribution over the Wings and Tail Surfaces of a PW-9 Pursuit Airplane in Flight. NACA. 18. Liu, T. and Sullivan, J.P. (2005). Pressure and Temperature Sensitive Paints. 1st Ed. Springer, Berlin. 19. Morris, M.J. (1995) Use of Pressure Sensitive Paints in Low-Speed Flows. IEEE 16th International Congress on Instrumentation in Aerospace Simulation Facilities, Wright-Patterson AFB, Ohio. 20. Golfarrelli, A., Zagnoni, M., Proli, P., Callegari, S., Talamelli, A., Sangiorgi, E., and Tartagni, M. (2004) Acquisition System for Pressure Sensor Network. Proceedings of IEEE Sensors 2004, Vienna, AT. 21. Kim, N.P., Holland, M.J., Chien, C.-P., Tanielian, M.H., Wu, J., and Wong, C.P. (2001) Aircraft flight tests and reliability improvements of MEMs pressure sensor assembly. Journal of SMT, Vol. January. 22. Karolys, A. and Swanson, B. (1999) A Network of MEMS-based Smart Sensors Can Enhance Data Gathering: A Study in Signal Conditioning. ChipCenter Questlink Technical Notes [Online Serial], Available: http://archive.chipcenter.com/TestandMeasurement/tn006.html [2006, June 6]. 23. Kannemans, H. and Volkers, D.F. (1990) In-Flight Pressure Distribution Measurements; instrumentation, data handling and comparison with windtunnel data. ICAS congress proceedings, 1496-1505. 24. C. Williams. Introduction to Sensors. [Online]. Available: http://newton.ex.ac.uk/teaching/CDHW/Sensors/#CPS [2008, July 16] 25. VTI Technologies. SCP1000 Series-Absolute Pressure Sensor. [Online]. Available: http://www.vti.fi/en/products/pressure-sensors/pressure_sensors/ [2008, August 7] 26. SCP1000 Series Product Family Specification. 2007, VTI Technologies. 27. EDrawSoft. Network topology diagrams. Available: www.edrawsoft.com/Network-Topologies.php 28. Application Note 3884: How far and how fast can you go with RS-485. (Jul 25, 2006) Available: www.maxim-ic.com/appnotes.cfm/an_pk/3884 [2008, July 9] 29. Taylor, J.W.R., ed. Janes All the World's Aircraft. 1979, Jane's Publishing Goup: London. 30. Harrap, M.J. (2007) An Airborne Laboratory for Undergraduate and Postgraduate Education. AaeE Conference, Melbourne. 31. Katz, J. and Plotkin, A. (1991). Low-speed aerodynamics: from wing theory to panel methods. Ed. McGraw-Hill, New York. 32. Anderson, J.D. (2005). Introduction to flight. 5th Ed. McGraw-Hill, New York. 33. Coleman, H.W. and Steele, W.G. (1999). Experimentation and Uncertainty Analysis for Engineers. 2nd Ed. John Wiley & Sons, Inc., New York.

121 34. Griffith, M., Engineering Instruction sheet-Pressure sensor installation, Auto Avia Design. 35. Cebeci, T., Shao, J.P., Kafyeke, F., and Laurendeau, E. (2005). Computational Fluid Dynamics for Engineers. Ed. Horizons Publishing Inc., Long Beach, CA. 36. Cebeci, T. (1999). An Engineering Approach to the Calculation of Aerodynamic Flows. Ed. Horizons Publishing, Long Beach, Ca. 37. XFOIL: Subsonic Airfoil Development System. [Online]. Available: http://web.mit.edu/drela/Public/web/xfoil/ [2008, August 14] 38. Drela, M. (1989) XFOIL: An Analysis and Design System for Low Reynolds Number Airfoils. Conference on Low Reynolds Number Airfoil Aerodynamics, University of Notre Dame. 39. Abbott, I.H. and Doenhoff, A.E.V. (1959). Theory of Wing Sections. Ed. Dover Publications, Inc., New York. 40. Hoerner, S.F. (1985). Fluid-Dynamic Lift. 2nd Ed. L.A. Hoerner, 41. Zhenghong, G. (1999) Research on the Hysteresis Properties of Unsteady Aerodynamics about the Oscillating Wings. Applied Mathematics and Mechanics, Vol. 20, No. 8, 895-907. 42. Hess, J.L. and Smith, A.M.O. (1967) Calculation of Potential Flow about Arbitrary Bodies. Progress in Aeronautical Sciences, Vol. 8.

122 Appendix A. Hess-Smith panel method equations

The basis of inviscid panel methods is Laplace’s equation, which is a simplification of the Euler equations describing fluid motion.

sLaplace' n :equatio 2φ =∇ 0 as defined is where ∇ 2 is defined as ∂ 2 ∂ 2 2 ≡∇ + ∂ 2 ∂yx 2 bygiven is function, potential The potential function, φ is bygiven ∂φ ∂φ u = , v = (11) ∂x ∂y by satisfied is Continuity is satisfied by 2ψ =∇ 0 as defined is function, stream thewhere stream function, ψ is defined as ∂ψ ∂ψ u = , v −= ∂y ∂x

Hess-Smith Panel method The Hess-Smith method is a panel method that uses a combination of source and vortex elements to represent the flow. The name comes from the two developers of the method [42]. In order to calculate the pressure distribution over and airfoil, the airfoil is first divided into a number of panels. At the midpoint of each panel a source element and a vortex element are located. The strength of each element affects the flow around the airfoil with the combined effect of all the elements creating a flowfield that would be expected over an airfoil. The method is only two-dimensional and it is inviscid, neglecting any effects due to viscosity of the air.

The potential function for a two-dimensional source element is given by

q φ = ln r s 2π (12)

Where r is the radial distance from the centre of the source and q is the strength of the source.

A-1 Similarly the potential function for a vortex element is given by

Γ φ −= θ v 2π (13)

Where Γ is the strength of the vortex element.

The airfoil is divided into a finite number of short straight-line panels (N) as shown below. Each panel is defined by a pair of points, with the points beginning at the trailing edge and proceeding clockwise around the airfoil.

The panelling of an airfoil with source and vorticity elements placed at each panel mid-point. Reproduced from [35]. r The velocity, V at any point (x,y) is represented by

r r r += vUV (14)

r Where U is the uniform flow velocity, given by r r r = + αα jiVU )sin(cos ∞ (15)

vr is the disturbance field due to the body. If the body is represented by source and vortex elements, each element at point j induces a velocity at a point (x,y) according to

r = r )(),( + r τ )( dssvdssqvyxv ∫ ∫ jjvjjs (16)

Where dsq jj is the strength of the source element at point j and τ ds jj is the strength of the vortex element at point j.

A-2 One of the boundary conditions for the problem is that the body of the airfoil forms a streamline of the flow. This is analogous to ensuring that at each panel midpoint the flow only has a tangential component, i.e no normal velocity

V n 0)( i == ,...... ,2,1 N i (17)

To solve the Laplace equation, at each control point, i the normal and tangential velocity components due to the source and vortex elements on all panels, j are computed. These components are summed together with the freestream velocity components. It is convenient to write the normal and tangential velocities as

N N n )( n nτ VBqAV −++= θα )sin( i ∑∑ij j ij j ∞ i (18) j ==11j

N N t )( t tτ VBqAV −++= θα )cos( i ∑∑ij j ij j ∞ i (19) j ==11j

n n t t Where Aij , Bij , Aij , and Bij are known as the influence coefficients. Specifically

n Aij denotes the normal velocity components induced at the i-th panel by a unit strength

n source distribution on the j-th panel. Bij denotes the normal velocity component

t t induced by a unit strength vortex element, while Aij , and Bij denote the tangential components induced by the source and vortex distributions. The influence components are computed using the following expressions:

⎧ 1 ⎡ r ji +1, ⎤ ⎪ ⎢ −θθ ji ln)sin( −+ )cos( βθθ ijji ⎥ ≠ ji n ⎪2π ⎣⎢ r , ji ⎦⎥ Aij = ⎨ ⎪1 = ji ⎩⎪2

⎧ 1 ⎡ r ji +1, ⎤ ⎪ ⎢ −−− θθβθθ ln)cos()sin( ⎥ ≠ ji At = 2π ijji ji r ij ⎨ ⎣⎢ , ji ⎦⎥ (20) ⎪ ⎩0 = ji n t t n ij −= ij ij = ABAB ij

Where the geometric parameters are given by

A-3 1 2 2 2 ji +1, [ jmi +1 −+−= yyxxr jmi +1 )()( ] 1 2 2 2 , ji []jmi −+−= yyxxr jmi )()( 1 1 += ),( += yyyxxx )( mi 2 ii +1 mi 2 ii +1 (21) ⎛ − yy ⎞ ⎛ − yy ⎞ θ = tan −1 ⎜ +1 ii ⎟, θ = tan −1 ⎜ +1 jj ⎟ i ⎜ ⎟ j ⎜ ⎟ ⎝ +1 − xx ii ⎠ ⎝ +1 − xx jj ⎠ ⎛ − yy ⎞ ⎛ − yy ⎞ β = tan −1 ⎜ jmi +1 ⎟ − tan −1 ⎜ jmi ⎟ ij ⎜ ⎟ ⎜ ⎟ ⎝ − xx jmi +1 ⎠ ⎝ − xx jmi ⎠

The above equations satisfy the irrotationality requirement regardless of the nature of j (sq ) and τ j (s ) , with the specific values of the source strengths and vortex strengths adjusted to achieve the condition of flow tangency. The Hess-Smith method adopts the approach of assuming the source strength j sq )( to vary from panel to panel with the individual strengths adjusted to achieve flow tangency. The vortex strength is assumed to be constant on all panels and the single value is adjusted to achieve the correct circulation over the airfoil. The correct circulation is achieved by invoking the Kutta condition, which is the other boundary condition for the problem.

In panel methods, the Kutta condition is applied indirectly by setting another property at the trailing edge that gives the equivalent effect. In this case, the upper and lower surface total velocities are set to approach a common limit at the trailing edge. Because the normal velocity component at the trailing edge is zero, the tangential velocities on the upper and lower panels at the trailing edge must be equal.

t −= VV t ( )N ( )1 (22)

Introducing the flow tangency condition and noting that j = ττ gives

N N n τ N VBqA θα =−++ ,0)sin( i = ,...... ,2,1 N ∑∑ij j ij ∞ i (23) j==11j

The Kutta condition equation and the above equation from a system of algebraic equations whose solution is achieved as follows

The equations above can be written in the form

A-4 ~ ~ = bxA (24)

Where A is a square matrix of order N+1

11 aa 12 L a1 j L 1N aa N +1,1

21 aa 22 L a2 j L 2N aa N +1,2 MMMMMMM A ≡ aa a aa i1 i2 L ij L iN Ni +1, (25) MMMMMMM

N1 aa N 2 L aNj L NN aa NN +1,

N + 1,1 aa N + 2,1 L a + ,1 jN L + ,1 aa NNNN ++ 1,1

And ~x is a vector containing the element strengths

~ = qqqx τ ),,...,,...,( T 1 Ni (26) ~ And b is a vector containing the right hand side of the linear equation formulation ~ = bbbbb ),,...,,...,( T 1 NNi +1 (27)

The elements of the matrix A are given by

= ,.....2,1 Ni = Aa n , ijij = ,.....2,1 Nj

N a = n , = ,.....2,1 NiB Ni +1, ∑ ij (28) j=1

n n Where Aij and Bij are given above.

The last row in the A matrix represents the Kutta condition and is formulated as follows

N t t 1 j j τ ∑∑ 1 j ++ VBqA ∞ −θα 1 )cos( j=1

N N (29) ⎡ t t ⎤ −= ⎢ Nj j τ ∑∑ Nj ++ VBqA ∞ −θα N )cos( ⎥ ⎣ j=1 j=1 ⎦

Or as

A-5 N N t t t t ∑ 1 j Nj j τ ∑ 1 j +++ BBqAA Nj )()( j=1 j=1 (30)

V∞ θα 1 −−−= V∞ −θα N )cos()cos(

This means that

t t , jAAa =+= ,.....,2,1 N + 1,1 jjN Nj

N a t += BB t )( NN ++ 1,1 ∑ 1 j Nj (31) j=1

~ The components of the b vector are given by

i = −Vb ∞ α −θi ),sin( = ,.....,1 Ni (32) N +1 Vb ∞ θα 1 −−−= V∞ −θα N )cos()cos(

The system of equations was solved by inputing all the elements of the A matrix and the b vector into Matlab and using the command x=inv(A)*b to perform a matrix inversion solution. A function was written to formulate the A matrix and b vector and to return the pressure coefficients at each panel location after calculating the required source strengths and vortex strength. The inputs to this function were vectors containing the x and y points of each panel node and the angle of attack, α. The pressure coefficients at each point were then computed by

2 ⎛V t ⎞ C 1−= ⎜ i ⎟ ,iP ⎜ ⎟ ⎝V∞ ⎠ where (33) Vt at panel,any i was bygiven N N t t t ()i ∑∑ij j τ ij ++= VBqAV ∞ −θα i )cos( j==11j

A-6 Appendix B. CAR 35 approval

B-1

B-2

B-3

B-4 Appendix C. Electrical circuit schematic

C-1 Appendix D. Software functions

Slave microcontroller (sensor modules)

SCP1000 Slave header.h Header file containing all the setup parameters

SCP1000 Slave.c Main file containing most functions

interrupt: Interrupt service routine

main: Main execution loop

PicSetup: Pic configuration

Parity: Calculated the parity of a single byte

LoadData: Loads data from the sensor into a 5 byte buffer

ResetPic: Perform a software reset of the pressure sensor and the PIC

SPI.c Contained the functions for communicating with the SCP1000 sensor

InitialiseSensor: Confirms sensor startup was successful

SetMode: Sets the mode of the sensor

GetData: Routine to retrieve pressure and temperature data from the SCP1000 sensor

WriteDirectAccess: Routine to write to a direct access register on the SCP1000 sensor

ReadDirectAccess: Routine to read a direct access register on the SCP1000 sensor

WriteIndirectAccess: Routine to write to a direct access register on the SCP1000 sensor

ReadIndirectAccess: Routine to read a direct access register

WriteByteSensor: Routine to write a byte to the sensor using SPI

ReadByteSensor: Routine to read a byte from the sensor using SPI

D-2 Fail: Routine to run if a pressure sensor has failed initialisation

CheckStatus: Routine to check the status of the SPI communications and to clear a write collision bit if set

ReadDirectAccessISR: Routine to read a direct access register from the interrupt service routine

ResetSensor: Routine to perform a software reset of the SCP1000 sensor

RS485.c File containing the functions used to communicate on the digital data bus

USARTSetup: Setup routine for the PIC communications module

SendDataBuffer: Routine to send the data buffer over the digital data bus

PutData: Routine to send a single byte of data over the digital data bus

Master microcontroller

SCP1000 Master header.h Header file containing all the setup parameters for the master software

SCP1000 Master.c Main file containing the main functions

interrupt: Interrupt service routine

main: Main execution loop

Pic_setup: PIC configuration

USART.c Contains the functions used to communicate to the PC and the slave PICs

comsetup: Setup routine for USART module and address detection

putbytePC: Send a single byte to the PC

putbufferPC: Send a buffer of data from the slave PICs to the PC

putbyteSlave: Send a byte to the slaves

D-3 RestartDaisyChain: Routine to restart the daisy chain sequence used by the slaves to send data over the bus.

PC software

SCP1000 header.h Header file for configuration and global variable settings

SCP1000 pressure belt.c Contains all the main functions for displaying the pressure belt data for the PC software. The graphical user interface functions have not been listed here.

main: Main execution loop.

GetSensorInformation: Retrieves startup information from the pressure belt such as the addresses of the sensor modules connected.

RunStop: Starts or stops the data acquisition from the pressure belt system.

DataAcquisitionThread: this function retrieves and processes all the data from the pressure sensor modules. This function executes in a separate thread to the graphical user interface.

ConfigureFileOutput: Configures a file for the data to be saved in.

GroundRecord: Takes a reading from all the sensors on the ground for zeroing.

StartTrigger: Sends a trigger signal for the sensors to take a measurement.

AllTrigger: Function used to carry out a triggered acquisition at a fixed frequency.

ResetSensors: Sends a reset command to all sensor modules.

PlotRawData: Function to plot the results from the sensor modules on a stripchart.

PlotPressureCoefficient: Function to plot the calculated pressure coefficients.

UpdateInstrumentation: Function to update the auxiliary instrumentation displays.

CalculateLiftCoefficient: Function to calculate the lift coefficient from the measured pressures.

D-4 Data functions.c Contains all the functions for directly communicating with the sensor modules.

GroundInitialise: Function to command the sensor modules to perform a ground initialisation.

SendCommandSlave: Function to send a command to a sensor module.

SendCommandMaster: Function to send a command to the master module.

SendBytes: Function to send a byte out using the COM port on the computer.

ConvertPressure: Function to convert the raw digital data from the sensors into pressure values.

ConvertTemperature: Function to convert the raw digital data from the sensors into temperature values.

CheckParity: Function to perform a parity check on the digital data from the sensor modules.

PressureFromQNH: Function to convert the field QNH from the airport to a reference pressure for zeroing the sensor modules on the ground.

SetAddress: Function to set the previous sensor module address for each sensor module to follow during the daisy chain data dumps.

ConfigureDAQ: Function to configure the data acquisition system to acquire data from the aircraft data box.

DataBoxConvert: Function to convert the raw voltages from the data acquisition system into actual values representing the aircraft angles.

D-5