Winglet

Gulfstream V model winglet flutter tests at NASA Langley transonic wind tunnel

The term "winglet" was previously used to describe an additional lifting surface on an aircraft, e.g., a short section between wheels on fixed undercarriage, but today it refers to a near-vertical extension of the wing tips. The upward angle (or cant) of the winglet, its inward or outward angle (or toe), as well as its size and shape are critical for correct performance and are unique in each application. The wingtip vortex, which rotates around from below the wing, strikes the cambered surface of the winglet, generating a force that angles inward and slightly forward, analogous to a sailboat sailing close hauled. The winglet converts some of the otherwise-wasted energy in the wingtip vortex to an apparent thrust.

This small contribution can be worthwhile over the aircraft's lifetime, provided the benefit offsets the cost of installing and maintaining the winglets. Another potential benefit of winglets is that they reduce the strength of wingtip vortices, which trail behind the plane. When other aircraft pass through these vortices, the turbulent air can cause loss of control, possibly resulting in an accident. This possibility is greatest near airports and the minimum spacing requirements between aircraft operations at airports is largely due to these factors. Aircraft are classified by weight (e.g. "Light," "Heavy," etc.) in large part because the vortex strength is proportional (not linearly) to the amount of lift being generated by the airfoil. For this reason, wingtip vortices are typically most dangerous when an aircraft is at high lift coefficient and at a heavy gross weight. During takeoff, for example, flaps and slats are typically partially extended, the aircraft is at its heaviest and a large amount of lift is generated as the aircraft reaches rotate velocity and transition to climbout.

Winglets and wing fences also increase efficiency by reducing vortex interference with laminar airflow near the tips of the wing,[3] by 'moving' the confluence of low-pressure (over wing) and high-pressure (under wing) air away from the surface of the wing. Wingtip vortices create turbulence, originating at the leading edge of the wingtip and propagating backwards and inboard. This turbulence 'delaminates' the airflow over a small triangular section of the outboard wing, which destroys lift in that area. The fence/winglet drives the area where the vortex forms upwards away from the wing surface, since the center of the resulting vortex is now at the tip of the winglet.

Aircraft such as the Airbus A340 and the Boeing 747-400 use winglets. Other designs such as some versions of the Boeing 777 and the Boeing 747-8 omit them in favor of raked wingtips. Large winglets such as those seen on Boeing 737 aircraft equipped with blended winglets are most useful during short-distance flights, where increased climb performance offsets increased drag. Raked wingtips are now preferred[citation needed] over small winglets for long-distance flights, where increased fuel economy during the cruise phase is more important.

The Rutan VariEze, the first aircraft to use winglets in 1975

[edit] History and applications

The initial concept dates back to 1897, when English engineer Frederick W. Lanchester patented wing end-plates as a method for controlling wingtip vortices.[4] In the United States, Scottish-born engineer William E. Somerville patented the first functional winglets in 1910. Somerville installed the devices on his early biplane and monoplane designs.[5]

Vincent Burnelli received US Patent no: 1,775,474 for his "End Plating Wing Tips" on August 26, 1930.[6]

Dr. Sighard F. Hoerner was a pioneer in the field, having written a technical paper published in 1952 that called for drooped wingtips whose pointed rear tips focused the resulting wingtip vortex away from the upper wing surface. Drooped wingtips are often called "Hoerner tips" in his honor. Gliders and light aircraft have made use of Hoerner tips for many years.[7][8]

Winglet on KC-135 Stratotanker with attached tufts showing airflow during NASA tests in 1979–1980

[edit] NASA development

Hoerner's concept was further developed by Richard T. Whitcomb, an engineer at NASA's Langley Research Center, in response to the sharp increase in the cost of fuel after the 1973 oil crisis. Whitcomb's designs were flight-tested in 1979–80 by a joint NASA/Air Force team, using a KC-135 Stratotanker based at the Dryden Flight Research Center.[4] A Lockheed L-1011 and McDonnell Douglas DC-10 were also used for testing, and the latter design was directly implemented by McDonnell Douglas on the derivative MD-11, which was rolled out in 1990.[4] NASA's own most notable application of wingtip devices is on the Boeing 747 Shuttle Carrier Aircraft. Located on the 747's horizontal stabilizers, the devices increase the tailplane's effectiveness under the weight of the Space Shuttle orbiter.

Beechcraft Starship Model 2000

[edit] Composite aircraft

Even before NASA did flight testing on winglets, Burt Rutan incorporated them in his innovative Rutan VariEze homebuilt aircraft design, which made its first flight with winglets on May 21, 1975. The VariEze pioneered glass-reinforced plastic composite construction in homebuilt aircraft, which simplified fabrication of the winglets. He reduced the resulting drag penalty by assigning double duty to the winglets; they also serve as vertical stabilizers and rudders in his canard, pusher configuration aircraft. They were also used similarly on the derivative Rutan Long-EZ and reappeared on his Beechcraft Starship business aircraft design that first flew in 1986.[9] Conventional winglets were fitted to Rutan's Rutan Voyager, the first aircraft to circumnavigate the world without refueling in 1987. However, the aircraft's wingtips were damaged when they dragged along the runway during takeoff, breaking off about a foot of each wingtip, so the flight was made without benefit of winglets.[10]

Gulfstream V with winglets

[edit] Business aircraft

Learjet exhibited the prototype Learjet 28 at the 1977 National Business Aviation Association convention. The Model 28 prototype employed the first winglets ever used on a jet and a production aircraft, either civilian or military. Learjet developed the winglet design without NASA assistance. Although the Model 28 was intended to be a prototype experimental aircraft, performance was so impressive that it resulted in a production commitment from Learjet. Flight tests, made with and without winglets, showed that the winglets increased range by about 6.5 percent and also improved directional stability. Learjet's application of winglets to production aircraft continued with newer models including the Learjet 55, 31, 60, 45 and Learjet 40.

Learjet 60 with winglets

Gulfstream Aerospace also explored winglets in the late 1970s and incorporated winglets in the Gulfstream III, IV and V. The performance of the Gulfstream V has been exemplary. Its operational range of 6,500 nmi (12,038 km) permits routine nonstop business travel for routes such as New York–Tokyo. The Gulfstream V also holds over 70 world and national flight records.[11]

Winglets are also applied to several other business jets to reduce take-off distance, enabling operation out of smaller secondary airports, and allowing higher cruise altitudes for overflying bad weather, both of which are valuable operational benefits for corporate travel. In addition to factory-installed winglets on new aircraft, aftermarket vendors developed retrofit kits, for popular jets and turboprops, to improve both aerodynamics and appearance. Winglets became so popular on this class of aircraft that the Dassault Group, whose French designers resisted applying them on their Dassault Falcon line until recently, were forced to run a contrarian marketing campaign. Cessna recently announced they were partnering with Winglet Technology, LLC of Wichita, Kansas, to test a new

wingtip device called Elliptical Winglets, which are designed to increase range and increase payload on hot and high departures.[12]

A Boeing 747-400 with winglets

[edit] Passenger aircraft

Boeing announced a new version of the 747 in October 1985, known as the 747-400, with an extended range and capacity. With that particular model, Boeing used a combination of winglets and increased span to carry the additional load. The winglets increased the 747-400's range by 3.5 percent over the 747-300, which is otherwise aerodynamically identical but has no winglets.[13] Winglets are preferred for Boeing derivative designs based on existing platforms, because they allow maximum re-use of existing components. Newer designs are favoring increased span, other wingtip devices or a combination of both, whenever possible.

In 2002, Boeing first flew a production Next-Generation 737 with its new Blended Winglets, six-foot extensions that decrease fuel consumption by about 4 to 6 percent. The airplane gained supplemental type certification in 2003, and the majority of 737s delivered today are equipped with the devices.[citation needed]

Schempp-Hirth Ventus-2 glider with factory winglets winch-launching

[edit] Gliders

In 1987, mechanical engineer Peter Masak called on aerodynamicist Mark D. Maughmer, an associate professor of aerospace engineering at the Pennsylvania State University, about designing winglets to improve performance on his 15-meter wingspan racing sailplane. Others had attempted to apply Whitcomb's winglets to gliders before, and they did improve climb performance, but this did not offset the parasitic drag penalty in high-

speed cruise. Masak was convinced it was possible to overcome this hurdle.[14] By trial and error, they ultimately developed successful winglet designs for gliding competitions, using a new PSU–90–125 airfoil, designed by Maughmer specifically for the winglet application. At the 1991 World Gliding Championships in Uvalde, Texas, the trophy for the highest speed went to a winglet-equipped 15-meter class limited wingspan glider, exceeding the highest speed in the unlimited span Open Class, an exceptional result.[15] Masak went on to win the 1993 U.S. 15 Meter Nationals gliding competition, using winglets on his prototype Masak Scimitar.[16]

PSU-90-125 winglet airfoil profile

The Masak winglets were originally retrofitted to production sailplanes, but within 10 years of their introduction, most high-performance gliders were equipped from the factory with winglets or other wingtip devices.[17] It took over a decade for winglets to first appear on a production airliner, the original application that was the focus of the NASA development. Yet, once the advantages of winglets were proven in competition, adoption was swift with gliders. The point difference between the winner and the runner-up in soaring competition is often less than one percent, so even a small improvement in efficiency is a significant competitive advantage. Many non-competition pilots installed them for handling benefits such as increased roll rate and roll authority and reduced tendency for wing tip stall. The benefits are notable, because sailplane winglets must be removable to allow the glider to be stored in a trailer, so they are usually installed only at the pilot's preference.

[edit] Advertising

Advertising on WestJet Boeing 737-700 winglets

Some airlines capitalize on the visibility of winglets to passengers. AirTran Airways, American Airlines, Southwest Airlines, WestJet and Ryanair advertise their websites on the inboard side of their 737's winglets.

[edit] Notable examples

Winglets are employed on many aircraft types, such as:

Rutan VariEze , the first aircraft to use winglets (1975) Learjet 28/29 , the first production jet aircraft to use winglets (1977) Glaser-Dirks DG-303 , an early glider derivative design, incorporating winglets as

factory standard equipment Airbus A310-300 , the first airliner to feature wingtip fences (1985)[18]

Boeing 747-400 , the first mainline airliner to feature winglets (1988) Ilyushin Il-96 , first Russian and modern jet to feature winglets (1988) Bombardier CRJ-100 , first regional airliner to feature winglets (1992)[19]

Tupolev Tu-204 , first narrow body aircraft to feature winglets (1994)

[edit] Wingtip fence

A detailed view of the wingtip fence on an Airbus A319

A wingtip fence is a winglet variant, with surfaces extending both above and below the wingtip. Both surfaces are shorter than or equivalent to a winglet possessing similar aerodynamic benefits. Wingtip fences are the preferred wingtip device of Airbus, employed on all their airliners except for the A330 and A340 families. The A350 however, will make use of winglets rather than wingtip fences.

[edit] Blended winglets

Boeing 737 with blended winglets

A blended winglet is attached to the wing with smooth curve instead of a sharp angle and is intended to reduce interference drag at the wing/winglet junction. A sharp interior angle in this region can interact with the boundary layer flow causing a drag inducing vortex, negating some of the benefit of the winglet. The blended winglet is used on

business jets and sailplanes, where individual buyer preference is an important marketing aspect.

Blended winglets have been offered as an aftermarket retrofit for the Gulfstream II, Hawker 800, and the Falcon 2000 with winglets designed by Aviation Partners Inc., a Seattle, Washington-based firm that develops and markets blended winglets. A joint partnership of Aviation Partners, Inc. and Boeing, Aviation Partners Boeing offers blended winglets for the Boeing 737 Classic and Next Generation models,[1] 757 and 767.[20] The 737 version is now standard on the Boeing Business Jet derivative. Many operators have retrofitted their fleets with these for the fuel cost savings.

Airbus tested two candidate blended winglets, designed by Winglet Technology and Airbus themselves, for the Airbus A320 family,[21] but determined that their benefits did not warrant further development. In December 2008, Airbus announced that, in conjunction with Aviation Partners, Inc., they are restarting their winglet testing program for the A320, stating they are putting into practice the lessons learned from tests two years before. The stated aim of the new tests is to consider "an integrated Airbus programme"[citation needed].

In 2009 Airbus launched a new blended winglet design which the company called a "sharklet", designed to enhance the payload-range performance of the A320 Family. Offered as a retrofit option, sharklets are expected to result in a reduced fuel burn of at least 3.5 percent over longer sectors, corresponding to an annual CO2 reduction of around 700 tonnes per aircraft. The A320 will be the first model fitted with sharklets, which will be delivered in 2012.[22][23]

[edit] Raked wingtip

Boeing 787 Dreamliner rollout showing raked wingtip

Raked wingtips are a feature on some Boeing airliners, where the tip of the wing has a higher degree of sweep than the rest of the wing. The stated purpose of this additional feature is to improve fuel efficiency and climb performance, and to shorten takeoff field length. It does this in much the same way that winglets do, by increasing the effective aspect ratio of the wing and interrupting harmful wingtip vortices. This decreases the

amount of lift-induced drag experienced by the aircraft. In testing by Boeing and NASA, raked wingtips have been shown to reduce drag by as much as 5.5%, as opposed to improvements of 3.5% to 4.5% from conventional winglets.[13]

While an equivalent increase in wingspan would be more effective than a winglet of the same length, the bending force becomes a greater factor. A three-foot winglet has the same bending force as a one-foot increase in span, yet gives the same performance gain as a two-foot wing span increase.[24]

For this reason, the short-range Boeing 787-3 design called for winglets instead of the raked wingtips featured on all other 787 variants.

Raked wingtips are installed on, or are planned to be installed on:

Boeing P-8 Poseidon Boeing 747-8 Freighter Boeing 747-8 Intercontinental Boeing 767-400ER Boeing 777-200LR Boeing 777-300ER Boeing 777 Freighter Boeing 787-8 Boeing 787-9

[edit] Non-planar wingtip

DG Flugzeugbau DG-1000 glider with raked, non-planar wingtip and winglet

Non-planar wingtips are normally angled upwards in a polyhedral wing configuration, increasing the local dihedral near the wing tip. These provide the wake control benefit of winglets, with less parasitic drag penalty, if designed carefully. The non-planar wing tip is often swept back like a raked wingtip and may also be combined with a winglet. A winglet is also a special case of a non-planar wingtip.

Aircraft designers employed mostly planar wing designs with simple dihedral after World War II, prior to the introduction of winglets. With the wide acceptance of winglets in new sailplane designs of the 1990s, designers sought to further optimize the aerodynamic performance of their wingtip designs. Glider winglets were originally retrofitted directly

to planar wings, with only a small, nearly right-angle, transition area. Once the performance of the winglet itself was optimized, attention was turned to the transition between the wing and winglet. A common application was tapering the transition area from the wing tip chord to the winglet chord and raking the transition area back, to place the winglet in the optimal position. If the tapered portion was canted upward, the winglet height could also be reduced. Eventually, designers employed multiple non-planar sections, each canting up at a greater angle, dispensing with the winglets entirely.

Closed surfaces at the end of winglets are a possible way to eliminate the wake vortices induced at the tips of a wing. An example of a closed-surface winglet is the Spiroid winglet, a design currently under development by Aviation Partners. Initial testing using a Gulfstream II test aircraft has shown the winglet design to reduce fuel consumption in the cruise phase by over 10%.[25] These Spiroid winglets have also been flight tested on a Falcon 50 aircraft.[26]

A Falcon 50 with a spiroid winglet

Non-planar wingtips (without winglets) are or will be employed on:

Schempp-Hirth Discus-2 Schempp-Hirth Duo Discus Airbus A350-800 XWB Airbus A350-900 XWB Airbus A350-1000 XWB

[edit] Actuating wingtip devices

There has been research into actuating wingtip devices, including a filed patent application,[27] though no aircraft currently uses this feature as described. The XB-70 Valkyrie's wingtips were capable of drooping downward in flight, to facilitate Mach 3 flight using waveriding.

[edit] Use on rotating blades

"Winged rotor" on AgustaWestland AW101 Merlin helicopter

C-130J Super Hercules showing scimitar propellers with raked tips

Detail view of the wingtip device on a wind turbine rotor-blade.

Wingtip devices are also used on rotating propeller, helicopter rotor, and wind turbine blades to reduce drag, reduce diameter, reduce noise and/or improve efficiency. By reducing aircraft blade tip vortices interacting with the ground surface during taxiing, takeoff, and hover, these devices can reduce damage from dirt and small stones picked up in the vortices.[28]

[edit] Rotorcraft applications

The main rotor of the AgustaWestland AW101 (formerly the EH101) has a special "winged tip"; pilots have found that this alters the downwash field and reduces brownout which limits visibility in dusty areas and leads to accidents.[29]

[edit] Propeller applications

Hartzell Propeller developed their "Q-tip" propeller used on the Piper PA-42 Cheyenne and several other fixed-wing aircraft types by bending the blade tips back at a 90-degree angle to get the same thrust from a reduced diameter propeller disk; the reduced propeller tip speed reduces noise, according to the manufacturer.[28] Modern scimitar propellers have increased sweepback at the tips, resembling a raked tip on an aircraft wing

Abstract

Have you ever noticed how some jet planes have small, vertical projections as the tips of the wings? They're called winglets. What are they there for?

Objective

The goal of this project is to measure the effects on flight performance when winglets are added to a paper airplane design.

Introduction

The Boeing jet in the picture at right has winglets at the tips of its wings. Why are they there? What do they do?

As an airplane moves through the air, the wings generate lift by creating an area of low pressure above the upper surface of the wing. The higher air pressure beneath the lower surface of the wing lifts the plane. At the tip of the wing, the high and low pressure air meet.

The air forms miniature tornadoes, called wing tip vortices that spread out behind the plane (see Figure 1, right). Wing tip vortices cause two problems:

1. the turbulent airflow they create can be strong enough to flip an airplane that encounters it;

2. they also increase the drag forces on the airplane that generates them, decreasing fuel efficiency.

Winglets break up wing tip vortices, alleviating both of these problems.

In this project, you will test paper airplanes built both with and without winglets and measure the effect on flight performance. When doing your background research, you should also study vertical stabilizers. In the simple designs used in this project, winglets will also function as vertical stabilizers.

Terms, Concepts, and Questions to Start Background Research

To do this project, you should do research that enables you to understand the following terms and concepts:

fuselage, airfoil, winglets, vertical stabilizer, horizontal stabilizer, drag, lift, center of lift, center of gravity, wing tip vortices.

Questions

What are the three forces acting on a glider in flight? What relationship between these forces is needed for stable flight? How will the addition of winglets affect these forces? How will the addition of winglets affect flight performance?

Bibliography

Figure 1. The diagram shows the expanding wing tip vortices generated by a passenger jet. (NASAexplores.com, date unknown)

You'll definitely want to check out the Gliders section (among others) of NASA's Beginner's Guide to Aeronautics. This site is packed with useful information on the science of flight:NASA, 2005a. "Guided Tours of the Beginner's Guide to Aeronautics," NASA, Glenn Research Center [accessed June 8, 2006] http://www.grc.nasa.gov/WWW/K-12/airplane/guided.htm.

Here are two links with alternative designs for folded paper airplanes. The second link (Palmer, 2000) has an excellent plan (PL-1, "Joe's Favorite") for testing with and without winglets:

o NASA, 2005b. "Folding Paper Airplane: How To Build a JET Model," NASA Glenn Research Center [accessed June 8, 2006] http://www.grc.nasa.gov/WWW/K-12/WindTunnel/Activities/foldairplane.html.

o Palmer, J., 2000. "Joseph Palmer's Paper Airplanes," [accessed June 15, 2006] http://www.josephpalmer.com/planes/Airplane.shtml.

Here are some sources of information on winglets:o NASAexplores, 2002. "Taming Twin Tornadoes," NASA's Aeronautics

Mission Directorate, NASAexplores.com [accessed June 8, 2006] http://www.nasaexplores.com/show2_articlea.php?id=02-025.

o Larson, G.C., 2001. "How Things Work: Winglets," originally published in Air & Space/Smithsonian, Aug/Sep 2001, available online: http://www.airandspacemagazine.com/ASM/Mag/Index/2001/AS/htww.html.

Materials and Equipment

To do this experiment you will need the following materials and equipment:

paper for making airplanes, tape measure to measure flight distance, an indoor location with open space to test-fly the planes. Optional: stop watch to measure flight time.

Experimental Procedure

1. Do your background research so that you are knowledgeable about the terms, concepts, and questions above.

2. Start with your favorite paper airplane design. Figure 2, below, shows one popular model (see the first suggestion in the Variations section, below, for ideas on optimizing the design). This NASA link has another design you can try: http://www.grc.nasa.gov/WWW/K-12/WindTunnel/Activities/foldairplane.html.

Figure 2. The simple, classic folded paper airplane. 3. Using your chosen design, build several identical paper planes. 4. Test-fly each plane at least 5 times, and measure the distance flown. Be careful to

launch the planes at the same angle, and with the same amount of force each time. Note any instabilities in the flight characteristics (nose dives, rolling, turning). Optional: you can also use a stop watch to measure the flight duration. Keep track of the data in your lab notebook.

5. Fold a small portion of each wing tip up to create equal-sized winglets on each wing, and repeat the test flights.

6. Calculate the average flight distance for each plane, both with and without winglets.

7. Did flight distance improve with winglets? Were there improvements in other flight characteristics?

Variations

Experiment with the design of the simple folded airplane to optimize the flight characteristics before trying winglets. For example, you can shorten the plane by folding back a portion of the nose before folding up the wings (step 3 in Figure 2, above). (What effect does this have on the center of gravity? What effect does this have on the center of lift?) You can alter the surface area of the wings slightly by experimenting with exactly where to place the fold in step 4 of Figure 2. Test your designs with multiple flight tests and keep track of the results in your lab notebook. Then use your best design to see if winglets improve performance even further.

Experiment to find the optimal size for winglets. Does it matter if you fold the winglets down or up? The simple folded airplanes used in this project normally lack vertical stabilizers.

Vertical stabilizers resist forces that would tend to make the plane yaw (nose moving from side to side). In this simple type of paper airplane, winglets can function as vertical stabilizers. Another type of paper airplane (made with laminated construction methods) generally does include a vertical stabilizer as part of the design. For more details, see the Science Buddies project What Makes a Good Aerodynamic Design? Test Your Ideas with High-Performance Paper Gliders. Do winglets improve the flight characteristics of high-performance paper gliders?

For a more advanced project on winglets using a wind tunnel, see the Science Buddies project Winglets in Wind Tunnels.

For more science project ideas in this area of science, see Aerodynamics & Hydrodynamics Project Idea

There are many factors which influence the amount of aerodynamic drag which a body generates. Drag depends on the shape, size, and inclination, of the object, and on flow conditions of the air passing the object. For a three dimensional wing, there is an additional component of drag, called induced drag, or drag due to lift. On modern airliners, the wing tips are often bent up to form winglets. Winglets were wind tunnel tested and computer analyzed by Richard Whitcomb of the NASA Langley Research Center in the mid 1970's.

Induced drag is a three dimensional effect related to the wing tips; induced drag is a wing tip effect. So if the wing tip represents only a small fraction of the total wing area, the induced drag will be small. Long thin wings have low induced drag. Wings with an elliptical planform also have lower induced drag than rectangular wings. For many years, wing designers have attempted to reduce the induced drag component by special shaping of the wing tips. The Wright Brothers used curved trailing edges on their rectangular wings based on wind tunnel results. The outstanding aerodynamic performance of the British Spitfire of World War II is partially attributable to its elliptic shaped wing which gave the aircraft a very low amount of induced drag.

The idea behind the winglet is to reduce the strength of the tip vortex and therefore cause the flow across the wing to be more two-dimensional. Flight tests at the NASA Dryden Flight Research Center have found a 6.5% reduction in the fuel use of a Boeing 707 type airliner when using winglets. Winglets must be carefully integrated into the total wing design, which explains why many different winglet designs appear on various airliners.

For a wing, the total drag coefficient, Cd is equal to the base drag coefficient at zero lift Cdo plus the induced drag coefficient Cdi.

Cd = Cdo + Cdi

The drag coefficient in this equation uses the wing area for the reference area. Otherwise, we could not add it to the square of the lift coefficient, which is also based on the wing area.

You can further investigate the effect of induced drag and the other factors affecting drag by using the FoilSim III Java Applet. You can also download your own copy of FoilSim to play with for free.

Mark D. MaughmerThe Pennsylvania State UniversityUniversity Park, Pennsylvania 16802

Abstract

Although theoretical tools for the design of winglets for low-speed aircraft were initially oflimited value, simple methods were used to design winglets that gradually became accepted asbenefiting overall aircraft performance. As understanding was gained, improved methods weredeveloped, which ultimately resulted a number of successful applications of winglets. The current

approach incorporates a detailed component drag buildup that interpolates airfoil drag andmoment data across operational lift-coefficient, Reynolds-number, and flap-deflection ranges.Induced drag is initially predicted using a relatively fast multiple lifting-line method. In the finalstages of the design process, a full panel method, including relaxed-wake modeling, is employed.The drag predictions are used to compute speed polars for both level and turning flight, yieldingpredicted performance that is in good agreement with flight-test results. These methods have beensuccessfully applied to the design of winglets to improve the cross-country soaring performanceof both span-limited and span-unlimited, high-performance sailplanes, as well as to improvevarious mission capabilities for several different categories of powered aircraft.

Nomenclaturespanwing chordsection lift coefficientwinglet heightprofile drag coefficient averaged over spaninduced-drag factorplanform areaairspeedaverage cross-country speedcrossover speedsink rateweightair density

wingwingletwing tip

Introduction

Over the past fifteen years, from initially being able to do little to improve overallsailplane performance, winglets have developed to such an extent that few gliders now leave thefactories without them. This change was brought about by the efforts of a number of people tobetter understand how winglets work, to develop theoretical methods to analyze their

performance, and to develop design methods that allow the benefits to be tailored such that gainsin cross-country performance are achieved over a wide range of soaring conditions. The story ofthis development is an interesting case study in engineering design, in which trial and error,theoretical analysis, and flight testing all contributed to the successful solution of a difficultproblem.The efforts at Penn State to develop winglets for high-performance sailplanes began inthe early 1980’s as a collaborative effort with Mr. Peter Masak to design winglets for the 15mClass competition sailplanes of that era. Although work had already been done in the area of non-planar wings and winglets, in practice it was found winglets provided little or no benefit tooverall sailplane performance.1-4 The widely held belief at that time, essentially the same as thatheld for transport-type aircraft, was that while climb performance could be improved, it could notbe done without overly penalizing cruise performance. Thus, it was with some skepticism thatefforts were undertaken to work on this problem.The first steps taken were directed toward the design of an airfoil specifically intendedfor use on a winglet. Although not a great deal was known at this time about exactly how asailplane winglet should operate, it was clear that a winglet does not operate exactly as a wingand, consequently, an airfoil intended for use on a wing would not be a good choice for a winglet.Thus, the PSU 90-125 airfoil was designed. This was a robust design that was intended to operateover a very broad range of conditions.From this point, a trial-and-error process was begun that used flight testing as the primarymethod of determining the important design parameters. Although vortex-lattice and panelmethods were of some value for gaining insight, they were unable to predict drag accuratelyenough to be of use in the actual design process. Likewise, because the beneficial influence of awinglet is due to it favorably altering the flow field over the entire wing, meaningful wind-tunnelexperiments require a full- or half-span model. Thus, unless the wind tunnel has a very large testsection, the high aspect ratios typical of sailplanes result in model chords that produce

excessively low Reynolds numbers. To address these problems, methods of simulating full-scaleflow fields with truncated spans have been explored but, in every case, the necessarycompromises produce results that are somewhat questionable.5 For these reasons, the parametersthat were deemed the least important were set to reasonable values, while the more critical oneswere determined from flight test. Using some of the results from earlier work on winglets fortransport aircraft,6, 7 along with some simple calculations, the winglet height, planform, and cantangle, as defined in Fig. 1, were fixed. The goal from this point was to establish the spanwise loaddistribution on the winglet that would interact in a favorable way with the wing and therebyproduce an overall drag reduction. Because the basic shape of this loading can be adjusted witheither twist or sweep, the twist was set, again being guided by the earlier work on winglets, andthe sweep iterated until the desired result was obtained. For minimum induced drag, if theplanform is somewhat close to elliptical, the load distribution would have spanwise liftcoefficients that are essentially constant. Thus, with the planform set, the sweep was adjusteduntil yarn tufts indicated a uniform stall pattern in the spanwise direction. The last designparameter to be determined was the toe angle. Because there seemed to be little benefit in havingthe winglet carry load beyond that of the wing, the toe angle was adjusted until both the wing andthe winglet stalled simultaneously, again as determined tufts.Although it took some time and competition successes, the winglets that were the resultof the process were the first ones that were generally accepted as beneficial to overall cross-

country performance over a wide range of thermal sizes and strengths.8 In 1989, one of thesedesigns was adopted by sailplane manufacturer Schempp-Hirth and became the “factory winglet”for the Venus. In retrospect, with the understanding that has come since, it seems that thisprocess, while systematic and logical, was accompanied with a great deal of luck. It now seemssomewhat remarkable that with the tools then at hand, it was possible to configure a winglet thatactually worked!

Finite Wing Aerodynamics and Winglets

In essence, the improvement in aircraft performance due to winglets results from theirability to reduce induced drag traded off against their added wetted area increasing the profiledrag. Profile drag is the drag due to the shape of the airfoil or wing section. It is a consequence ofboth the skin-friction drag due to air moving along the surface of the airfoil, as well as pressuredrag, due to pressures acting over the front of a body not being balanced by those acting over itsrear. This pressure imbalance is the result of flow separating over the rear of the body, as well astotal pressure losses in the boundary layer. To measure profile drag in a wind tunnel, a constant-chord wing using the airfoil of interest is made to span the width of the wind-tunnel test section.Thus, the flow is not free to come around the wing tips. As a consequence, the flow is two-dimensional. The absence of spanwise flow causes the wing sections to behave as though theybelong to a wing of infinite span. Profile drag depends on, among other things, the amount ofwetted area and the shape of the airfoil and its angle of attack. Profile drag increases with thesquare of the airspeed, V2.Induced drag is the drag that is a consequence of producing lift by a finite wing. Inproducing lift, there must be higher pressure on the underside of the wing than there is on theupper side. As this pressure difference “wants” to equalize, there is a flow around the wingtipfrom the high-pressure air on the underside of the wing to the low-pressure air on the upper side,as shown in Fig. 2. As depicted in Fig. 3, this results in spanwise flow on the finite wing that wasnot present on the infinite wing. This component of spanwise flow is present in the flow leavingthe trailing edge, that from the upper surface flowing inboard while that on the lower surfaceoutboard. At the trailing edge, these two streams meet with a spanwise component of velocitygoing in opposite directions. As a consequence, vorticity is shed from the trailing edge which,within a short distance downstream, rolls up into two well defined tip vortices.Clearly, the generation of tip vortices requires energy and one approach to calculating theinduced drag is through determining how much energy is contained in the trailing vortex system.

This vortex system can be idealized as the “horseshoe” vortex system depicted in Fig. 4. As aconsequence of producing lift, “an equal and opposite reaction” must occur. While there are manyways to describe the generation of lift, one is that the upward lifting force being producedrequires that a certain amount of air be given a downward velocity, or downwash, as is indicatedin the sketch. Thus, producing a given amount of lift is accompanied by the generation of acertain amount of downwash and as, a consequence, a certain amount of induced drag. Tominimize this drag, the amount of energy used in producing the required downwash must beminimized, that is, the energy that is “wasted” in creating unnecessary spanwise flow and in therolling up of the tip vortices must be minimized.In observing the flowfield around the wing sketched in Fig. 3, it should be clear that thegreater the span, the less the tip effect is felt on the inboard portions of the wing. That is, thegreater the span, the more “two-dimensional like” will be the rest of the wing and, consequently,the less its induced drag. As the span is taken to infinity, the downwash and induced dragapproach zero. Likewise, if the wing is not producing lift, there will be no downwash and thus noinduced drag. Induced drag is a function of the inverse of the square of the airspeed, 1/V2, and thesquare of the span loading, (W/b)2. Among other things, it also depends on the wing planform

itself and how efficiently it produces lift with respect to induced drag. As a reference point, themost efficient planar wing is that having an elliptical loading.17 Typical planar wings aresomewhat less efficient, while non-planar geometries can be somewhat better than the ellipticalcase.It has been known for over a century that an endplate at the tip of a finite wing can reducethe spanwise flow and thereby reduce the induced drag. Unfortunately, to be effective at this, theendplate must be so large that the increase in skin friction drag far outweighs any induced dragreduction. A winglet, rather than being a simple fence that limits the spanwise flow, carries anaerodynamic load producing a flowfield that actively interacts with that of the main wing toreduce the amount of spanwise flow. That is, the downwash (sidewash) produced by the winglet

opposes the spanwise flow on the main wing. This effect has been measured experimentally andis shown in Fig. 5, where it can be seen that the spanwise flow has been largely eliminated by thepresence of the winglet. In essence, the winglet diffuses or spreads out the influence of the tipvortex such that the downwash, and consequently the induced drag, is reduced. In this way, thewinglet acts like an endplate in reducing the spanwise flow but, by carrying the properaerodynamic loading, it accomplishes this with much less wetted area. Nevertheless, recalling thepenalty of profile drag with increasing airspeeds, the designer’s goal is that of gaining the largestreduction in induced drag for the smallest increase in profile drag.

The Winglet Design Process

To obtain the desired results over the entire range of operation of an aircraft, it isnecessary to design a new winglet for every application. The area, height, cant angle, sweepangle, twist angle, and the all important toe angle must be uniquely determined to achieve thedesired performance goals. Thus, even though the trial and error process described resulted in asuccessful winglet, much remained to do in the development of tools and methods for analysisand design. Through the efforts of a succession of excellent students,5, 9-11 a great deal has beenaccomplished at Penn State which has bettered this situation.The first accomplishment of these efforts was the design and testing of a new airfoil.With a much better understanding of the operating conditions of a winglet, the PSU 94-097 airfoilwas designed to have much less conservatism than its predecessor.12 Following this, theoreticalmethods have been developed and validated through comparison glides and flight-testmeasurements. As a result, the design tools are now quite reliable and the products of thesemethods typically meet their design goals without modification.13-15 Winglets have been designedfor a number of sailplanes and powered aircraft, including those used on the Schempp-HirthVentus 2ax, shown in Fig. 6, and those under development for the Discus 2, presented in Fig. 7.

Crossover-Point MethodThe first attempt to better quantify the winglet design process made use of what has been

termed the crossover point on the sailplane speed polar. This point corresponds to the speed atwhich the flight polars of the aircraft without winglets and with winglets intersect or,equivalently, where the change in sink rate due to the winglets is zero. As noted, the profile dragincreases as V2, while the induced drag increases with 1/V2. Thus, the crossover point is a simpleway to make the tradeoff between the profile-drag penalty and the induced-drag benefit. Belowthis speed, winglets are beneficial, while above it they are detrimental. The crossover point is theflight speed at which the benefit in induced drag due to winglets is equal to the profile-dragpenalty, that is, when

∆DPROFILE + ∆DINDUCED = 0

The more the induced drag can be reduced for a given increase in profile drag, the higher thecrossover point and the more effective the winglet.To understand the factors that determine the crossover speed, VCR, an expression can beobtained by equating the increase in profile drag due to winglet height with the resulting decreasein the induced-drag factor

VCR =

2Wρb

∆K(h)π∆hc C Dp,WL

4

where ∆K(h) is a function relating the reduction in the overall induced-drag factor to a givenincrease in winglet height, h. Originally, this function was estimated using results from earlierwork.6-7, 10 The lower the profile-drag coefficient of the added winglet area, CDp,WL, and the greaterthe span loading, the higher the crossover speed, whereas increasing the winglet height reduces it.This simple expression for VCR gives insight into how the crossover point can becontrolled through the geometry of the winglet. In the early stage of development, the crossover

point was simply set to be higher than the cruising speed corresponding to the strongest thermalstrength anticipated. The use of this expression resulted in winglets that generally improvedoverall performance and, although based on a simple concept, was as accurate as the somewhatcrude ability to predict the changes in induced drag due to changes in winglet geometry.

Modified Crossover-Point MethodAs the ability to predict the induced drag for a given wing geometry improved, 5, 9 thecrossover-point method was modified. Rather than equating the change in profile drag with thechange in induced drag in terms of winglet height only, the expression is written more explicitlyin terms of parameters describing the winglet geometry and the resulting aerodynamic influencesas

(S C Dp )WL - (S C Dp )WT +

4W 2

2

4VCR

πρ

K2K1- 22b

b1 2

= 0

where the “WT” subscript corresponds to the wingtip region that is removed to mount the winglet,

the subscript “1” to the original wing, and “2” to the one modified with winglets. The weight ofthe aircraft, W, is considered to be unchanged by the wingtip modification. For restricted spanclasses, of course, b1 = b2. The problem for the winglet designer is to minimize the profile-dragincrease due to adding the winglet, to maximize the drag reduction resulting from removing theoriginal wingtip to mount the winglet, and to achieve the greatest induced-drag reduction bymaking the induced-drag factor, K2, as small as possible relative to K1. Likewise, the net areaincrease should be minimized, as should the profile drag coefficient corresponding to any addedarea. While this expression does not capture all of the details of winglet design, it does capturethe essence.Using either of the closed-form relations presented to guide the winglet design, atraditional drag buildup was performed to predict aircraft speed polar. Then crossover speed isadjusted, primarily using the toe angle, to allow the winglet to benefit performance over somepart of the operational speed range. Shifting the crossover speed not only affects the speed rangeover which a benefit is achieved, but also the magnitude of that benefit across the chosen range.Shifting it to higher speeds reduces the performance gains due to the winglet at lower speeds,whereas shifting it to lower speeds achieves a much larger drag reduction, but only over a smallportion of the flight polar.

A number of winglets were designed, fabricated, and flight tested using this method, andwhile based on simple ideas, these efforts contributed to the basic understanding of wingletdesign. First, whether it be with up-turned tips or winglets, it is beneficial for the design to be“out-of-plane.” Second, while a great deal of work has been directed toward determining theoptimum geometries for minimum induced drag,9, 16 experience has shown that too muchemphasis on this optimum penalizes the profile drag far more than can be offset by the induced-drag reduction.13-15 The design goal is to minimize the overall drag, not just one component of it.

For example, the optimum loading for minimum induced drag must be continuous across thejuncture between the wing and the winglet, which requires the chords at the juncture to be thesame, or that the lift coefficient at the root of the winglet to be proportionally greater than that ofthe wingtip. Either way, the amount of wetted area or the increase in lift coefficient results inprofile drag that is considerably greater than that of current designs. Thus, although not optimalwith respect to minimizing induced drag in accordance to classical theory,16 winglets as currentlydesigned achieve most of this reduction, and do so with a much lower profile drag increase thanwould otherwise be the case. In short, much of the optimal induced-drag reduction predictedtheoretically is obtained by adding winglets to the wing. Once this is done, minimizing the profiledrag of the winglet is paramount.

Present Design Approach for High-Performance SailplanesMany of the comments on winglet design presented thus far are applicable to any low-speed aircraft, while the details of the design methodology depend to a large extent on theparticular mission of the intended aircraft. In the case of high-performance sailplanes, the broadnature of the mission profile greatly complicates the choice of an optimum crossover speed. Inweak conditions, gains in climb offset losses in cruise. Conversely, in strong conditions, notpenalizing high-speed cruise is of the most importance to overall cross-country performance.While the crossover-speed method is effective for predicting the change in aircraft performancedue to the addition of winglets, and it does ensure some benefit, its use will generally not producethe best design. For high-performance sailplanes the optimal configuration cannot be determinedwithout specifically taking into account the impact of the winglets on the average cross-countryspeed. To do this, a fast, accurate prediction of the aircraft performance has been developed andcombined with a sailplane cross-country performance model, allowing the calculation ofMacCready average cross-country speeds for specific weather conditions and aircraftconfigurations.11, 13, 15 These average cross-country speeds are then used as the metric to

determine the suitability of a design. This approach allows the entire flight profile to be taken intoaccount in the design and yields a simple result encompassing the broad range of contributingfactors.Previous methods were not able to accurately and rapidly account for small changes in anaircraft configuration. The simplifications typically used, such as approximated airfoilcharacteristics and parabolic flight polars, introduce errors that are of the same order as theimprovements due to winglets. While useful for exploring trends and the basic characteristics ofwinglets, these methods are not accurate enough for design.

Performance PredictionThe calculation of sailplane performance is a major component of the winglet designproblem. The performance evaluation must have sufficient resolution to account for the effect ofchanges to the winglet geometry. Because these effects can be relatively small and errors orinconsistencies in other portions of the calculation can overshadow them, it is important that allaspects of the performance calculation be accurately determined. The accuracy necessary forsuccessfully undertaking activities such as winglet design is obtained through the use of aperformance program that has been developed to predict the straight- and turning-flight polars ofsailplanes.11, 13-15 In addition to the drag contributions of the major components of the sailplane,

the program accounts for the effects of airfoil characteristics, trim drag, static margin, flapgeometry, and flap-deflection scheduling. The most important element of the method is theanalysis of the wing-planform aerodynamics.Essential to the analysis method is the interpolation of the airfoil data. Wing profile dragis such a large portion of the overall drag that small errors in its determination can eclipse theeffects of winglets. To accurately provide such data, it is necessary to interpolate the airfoil dragand moment data over the operational ranges of lift coefficient, Reynolds number, and flapdeflection.The other essential component for predicting the planform aerodynamics is thedetermination of the span efficiency and lift distribution. The lift distribution directly affects the

wing profile drag, and the planform efficiency dictates the induced drag of the wing. Because thisis where the benefit of the winglet is quantified, an accurate method of determining these twoitems is of critical importance. In the present approach, use is made of both a multiple lifting-linemethod and a three-dimensional lifting-surface panel code. The multiple lifting-line method,which has been integrated directly into the performance program, has several chordwise liftinglines, each having a second-order vorticity distribution.2 This produces a continuous sheet ofvorticity that is shed into the wake. The method allows the spanwise lift distribution and induceddrag of non-planar wing geometries to be predicted with reasonable accuracy and less computa-tional effort than is required by a three-dimensional panel method. Although not accounting forthe consequences of thickness and a free wake, the multiple lifting-line procedure is able toquantify the effects of winglets. For initial design iterations, the increased speed of the multiplelifting-line method more than offsets the small loss in accuracy.For the final detailed design of the winglet, use is made of a panel method program thattakes free-wake effects into account.9 For the calculation of induced drag; the program applies theKutta-Joukowsky theorem at the trailing edge.17 This eliminates some of the problems associatedwith attempting to account for wake relaxation in the far field using a Trefftz-plane approach.While the differences in results between a relaxed wake and a fixed wake analysis are generallysmall, these differences can be important in determining the final winglet toe and twist angles.5The turning-flight performance of the sailplane is obtained by adjusting the straight-flightpolar for bank angle and load factor. By these means, the minimum sink rate, optimal bank angle,and optimal flight velocity as a function of turning radius are determined. The effects of deflectedailerons and the curved flow field are neglected.

Analysis of Cross-Country PerformanceWith straight- and turning-flight polars available, an analysis of crossover speeds ispossible but, as mentioned previously, a more rigorous means of evaluating designs is desirable.

This task is accomplished with a program that calculates the MacCready average cross-countryspeeds for a given configuration using the straight- and turning-flight polars generated by theperformance program.11, 13-15The thermal model used in this analysis has a distribution of vertical velocity that variesparabolically with thermal radius. Thus, the thermal profile is specified in terms of the magnitudeof the vertical velocity of the rising air at the core and the radius. The thermal profile has asignificant impact on the cross-country performance of a sailplane, and the most realisticperformance index would result from some particular mix of thermal strengths and profiles. Thiscould be done, but instead a single, representative thermal profile is used here, as this greatlysimplifies the interpretation of the results while still yielding a meaningful comparison betweensailplanes having different winglet geometries.To obtain the optimal climb rate for a particular configuration, the thermal profile issuperimposed over the predicted turning polars. The straight flight polar is then searched for theinter-thermal cruise speed to optimize the MacCready cross-country speed. The result is a trade-

off of climb and cruise performance, properly weighted to account for the variations in soaringconditions over which the sailplane might be operated.

Cross-Country Performance Gains: A Case Study

To see the performance increases that are possible with winglets, the predicted speedpolars for the Schempp-Hirth Discus 2, with and without winglets, ballasted and unballasted, areshown in Fig. 8. Although gains are demonstrated, they are difficult to assess because of thescales used on the polars shown. Thus, these data are replotted in terms of L/D verses velocity inFig. 9. In addition to demonstrating the gains in carrying water ballast at higher cruising speeds,the benefit of winglets can now be seen. To get an even better idea of the gains in L/D, in Fig.10these data are again replotted in terms of the percentage increase in L/D relative to the unballastedand ballasted glider without winglets. It should be noted that this winglet is such that thecrossover points occur at airspeeds that are above the maximum allowable. As already noted, the

crossover point that was so important in earlier winglet designs is no longer a factor in currentdesigns. This is because experience has demonstrated that even though better overall performancecould be achieved using the crossover point concept, this approach can result in a very largeperformance penalty if the winglets are operated much above the crossover speed. The problem isthat during inter-thermal cruise in very strong conditions, there are strong psychological andstrategic reasons for a pilot to “stay with the pack.” Unfortunately, the glider with wingletssuffers a very large performance penalty for flying faster than the crossover speed, which theglider without winglets does not. Thus, as is typical of the more recent designs, for this designthere are no allowable flight conditions at which the winglets penalize performance. While thepercentage gain in L/D does not appear to be very great, it is a gain that comes without anypenalty at higher speeds.The influence of winglets on the percentage change in average cross-country speedrelative to that of the baseline aircraft, that is without ballast and without winglets, is presented asa function of thermal strength in Fig. 11. The winglets improve the cross-country performance forall the thermal strengths considered, that is, for thermals having a 150 m radius and strengths,averaged across the diameter, of up to 6 m/s. As expected, the performance gain due to wingletson the unballasted glider is very significant for weak thermals as the winglets allow for someclimb rate, whereas without winglets, it is minimal or zero. As the thermal strengths increase, thebenefit due to winglets decreases; however, for this glider winglets do not hurt cross-countryspeed even for average thermal strengths of more than 6 m/s. The point at which full water ballastbecomes beneficial is indicated by the crossing of the unballasted and ballasted curves at anaverage thermal strength of just above 4 m/s, corresponding to a climb rate with full ballastpredicted to be about 2.7 m/s. As indicated, ballast causes a reduction in average cross-countryspeed for average thermal strengths of less than 4 m/s. For thermal strengths greater than this,

winglets improve the cross-country speed, but only by a half-percent or so. The glider withwinglets, however, can carry ballast to slightly weaker conditions without penalty than the gliderwithout winglets can.

Other Considerations

In designing winglets for a variety of sailplanes, as well as for a few non-sailplaneapplications, it seems to be true that all wings can be improved with winglets, although the betterthe original wing from an induced drag standpoint, the smaller the gain possible with winglets(and the more difficult is the design process). The case presented here, in fact, represents one ofthe smallest gains due to winglets thus far achieved. It is sometimes heard that winglets were tried

on “such and such” a glider but did not work. What this actually says is that a poor design did notwork. As an example of how critical some of the design issues can be, the effect of the winglettoe angle on the Discus 2 winglet design is presented in Fig. 12. Obviously, a small deviationfrom the optimum can cause the winglet to become a speed brake. Furthermore, as suchparameters are unique to each type of glider, each glider must have winglets specifically designedfor it. Rules of thumb regarding winglet design can be disastrous. It is certainly true that it ismuch easier to make a glider worse with winglets than it is to make it better!In some cases, it has been found that the winglets fix some problem with the originalwing. For example, in the case of a flapped glider, it is important that the flaps/ailerons extend tothe wingtip. Otherwise, when the flaps are deflected upward for high-speed cruise, the tips areloaded far more than they should be for the optimal spanwise loading. Although only a smallportion of the wing is seemingly influenced, it results in very significant induced drag increase. Inthese cases, cutting the tip back to the aileron in order to mount the winglet can result in gains,especially at high speeds, that would not be expected by the addition of winglets. In addition, itshould be noted that although the current generation of Open-Class gliders still benefit from tip

treatment, unless the wing loadings can be increased dramatically, increasing spans eventuallyreach the point where the penalty of any wetted area addition cannot be overcome by an induceddrag benefit. This is true whether the additional area is due to a span increase or a winglet.Nevertheless, because of the fact noted that a winglet can achieve a given reduction in induceddrag with less wetted area than a span extension, it has been the case that if a span extensionbenefits performance, then it is benefited even more if a winglet is added to the extension.From the understanding of how winglets achieve an induced drag reduction, it alsobecomes clear that they can yield other performance and handling qualities gains as well. Inparticular, it has been found that winglets improve the flow in the tip region and thereby improvethe effectiveness of the ailerons. This is in part due to the local angle of attack in the vicinity ofthe ailerons being reduced less by the reduced downwash velocities, as well due to the reductionof spanwise flow, helping to keep the ailerons effective. One of the benefits of greater controleffectiveness is that smaller aileron deflections are required for a given rolling moment. This notonly results in less drag for a given roll rate, but it also allows for the achievement of higher rollrates. Likewise, woolen tufts attached to glider wings have shown that much of the flowseparation that is observed over the inboard tip during turning flight is essentially eliminated bythe presence of a winglet. In addition to the resulting reduction in drag, winglets benefit safety inthat the ailerons now remain effective much deeper into a stall than before.

Closing Comments

Although the performance gains achieved with winglets are only a few percent atmoderate thermal strengths, such small differences can be an important factor in determining theoutcome of many cross-country flights or contests. For example, in a recent U.S. Open ClassNationals, less than 1.5% of the points awarded to the first-place competitor separated the first sixplaces, far less than the performance advantage that can be achieved using winglets.So, since their shaky introduction many years ago, the acceptance of winglets is nowwidespread. Shortly after their introduction to sailplane racing, only 19 of the 105 gliders

competing at the World Championships in Uvalde, Texas in 1991 used winglets. At the presenttime, sport and racing sailplanes in almost every class make use of winglets or some type of tiptreatment. Thus, after over a decade of winglets being applied to sailplanes, it is clear that thebenefits are far reaching. If properly designed such that the profile drag penalty is of noconsequence over the range of airspeed at which the glider is flown, then there seems to be noreason whatsoever not to take advantage of the performance and handling qualities benefits thatwinglets offer

Finally, although some of the spinning characteristics of gliders with winglets have beenexplored, the testing has not been extensive. The anecdotal evidence, however, generallyindicates that gliders with winglets are more reluctant to spin, but once they do, the altituderequired for recovery is somewhat greater than for the glider not equipped. Given the largenumber of glider fatalities that are a consequence of stall/spin accidents during approach, forwhich the altitude required for recovery is already insufficient, a question worth pondering iswhether or not even the most basic training gliders might benefit from the installation of winglets. The present report describes the numerical investigation of theaerodynamics around a wind turbine blade with a winglet usingComputational Fluid Dynamics, CFD.

Five winglets were investigated with different twist distribution andcamber. Four of them were pointing towards the pressure side (upstream)and one was pointing towards the suction side (downstream). Additionally,a rectangular modification of the original blade tip was designed with thesame planform area as the blades with winglets.

Results show that adding a winglet to the existing blade increase the forcedistribution on the outer approx 14 % of the blade leading to increasedproduced power of around 0.6% to 1.4% for wind speeds larger than 6 m/s.This has to be compared to the increase in thrust of around 1.0% to 1.6%.Pointing the winglet downstream increases the power production evenfurther.

The effect of sweep and cant angles is not accounted for in the presentinvestigation and could improve the winglets even more.

Contents

Introduction 4

1 Winglet design 5

2 Generation of computational mesh 7

3 Results 8

4 Discussions 16

References 16

3

Introduction

The following report is made during the EFP 2005 project “Program for Forskning iAnvendt Aeroelasticitet” under the milestone “Avanceret Rotoraerodynamik – herundertip- og rodaerodynamik”, (Contract number: ENS-33031-0077) which is carried out bythe Aeroelastic Design Group, Wind Energy department at Risø National Laboratory andthe Fluid Mechanics Section at the Technical University of Denmark.

In short, the purpose of adding a winglet to a rotor blade design is to decrease theinduced drag from the blade by changing the downwash distribution. The art is then todesign a winglet, which carries an aerodynamic load such that the vortex from thewinglet spreads out the influence of the tip vortex decreasing the downwash and reducesthe induced drag. Or in other words; to design a winglet such that the extra form drag, orprofile drag, of the winglet is smaller than the decrease in induced drag such that thetotal drag is decreased. For further reading the theory and physics of winglets on asailplane is thoroughly described in ref. i .

In sailplane applications it is most efficient to deflect the tip towards the suction side(pointing upwards) partly due to ground clearance. This leads to a lift on the wingletpointing inwards leading to a wake expansion and thereby emulating the effect of a spanincrease. In wind turbine applications it is more convenient to point the winglet towardsthe pressure side (pointing upstream) to avoid tower clearance issues. As a consequencethe lift force on the winglet will point outwards. Additionally, since the tip vortex is shedupstream of the rotor plane it will eventually interfere with the vortex sheet from theremaining blades.

The present report describes the aerodynamic investigation of a design change of the tipof a modern wind turbine blade by replacing the original tip with a winglet usingcomputational fluid dynamics. Five different winglets are investigated together with arectangular shaped tip.

Firstly, a discussion is given of the methodology of generating the blade surface mesh by

means of deflecting the outer part of the blade to an angle of approximately 90°.Secondly, the volume mesh generation is described, where special care has to be taken inorder for the computational cells to be as orthogonal as possible.

Finally, fully 3D computations have been performed on rotors and compared with thestandard rotor without winglets. Results are presented as increase/decrease in mechanicalpower and thrust as well as spanwise and cross-sectional force and moment distributionsand visualisation plots.

4

1 Winglet design

In the present work it was decided that the outer part of a modern type wind turbineblade should be modified to investigate the effect of adding a winglet. The height of thewinglet should be approximately 1.5 % rotor radius and the cant angle (see Figure 1)should be kept at approximately 90º. The sweep angle is set to 0º. Finally, to maximizethe effect of the winglet a rectangular tip was assumed.

Figure 1: Geometric quantities used to define winglet (from ref.1)

Five different winglet designs have been computed. The first winglet, winglet1, has asymmetric NACA 64-018 airfoil section and approximately 0º twist to obtain a Cl of 0.0as a reference, (green blade in Figure 2). The second winglet, winglet2, has a NACA 64-518 airfoil section and a twist of approximately -2º (positive nose down) at the tip toobtain a design Cl of approximately 0.6 (red blade in Figure 2). The third winglet,winglet3, is identical to winglet2 except that it has a tip twist of -5º instead. Based on theresults of the first three winglets (see below) two more winglets were designed. In Figure3 the two final winglets are shown. winglet4 (white blade in Figure 3) has a tip twist of+3º, while winglet5 (purple blade in Figure 3) has the same twist as winglet2, but thewinglet is bended towards the suction side instead. Finally, a rectangular shaped tip,rectangular, (not shown in the figures) without winglet, was designed for comparison.For all blades it was decided to keep the rotor radius constant.

Figure 2: winglet1 (green), winglet2 (red) and winglet3 (blue) compared to the originalblade (orange).

5

Figure 3: winglet1 (green), winglet4 (white) and winglet5 (purple) compared to theoriginal blade (orange)

The winglets were made be generating a straight blade extending 1.5 % rotor radiusfurther. Following, the 1.5% extension was deflected 90º towards either upstream ordownstream direction using a spline representation with seven control points. In this waythe final radius was kept the same as the original blade. Looking at the very tip of the

winglet it is seen that winglet2 and winglet3 are twisted at the tip compared to winglet1.Figure 4 shows the chord and twist distribution of the outer 6 % of the blades beforedeflection of the winglet. (winglet2, winglet3, winglet4 and winglet5 have the samechord distribution as winglet1 and are therefore not shown.) It is seen that the originalblade as well as the rectangular tipped blade extend to r/R = 1, while the blades withwinglets extend to r/R =1.015. It is seen that the original blade is designed with apositive twist at the tip (towards lower angles of attack) to minimize the loading.winglet2 and winglet3 are designed to increase loading and therefore have a negativetwist (towards higher angles of attack). winglet4 was designed to have the same positivetwist at the tip as the original blade. As will be shown below winglet2 had the bestoverall performance and therefore winglet5 was designed with the same twist aswinglet2, however, winglet5 is bending towards to suction side.

0.02

0.015

chord/Rotor radius [-]

0.01

0.005

originalwinglet1rectangular

00.94

0.95

Figure 4: Chord (left) and twist (right) distribution of the outer part of the blades.

0.96

0.97 0.98 0.99r/Rotor radius [-]

1

6

Figure 5 shows the rotor equipped with winglet2 to illustrate the size of the wingletcompared to the size of the rotor.

Figure 5: Rotor geometry with winglet2

In Figure 5 it is seen that the computations are made on a rotor only configurationneglecting the nacelle and tower.

2 Generation of computational mesh

The surface mesh is generated using Gridgen, a commercial mesh generator developedby Pointwise, Inc.

Figure 6: Surface mesh on winglet2.

The volume mesh away from the blade surface is made using the Risø hyperbolic meshgenerator HypGrid ii , which has shown to be more efficient for volume meshing since

7

restrictions on orthogonality and grid topology are very important. The number of cellson one blade is 256 in the chordwise direction and 128 in the spanwise direction with anadditional block of 64x64 on the tip resulting in 36.864 cells on one blade. Away fromthe surface 128 cells are used with careful control of the stretching. Since the outerboundaries should not affect the solution around the rotor these are placed approximately10 rotor diameters away. The total number of cells is 432 blocks of 323 cells resulting inapproximately 14 mio. cells.

Prologue The Object of Interest Too much Stress?Bumpy Roads Induced Drag And the Big Guys? Epilogue

Prologue

Most modern high performance flying wing aircraft with nearly elliptical lift distributions have vertical wing tip extensions, called winglets. Flying wings which use the bell shaped Horten lift distribution usually have no need for vertical fins.Winglets are also used on some conventional aircraft, as they can reduce the induced drag similar to an extension of the wing span. Compared with the span extension, winglets usually produce lower additional bending moments in the wing spar, which makes them useful for retrofitting existing wings without increasing the mass of the basic wing.

In 1992, the DLR Sailplane Symposium at Stuttgart saw a presentation from Mr. Waibel of the well known Schleicher sailplane factory. He presented interesting results from

simple experiments, using a Volkswagen Golf (aka Rabbit), equipped with a winglet. This paper shows the results of numerical studies which were conducted to apply his findings to the winglets of tailless models. It is not the aim of this paper to seek information about the optimum winglet shape and its effect on induced drag, but to understand what happens in the region where the winglet joins the wing and why the reduction in induced drag can be spoiled by additional friction drag in this region. Thus we will concentrate on the boundary layer effects.

The cover page of the 1992 «Competitors Achmer News» I found a detailed technical drawing of three tailless planes with winglets (it could also have been an explosion-sketch, as we call it in Germany, of the prototype flying wing «Delaminator») as shown in the title graphics above. The three tailless planes show interesting corners between the wing and the winglets, even an asymmetric layout, which might be optimized for left turns on the northern hemisphere. This drawing, which was very sketchy due to competition reasons, forced me into a more thorough investigation of this region of flying wing models.

The Object of Interest

For the investigation, a simple flying wing model, equipped with the MH   60 airfoil was defined and a three dimensional panel model of the configuration was created (each airfoil section was represented by 60 chordwise panels, spanwise clustering was used to refine the tip and root regions, resulting in about 600 to 800 panels per half wing, depending on the configuration examined). Only the region close to the wing tip was modified, which resulted in four different configurations:

a rounded wing tip without winglet, a winglet with a large, rounded corner, a winglet with a sharp corner, and a winglet with a sharp corner, but moved downstream 50% of the tip chord. The

maximum thickness of the winglet section is approximately located at the trailing edge of the wing tip.

All winglets had the same airfoil as the wing. The following images show the tip region of these four configurations.

Too much Stress?

The boundary layer is very sensitive with respect to the pressure rise, which occurs behind the point of maximum thickness of the airfoil. If the pressure rises too quickly (which corresponds to a steep velocity gradient), the boundary layer will transition early at best or even worse separate from the surface. Flow separation causes large additional drag.

Comparing the flow field around a single airfoil and the airfoil winglet combination, a fundamental difference becomes obvious: whereas the free airfoil feels a pressure rise in one dimension only, an additional spanwise pressure rise is imposed on the flow in the corner between winglet, as depicted below.

The external flow or pressure field is imposed on the boundary layer flow, which must overcome both pressure rise components in the winglet corner. The result is a high risk of boundary layer separation, if the winglet consists of a simple, bent up wing tip. To remedy this unfortunate situation, the winglet can be attached with an arc segment of large radius, to avoid the rectangular corner. A second way to decouple the pressure rise regions of wing and winglet is to shift the winglet downstream. By doing so, we move the front part of the winglet into the region of the pressure rise of the wing and the winglets pressure rise region is moved behind the wings trailing edge. Thus the winglets region of a favorable pressure gradient (the a region, where the flow accelerates) partially cancels out the wings pressure rise, resulting in a more favorable situation then without the winglet.

The figure below shows the velocity distributions of the wing sections close to the wing tip as well as isobar patterns on the wing itself. The same four configurations are shown.

A first impression of the flow field is given by the isobar plots in the lower part of the picture. First it can be seen, that the influence of the wing tip shape is limited to a fairly small region of the wing, maybe 10% of the span. A second observation is the fact, that all winglets hinder the spanwise flow and thus the pressure drop at the wing tip, which is seen in the first configuration, the free wing tip.

The change of the flow pattern close to the tip is also visible in the velocity distributions in a section at 99% of the span, which is shown in an enlarged view below.

Bumpy Roads

Velocity distributions close to the wingtip of the different configurations.

The graph above shows the local velocity on the surface of the wing, at the 99% span station. The upper set of lines represents the velocity on the upper surface, the curves falling more closely together are for the lower surface, which is of no interest here. Starting with the black line for the wing tip without winglets, we see that all winglet configurations raise the velocity on the upper surface, but at different places.

The winglet with the smooth, large fairing (red line) shifts the velocity distribution to higher velocities, without a large distortion. Close to the trailing edge, the velocity must reach the value of the lower side, resulting in a steeper gradient there. This configuration is not increasing the stress on the boundary layer very much, but it has a relatively large surface and is more work to build.

The winglet with the sharp corner (green) shows a quite distorted velocity distribution, with a suction peak close to the leading edge, which can lead to premature transition into turbulent flow. The following region up to 60% of the chord shows a flat pressure gradient, which favors laminar flow, bit it is followed by a steeper velocity drop towards the trailing edge, which increases the risk of flow separation.

Moving this winglet downstream (blue) shows almost no change in the first 50% of the chord and the velocity distribution up to 80% of the chord is much flatter than the original airfoil. This can lead to a larger area of laminar flow. The pressure rise towards the tailing edge is not much steeper than that of the base airfoil, because the trailing edge velocity is also raised. The result will be similar to the well rounded fairing configuration, probably even better.

Induced Drag

All the investigations above concentrated on the boundary layer and the associated friction drag. Of course the main idea behind winglets on conventional airplanes is to reduce the induced drag, and all the discussion about pressure distributions and boundary layer effects would be incomplete and questionable, if the different configurations would show very different induced drag figures. On the other hand, the main reason for the usage of winglets on tailless airplanes is not the possible reduction of the induced drag, but the search for the most effective arrangement of vertical fins to achieve directional stability - the influence on drag is an additional benefit for tailless planes.

Lift coefficient versus (induced) drag, as calculated by the panel method.

The induced drag polar above shows, that all three winglet configurations show almost identical results in term of induced drag. There seems to be a small advantage for the last configuration, but this is close to the limits of the numerical method. All configurations outperform the free wing tip model - as expected, the drag reduction is getting larger with increasing lift coefficient. The graph does not include the friction drag, which reduces the benefit of the winglets due to the additional surface. Taking friction into account, favors the classical wingtip at low lift coefficients, but the total drag of the winglet configurations will still be lower at higher lift coefficients. For the flying wing, the winglets make a vertical fin obsolete, whose friction drag would have to be added to the drag of the conventional wing tip configuration.

And the Big Guys?

If we look at full scale aircraft with large winglets, we see, that most of them have shifted the winglet downstream, either by reducing the chord length at the root of the winglet (VFW, NASA) or by moving the complete winglet downstream (Voyager). If possible, they add a curved fairing between wing and winglet. The sweep of the NASA and VFW winglets is necessary due to the higher flight Mach numbers, for a model airplane no sweep is needed (only to please your eyes). Also, the dihedral of the winglet is of minor importance as is the small forward winglet-let pointing downward.

Epilogue

We still have not reached the state of the art with our primitive efforts...

Adaptive Winglet Design, Analysis andOptimisation of the Cant Angle for EnhancedMAV Performance

Chen-Ming Kuo and Christian BollerUniversity of Saarland, Materials Science & Technology Dept.Chair of Nondestructive Testing & Quality AssuranceCampus E3.1, 66123 Saarbrücken, Germany

ABSTRACT

Adaptive air vehicle structures are an interesting option forenhancing an air vehicle’s performance. This has been shown tobe true even for MAV where a variety of solutions with regardto adaptive wings and tails have been presented so far in thepast. Within the paper to be presented here an adaptive wingletfor a MAV of 40cm in span and 25cm in length will bedescribed. This is the size of a bird where bio-inspiration hasbeen a good source for generating the adaptive winglet idea.Based on a modular MAV design and a flexible CAD modelbeing already available, the adaptive winglet with variable cantangle could be simulated, designed, realised and validated. TheCFD simulations of the MAV with winglet were done forvarious flight conditions and represented by factors such asbest aerodynamic efficiency, stability and manoeuvrability.

Optimum winglet angle search for best performance was doneby using Genetic Algorithms. In order to expand the limiteddata points without doing too much CFD simulation, a newtechnique of grey prediction (using the rolling model) has beenapplied. Results predicted with this procedure are rather closeto CFD results, with slightly less than 10% error in general andthe optimum winglet angle to be determined very much inaccordance to reality. With all the data from differentsimulations and algorithms used a working prototype with themechanism for an active adaptive winglet was realised and itsperformance shown in hardware.

formatting of papers if necessary.However, passive adaptive multiple winglet requires deepunderstanding of Fluid Structure Interaction, which is verytime consuming and requires large amount of computationalresources. In order to avoid the complexity of FSI, theproblem is simplified to active control winglet (ie, no FSIeffect).

Fig. 1 A bird’s wing during flight.[1]

Instead of having multiple winglets, one winglet oneach side of the wing is used (like conventional aircraft), andangle of the winglet can be varied. Also, this setup can bethe first step to understand the effect of the adaptive winglet.

2

MAV TEST PLATFORM

1

INTRODUCTION

The test platform MAV has 40cm wing span with lengthof 25cm, detail dimension drawing can be seen in Fig. 2.The platform is stable with long endurance, which has beenuse to demonstrate and evaluate of different adaptivestructures before [1],[2],[3]. therefore, it is very suitableplatform for the study in this paper.The MAV weights 200g (the test version), with flightspeed of 8m/s. It has unique of vector thrust propulsion unit,allows the low energy consumption for entire flight.

Most of the commercial long range aircraft has installed

winglet to decrease the induce drag to save more fuel, thisfeature can be also found on the bird. Bird use its feather atwingtip as “multiple winglet”, which can be seen Fig. 1.Each feather has different angle respect to the wing, and theyare passively adapted to the different flight conditions,which is different from the fixed angle winglet in theconventional aircraft (only designed for cruise).Such adaptive winglet feature can be studied andimplemented on the platform MAV to evaluate itsperformance and usefulness.All parts of this document have their own style (sectionheadings, subsection headings, text, captions, etc.). You canselect a style for your text, by selecting the text and thenselecting the appropriate style in the style bar (left of thefont type and size). Regular text should have the style‘Text’. The EMAV 2009 preserves the right to adjust the

Email address: chen.kuo@mx.uni-saarland.de

Fig. 2. The dimension of the MAV. Note: all units are in mm.

3

DEVELOPMENT PROCEDURES

Since the MAV is quite a new field, therefore there is no

actual research about the winglet size and angle, therefore,everything need to start from ground in this case.There are varies adaptive structure has been design onthe platform MAV; therefore, the adaptive winglet designmust be able to be independently implemented on platformand integrate with other adaptive structures. For this reason,the adaptive winglet in with integration to platform is shownin Fig. 3

Fig. 6. The aeroelasticity test of the MAV in wind tunnel. Note. The markersystem are provided by Fraunhofer ITB Germany.

For the fair test, all the mesh for CFD are kept constant inall simulation, boundary layer assume to be <15mm (by theexperience). Hybrid mesh was used for boundary layer andair around it to reduce the computational resources,, andtypically 2 million cell were in the mesh.

Fig. 3. Left, platform MAV with winglet (iso view). Right, back-view of

the MAV with winglet.

However, instead simulate all the different angle ofwinglet, only few are selected to speed up the design andanalysis process, and selected one are: -50,-30,0,20,45 and65 degree, the CAD model of each can be seen in Fig. 4

Fig. 7. Left, mesh of the flow region. Right, mesh around the MAV.

The overall design and development process blockdiagram can be seen in Fig. 8.

Fig. 4. CAD model of different winglet angle, from the top left to bottomright are:-50,-30,0,20,45,65 degree.

Because aerodynamic performance is not the only factorconsider in this paper, therefore the simulation setup mustconsider varies different flight condition in order to evaluatefactor such as stability and maneuverability [4].CFD simulation assume the flow speed of 12m/s due tothe propeller plus its own flight speed (ie the Re number is at192,000). This condition has been proved in wind tunneltest in DLR in Göttingen.Since the MAV geometry is complex, the use of K ωmodel become difficult and required high computationalresources in storing the detail of boundary layer mesh, also,the simulation of the MAV flight is mostly with-in the stallregion, therefore, standard K εmodel is selected for thisstudy. The results has also be tested in the wind tunnel inDLR for aeroealsticity, the accuracy is of Fluid solidinteraction simulation are well fit in around 5%, whichshows the reliability of the CFD model. (please note, themodel is not suitable for testing in stall region, which all theexperimental and simulation are all carried out in pre stallregion)

Fig. 8. Block diagram of the design and development process.

4

NUMERICAL METHOD

Fig. 5. Wind tunnel test model

Genetic AlgorithmGenetic algorithm (GA) is a computational method for

searching the solution of the optimal point. Traditionally,solution are represented in binary as strings of 0 and 1,therefore, the input value (analogue) in this case mustconvert to binary.The algorithm starts from some randomly generated“Parents generation”, and fitness of every individual isevaluated, multiple individuals are stochastically selectedfrom the current population, and modified (cross over andsome randomly mutated).In this case the input value is the angle of winglet, sincethe range is between 0 to 360 degree, therefore, 10 digit

binary is selected as gene. In order to make sure there isoptimal solution before end of generations, therefore, 100generations are set.Biologically, mutation only sometimes generate betteroffspring, therefore, the rate is set to be 10%, and cross rateis set to be 60%.Base on the available data point, functions’ are roughlyestimated, and then G.A uses this function to search themaximum and minimum point (aerodynamic efficiency,stability and maneuverability).

Grey predictionGrey predication is treating system as a grey system. Ingrey system theory, a dynamic model with a group ofdifferential equations called grey differential model. Thegrey derivative and grey differential equation are definedand proposed in order to build a grey model.[5]There are many different type of grey model, and commonone are GM(1,N) and GM(1,1) model, Since data predictionis required, therefore GM(1,1) model should be selected,and by definition, at least four set of data is needed; in thiscase, the input is angle of the winglet, output is the stabilityand maneuverability, therefore, 6 data point can be collected(which satisfy the condition of using grey model). Equation(1.1) and (1.2) is the solution for GM(1,1) model.

b) (0) b

x ( K + 1) = x(1) − e − ak + , aa(0)

(1)(1)ˆˆˆx( K + 1) = x( K + 1) − x( K ) [1]

R ig id W in g0 d e g re e2 0 d e g re e4 5 d e g re e6 5 d e g re e-3 0 d e g re e-5 0 d e g re e

25

30

R ig id w in g0 d e g re e2 0 d e g re e4 5 d e g re e6 5 d e g re e-3 0 d e g re e-6 5 d e g re e

0 .8

0 .9

1

R ig id w in g0 d e g re e2 0 d e g re e4 5 d e g re e6 5 d e g re e-3 0 d e g re e-5 0 d e g re e

5

RESULTS

CFD simulated resultsThe first simulated results include the winglet cant angleof 0, 20, 45, 65, -30, and

-50. Each model has go tough series of CFD simulationfor different flight conditions, results of longitudinal andlateral stability are compared with the MAV without anywinglet. All simulated results are shown from Fig. 9 to Fig.12.

1

T im e s e c

1 .5

L o n g it u d in a l s t a b ilit y o f + 2 U v e lo c it y in p u t

R ig id w in g0 d e g r e e ( W i n g le t )2 0 d e g re e4 5 d e g re e6 5 d e g re e-3 0 d e g re e-5 0 d e g re e

5

10

15T im e s e c

20

25

30

Fig. 9. Longitudinal stability results of the MAV (change in U velocity)

LongitudinalAll three different output results shows the most stableconfiguration is where the cant angle is equal to 0, and mostunstable configuration is -50. Note: these results are only

base on the simulated cases, which does not mean thedefinite results.

L o n g it u d in a l s t a b ilit y o f c h a n g e in W v e lo c it y r e s p o n s t o + 2 m / s in U

0 .3

0 .2

C h a n g e in W v e lo c it y ( m / s )

0 .1

0

-0 .1

-0 .2

-0 .3

-0 .4

0

Fig. 10. Longitudinal stability results of the MAV (change in W velocity)

L a t e r a l s t a b ilit y o f b e t a in r e s p o n s e o f + 0 . 0 8 7 ( 5 d e g r e e ) in s id e s lip

0 .1

0 .0 8

C h n a g e in b e t a ( r a d )

0 .0 6

0 .0 4

0 .0 2

0

-0 .0 2

0

0 .1

Fig. 11. Lateral stability results of the MAV (change in β)

2 .5

2

C h a n g e in r ( r a d / s )

1 .5

1

0 .5

0

-0 .5

0

Fig. 12. Lateral stability results of the MAV (change in r)

LaterallySimulated results in all cases shows the additionalwinglet cases damping ratio to drop, and MAV response togust faster (but with overshoot). This is very logic, sincewinglet (doesn’t matter which angle it is at) createsadditional surfaces area in XZ plan (body coordinate systemof the aircraft), and these additional surfaces cases additionalforces and moment to be generated during lateraldisturbance, in theory this should increase the speed ofreaction, however, because the winglet in this experimentwas located slightly in front of the C.G (due to theZimmerman profile and vector thrust propulsion unit),therefore, the additional forces is resulting decreasing theoverall damping ratio of the motion (speed up the response).This results also agree with Corneil’s work in early 80s [6].However, the with this increase in the response speed togust, means decrease maneuverability, and from Fig. 11 andFig. 12 shows the best maneuverability occur with 65winglet configuration, where the response is slowest (if

5

10

15T im e s e c

20

0 .2

0 .3

0 .4

L a t e r a l s t a b ilit y o f r in r e s p o n s e o f + 0 . 0 8 7 b e t a ( 5 d e g r e e )

0 .5T im e s e c

0 .6

0 .7

0 .5

consider MAV with winglet configuration only).Genetic algorithm search for best performance of thewinglet angleSince data obtain from the CFD simulation are limited,and with these limited data, one can only identify the bestresults from select cases.However, the best results may not be in those exact cases(ie, best longitudinal stability may occur in 32 degree, whichwas not the simulated case). In order to solve this problem,genetic algorithm model was written for search the bestaerodynamic performance, stability and maneuverability.The research results shows from Fig. 13 to Fig. 15, andcompiled results table is shown at Table 1.

G e n e t ic A lg o r it h m s e a r c h fo r b e s t C L / C D

Table 1. Results of G.M search for best and worst performance of thevariable angle winglet angle. Note: min stability represent bestmaneuverability.

The results from the G.A clearly shows some data thatwas not visible by just looking at selected CFD cases, and

these information can be used for further autopilot designdata base or pilot.

Using combination of Grey prediction and Geneticalgorithm for searching best performanceG.A model is base on the already know data point andfunctions to search the best results, however, more data pointit has, more accurate the function can be, hence, moreaccurate the G.A results will be.As mentioned before, Grey prediction (G.P) uses thenumbers of exciting data to predict the next set of data,which is completed different to G.A. In this project, rollingmodel method is used, and results is shown at table 2

Longitudinal dampingLateral damping ratioratio-500.01760.53689-300.0240.64800.13780.5594200.090.5275450.07870.52334650.094110.507812850.10740.47611050.11630.45541250.09950.44301450.0983

0.41071650.09700.39071850.09580.3717Table 2. The results of grey prediction of the damping ratio for bothlongitudinal and lateral motion. Note: black shows CFD results, and orangeshows the G.P value. Note: 180 degree is not the same as 0 degree, seefigure.2 for the orientation angle of the winglet.

Angle of winglet

50G e r n e r a t io n s

60

70

80

90

100

Fig. 13 G.M search of minimum value of the function for CL/CD

20

30

40

G e n e t ic a lg o r it h m s e a r c h fo r b e s t lo n g it u d in a l s t a b ilit y

20

30

40

50G e n e ra tio n s

60

70

80

90

Results from the G.P are later used in the G.A modeling.Even the data now is bigger; solution still converged beforeit gets to the default number of generation (100 in thisproject), results are shown in Fig. 16,Fig. 17 and compiledresults in Table 3.Even though the longitudinal results from Table 3 andTable 1 shows very similar, there is still have some slightlyvariation. However, the story is completed different in thelateral mode; there is 5 degree different for the max stability,but completely different story about the minimum stability.The main reason for such different is because before G.Pmodeling, the data point is only up to 65 degree, andaccording to the data collected, which is not enough for G.Mto solve the case, therefore the search stop at 65 degree.With G.P model, the data expanse to 185 degree, and withsufficient data point, much better G.A search can bepreformed.

100

Fig. 14 G.M search of minimum value of the function for Longitudinalstability (base on the damping ratio of the motion)

G e n e t ic a lg o r it h m s e a r c h fo r t h e b e s t la t e r a l s t a b ilit y

m ix t u r e o f G r e y P r e d ic t io n a n d G e n t ic A lg o r it h m fo r lo n g it u d in a l m o t io n

20

30

40

50G e n e ra tio n s

60

70

80

90

Fig. 15. G.M search of minimum value of the function for lateral stability(base on the damping ratio of the motion)

CL/CD

35.6452

-41.16

Longitudinalstability0

-42

Lateral stability

-39.638

65

7 .6 2 5 8

100

7 .6 2 5 8

m in v a lu e o f t h e f u n c t io n

7 .6 2 5 8

7 .6 2 5 8

7 .6 2 5 8

7 .6 2 5 8

7 .6 2 5 8

7 .6 2 5 8

0

10

20

30

40

50

60

g e n e r a t io n

70

80

90

100

Fig. 16 Mixture of G.P and G.M search of minimum value of the functionfor longitudinal stability (base on the damping ratio of the motion)

m ix t u r e o f G r e y P r e d ic t io n a n d G e n t ic A lg o r it h m fo r la t e r a l m o t io n

20

30

40

5060G e n e r a tio n

70

80

90

maneuverability occur at the 187 degree. The reason for thiseffect is also to do with where the lift location and value.When lift is closer toward to the body, the moment for thelateral motion decreased when the same disturbance occurs,therefore the MAV resist the change.All the results in the study indicate that all the bestperformance of MAV occur at different angle of winglet,therefore, for the in order to achieve most efficient flight,active control of winglet is required, and results from thisstudy can be used for future autopilot design.One other discovery in this study was that even useasymmetric variable angle winglet for maneuver is notefficient enough. Consider MAV with left side of wing haswinglet of 90 degree (flat), and the right side of 0 degree,then the lift one the left is higher, which cause the rollingmoment to the right, but side down wash on the right side islower (because of the winglet), this will make the yawingmoment to the left, therefore it is in conflict with rollingmoment created.Further more, when MAV is turn left with thisconfiguration, the right side winglet is acting like verticaltail, which will generate side force to the right to resist therolling moment. These motion can be seen in Fig. 19, andwas also confirm with Bourdin’s [9] and Corneil’s [6] work.

100

Fig. 17. Mixture of G.P and G.M search of minimum value of the functionfor lateral stability (base on the damping ratio of the motion)

Longitudinal stabilityLateral stabilityMax-3.0158-34.3046Min-43185.1678Table 3. Results of mixture of G.P and G.M search for best and worstperformance of the variable angle winglet angle. Note: min stabilityrepresent best maneuverability.

Aerodynamically, the best efficient shows at 35.65degree, which if recall Fig. 1, bird’s winglet angle is not at

90 degree during the steady flight. In fact, when eagle isduring the steady flight, the winglet angle is around 30 to 40degree [7], which is in the range results obtain from thisproject.Looking stability/maneuverability, longitudinally, thebest stability at 0 degree and best maneuverability is atround -42 to -43 degree. Bird’s winglet can not go tonegative degree, since it is only control passively, but ifconsider the shape of wing during when bird during dive toattack on its target, the wing is rather at “M” shape [8]. ThisM shape decreases the lift coefficient, and vortex centermove more toward to the body side (where the weight is),this similar effect can be seen when the winglet is at negativeangle (Fig. 18), and therefore, it makes the bird and MAVeasier to move in the longitudinal direction.

Fig. 19. Asymmertic winglet MAV during the turn. (front view at right)

6

HARDWARE PROTOTYPING

Since the individual moment of winglet does not haveany benefit (can see from Fig. 19), therefore, 2 winglet mustbe able to varies its angle together. For this requirement,only one servo is needed.The condition is very similar to the Ref[1], thereforeonly small modification is needed form the original design.

Fig. 20. The deflection of the winglet on the MAV. Note the mechanicallinkage.

7

CONCLUSION

Fig. 18. The stream line of MAV with different angle of winglet. Note: thelocation of the vortex center is clearly different.

Laterally, results of CFD simulation shows theconfiguration without winglet has best stability, reason hasbeen discussed before. However, if looking at wingletconfiguration, the best stability occurs at -34 degree from theG.A, and -39 degree with G.P & G.A modeling. Max

Study in this project shows the how the with a simple

winglet configuration, the overall performance ofaerodynamic efficiency, stability and maneuverability can bechanged. Optimal angle of aerodynamic, stability andmaneuverability performance has been identify with useeither CFD data only, G.A only, and mixture of G.P and

Traditional design and analysis process is timeconsuming, however, using limited data point from CFD,using mixture of G.P rolling model and G.A generating moredata point, and search the optimal point is fast and moreefficient.Optimal angle for aerodynamic performance is around35 degree. Longitudinally, most stable at 0 to -5 degree, andmost maneuverable at -42 to -43 degree. Laterally, moststable at -34 to-39 degree, and most maneuverable at 185degree. These results pattern also confirm with the naturalbird’s flight.

8

FUTURE WORK

This paper shows the optimal angle for MAVperformance, these data can be given to the autopilot designor flight programming.Second step, with optimal angle and aerodynamic dataknown for the max CL/CD, therefore, passive wingletdevelopment can be possible.Lastly, in order to get even more bio-inspired, bothactive and passive multiple wing let can be design anddevelop for the future.