See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309209815

HIGH PERFORMANCE SAILPLANE DESIGN STRATEGY USING INVERSE DESIGN

AND OPTIMIZATION TECHNIQUES

Conference Paper · April 2007

CITATION READS

1 4,858

1 author:

Krzysztof Kubrynski
Instytut Techniczny Wojsk Lotniczych
13 PUBLICATIONS 45 CITATIONS

SEE PROFILE

All content following this page was uploaded by Krzysztof Kubrynski on 18 October 2016.

The user has requested enhancement of the downloaded file.

Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

HIGH PERFORMANCE SAILPLANE DESIGN STRATEGY USING INVERSE DESIGN AND

OPTIMIZATION TECHNIQUES

Krzysztof Kubrynski
Institute of Aeronautics and Applied Mechanics
Warsaw University of Technology
Warsaw, Poland
[email protected]

ABSTRACT devices, as speed flaps or winglets). On the other hand the

The Paper presents process of aerodynamic main objective of modern sailplane design is the
development of the new wing for high performance racing maximization of overall performance that can be measured
class sailplane Diana-2 (based on the previous over 15- by cross-country speed at specified thermal conditions, that
years old SZD-56 Diana design). The main objective was depends not only on low drag or high lift to drag ratio of a
improving performance measured by increased average plane. To address and study of such a problem, a
cross-country speed. The design process consists of mathematical model of cross-country flight must be
analyses of cross-country flight problem, optimization of established and used to optimize/design of the sailplane
planform and wing sections, and finally removing adverse design parameters. Such an approach must take into
interference effects. Optimization and inverse design account the thermal model, sailplane aerodynamic
methods were the main tools in design process. characteristics (dependent on various design parameters
and “aerodynamic technology”), and mass (water ballast
INTRODUCTION amount). The problem is very complicated (especially
Good overall performance of modern flying vehicles weather and thermal conditions), so only a very simplified
depends on various design parameters, including basic approach can be used.
design concept, internall solution of a strength structure, The SZD-56 Diana sailplane was designed in 1989.
materials used, aerodynamic efficiency, flight mechanics Its first flight was in 1990. In respect to structure and
etc. All these problems must be solved in the most efficient technology the sailplane was (and it still is) the most
manner to assure success. Even most revolutionary advanced one. It has many unusual features, e.g. very light
solution in one of these fields does not lead to good overall structure (175 kg empty weight vs. typical 230-240 kg),
characteristics. Of course aerodynamic efficiency is very very thin (13%) and high aspect ratio (27.6) unusuall,
important among all these subjects. monocoque wing. Unaffected by a massive spar, the '56
Computational methods of fluid dynamics play an wings hold their precise, unwavering contours year after
important and growing role in aerodynamic development year. Additionally, the spar-less design leaves extra room
of modern flying vehicles, especially at high speeds. Low for ballast: the '56's wet wings hold 160 liters. In the late
speed aerodynamics also more and more frequently uses 80's and early 90's, glider aerodynamics progressed quickly
advanced design tools, significantly improving final thanks to both extensive wind tunnel experimental work
performance. Ones of the most interesting examples are and the development of advanced computational methods
gliders. Modern high performance gliders reaches very in fluid dynamics. The designer of Diana did not have
high level of aerodynamic efficiency measured by lift to these kinds of resources at that time. Other gliders built 3
drag ratio at entire range of speeds. Maximum L/D reaches to 6 years after the Diana had a chance to use all these new
value over 60, which means glide angle below 1O. Thanks tools and ideas, which allow for reducing the profile drag
to modern aerodynamic concepts and advanced and lowering adverse aerodynamic inference. Because of
computational methods minimum drag of nowadays these advantages, these newer gliders achieved much better
sailplane airfoils have reached value well below 0.004 at performance. Consequently, Diana, one of the most
Reynolds number about 3mln. Farther improvement of technologically advanced sailplane but equipped with less
aerodynamic efficiency is nearly impossible without using modern aerodynamics (a Wortmann type, based on 70’s
new aerodynamic concepts (such as active boundary layer technology, free transition type airfoil), fell into
control) which must be costly and rather impractical for background. In 2003 the decision was made to design new,
widespread operating. But there is still some space for advanced aerodynamics for Diana. In the case of high
improvement of overall performance by careful design and performance racing class sailplanes, the wing produces
optimization of all elements (including some assistant most of the drag. Because of this fact and the expected
Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

costs, modernization was restricted mainly to the design of created. Three thermal families are specified and studied:
a new wing. A:narrow, B:wide, C:middle thermal. The rate of climb for
Design of the new aerodynamics took a few stages a particular thermal shape, specified sailplane mass, and
subseqently shortly described in the paper: the aerodynamic characteristics can be specified as
difference between thermal lift and sailplane sink velocity
- cross-country flight performance analysis, ( wC = wT − w SINK ), where the thermal lift at a specified
including investigations of various design and circling radius and sink velocity depend on sailplane speed
aerodynamic features on final flight efficiency and bank angle in circling, mass and aerodynamic
- optimization of wing planform characteristics, and can be expressed as:
- optimization/design of wing two-dimensional
sections for various span stations
- removing of adverse aerodynamic interference wT = wT (THERMAL _ MODEL , wT 0 , RCIRC ( VCIRC , ϕ CIRC ))
effects (1)
wSINK = wSINK (VCIRC , ϕ CIRC , MASS , AERODYNAMICS )

CROSS-COUNTRY AND CIRCLING

PERFORMANCE ANALYSES Finally equation for altitude balance can be written:
A typical cross-country flight pattern is shown in figure
1. It includes circling in a thermal, interthermal descent, (2)
and cloud streets flight. Thermal strength (including (wT − wSINK ) ⋅ tCIRC = (wSINK + wd ) ⋅ L − LCS + (wSINK − wCS ) ⋅ LCS
Vd VCS
diameter and vertical velocity distribution), interthermal
descent strength rd (expressed as a fraction of max. thermal
strength), and the cloud street’s relative length rlcs and and the average cross-country speed can be written down
strength rcs (expressed as rd) are the parameters describing in the form:
thermal conditions. The specification of thermal
characteristics is the most challenging problem as they V CC = f (THERMAL _ MODEL , w T 0 , rd , rCS , rl CS , (3)
depend on a great number of factors (weather conditions, MASS , AERODYNAMI CS , V CIRC , ϕ CIRC , V d , V CS )
geographical region, ground features, altitude, time, etc.)
and are in principle governed by stochastic rules. A very
limited amount of measured data is available in the It is seen that final average cross-country speed depends on
literature, making this problem even more difficult. the thermal model, thermal strength wT0 , sailplane mass,
aerodynamic characteristics, circling speed and bank angle
and, additionally, on the weather model (rd, rlCS, rCS),
speed at interthermal glide, and along cloud streets. For a
specified weather model, thermal strength, mass, and
aerodynamic characteristics of the plane, the optimum
flight and circling parameters, as well as final average
cross-country speed, can be easily found using nonlinear
programming methods. Such modeling of gliding tactics is
equivalent to that of MacCready[2], which is not the one
actually applied by pilots due to relatively low value of
probability of arrival. Advanced mathematical analysis
shows, however, that reasonable modifications to the
MacCready rules have little influence on the final average
cross-country speed – increasing significantly the
probability of arrival[3,4]. This justifies the treatment
applied in the present work. Detailes of the analzes can be
found in[5]. Sample results for final design and one of the
weather with middle thermal is presented on the Fig. 2. It
Fig. 1 Assumed cross-country flight pattern. shows that higher max. lift coefficient at circling have
large influence on climb efficiency and significantly
improve final cross-country speed. In the same menner
Horstmann[1,2] models are probably the most realistic also other sailplane design parameters can be studied and
approach. The direct application of those models to the optimized. Compared to the old sailplane version the new
current problem is, unfortunately, not possible as one has slightly larger wing (but still have highest aspect
continuous spectrum of thermal strength is required. A ratio among all 15-m span sailplanes) and winglet that
continuous family of thermals based on the assumption of allow ro reduce induced drag by about 4%.
a linear interpolation between the Horstmann models was
Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

Fig. 3. Sketch of airfoil behavior at angle

of attack changes due to vertical gust.
Fig. 2 Influence of circling lift coefficient and thermal
strength on climb rate and average cross-country speed,
middle thermal (C). The very desirable feature of the sailplane’s airfoils
besides low drag should be: high maximum lift coefficient
and monotone lift vs. angle of attack characteristics up to
BASIC AIRFOIL DESIGN CLmax. Experience with airfoil design methodology
Modern, low-drag glider airfoils have specific features applying inverse boundary layer method [7] as well as
[6]. The lift characteristics at higher angles of attack have a utilization of optimization [8] and inverse design [9]
local decrease of lift with increasing angles of attack (has a procedures in developing two-dimensional airfoils allow
local minimum). The reason of such features is the abrupt designing airfoils with desired features. On the Fig. 4
forward movement of transition point along the upper computational characteristics of a typical modern sailplane
surface and the thickening of boundary layer on a highly airfoil and one of the airfoils designed for the new Diana
laminarized airfoils. According to an unpublished paper wing at two flap angles are presented. It is seen that
presented at OSTIV Congress in 2003 by A. Dushyn and besides the slightly lower minimum drag, much higher
L. L. M. Boermans (“Sailplane climb performance in maximum lift coefficient and monotone lift characteristics
thermals due to dynamic effects”), this local minimum in the new airfoil is much more forgiving for rain or bugs on
the lift curve can have significant influence on sailplane’s the surface (simulated by forced transition near the leading
behavior during entering thermal and flying in a turbulent edge - that leads to increase drag and lowering lift)
thermal. The situation is explained schematically in Fig.3. compared to existing airfoils. Such features are possible
If an angle of attack during circling is near local maximum thanks different design pressure distribution on the airfoil.
of CL (and close to the upper limit of the low drag bucket) Generally speaking the new airfoil is thinner (thinnest
a downward gust (one that decreases the angle of attack) among all 15-m span sailplanes), has larger leading edge
leads to some loss of lift and some loss of altitude. An radius, lower pressure gradients in the pressure recovery
upward gust (one that increases the angle of attack) leads region and slower movement of boundary layer transition
to additional lift in the case of a monotonic lift-angle of point on the upper surface with increasing angle of attack.
attack relationship, and an increase in speed and climb rate. Costs of such solution are: slightly lower level of laminar
However, in the case when the lift curve has a local boundary layer stability (result in narrower low drag
minimum as noted, an increase in angle of attack leads to a bucket) and higher sensitivity for proper performance flap
loss of lift and altitude. settings.
A variation of angle of attack due to a gust during Comparison of the geometry and computational
circling can reach up to 5 deg. Thus, there can be a characteristics of designed basic wing airfoil and the old
significant problem in circling at such a value of the lift one is presented on the Fig. 5. The new airfoil is slightly
coefficient due to both the loss of climb efficiency and the thinner (but have much better strength and stiffness
danger of stall. This means that circling at lower CL is characteristics). Laminar flow extend to 65-75% of upper
necessary for efficiency in case of an airfoil with lift surface (depending on flight conditions) and 92% of lower
characteristics containing local minimum. surface (where transition is forced by means of pneumatic
Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

turbulators in order to prevent laminar separation). Airfoil

drag is reduced by about 20÷25% along entire range of lift
coefficients compared to the old one.

a)

b)

Fig. 5 Geometry and drag envelope of the old and new

Diana airfoil.

Planform Specification
Wing planform should ensure best possible wing
characteristics including low induced drag (optimum load
distribution), high maximum lift coefficient and proper
stall characteristics (with no tendency to wing drop). It is
worth mentioned that real observations and flight tests does
not fully confirm classical considerations regards optimum
planform [11,12] due to nonlinear wing-tip flow features
and three-dimensional flow effects. Especially important
observation [11] is that wing planform and winglet can
Fig. 4 Computational characteristics of Diana-2 airfoil and signifficantly affect induced drag (more then classical
typical modern airfoil – clean and wet conditions finite wing theory predicts) and high lift and stall
simulated, a) no flaps, b) flaps down. characteristics.
The optimum load distribution along the wing/winglet
span was determined by applying third Munk’s theorem
WING DESIGN [13]. The optimum wing planform (more precise: local
Basic design procedure for the wing-winglet chord value) minimizing induced and profile drag is
combination consist of several stages[10]: planform possible to evaluate by division of the above optimum load
optimization, specification of flow conditions (local by optimum sectional lift coefficient. Such airfoil could
Reynolds number and sectional lift coefficients) at various minimize both: induced and profile drag. Modification to
wing stations and flight conditions (including speed, flap such a planform is undertaken in order to achieve proper
setting and sailplane mass variation), optimization of two- stall progression along the span by decreasing slightly
dimensional airfoils for each of wing sections (and local sectional lift coefficients in the outer wing stations (by
flow conditions), building three-dimensional wing about 2% at the end of the planar part of the wing).
(combining optimum planform and 2D airfoils), and finally Trailing edge line is nearly unswept in order to minimize
removing adverse interference effects via inverse design/ span-wise pressure gradients and prevent boundary layer
optimization process. washout effect at high angles of attack.
Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

All the design parameters, such as planform and final two-dimensional characteristics). Sample
profiles at subsequent span stations are, of course, aerodynamic characteristics and airfoil geometry designed
subsequently updated to achieve the properties dictated by for flow conditions from figure 7 are presented on figure 8.
the iterative design process. The final planform and wing
sections were obtained as a result of using three-
dimensional optimization methods during detailed
aerodynamic design and are seen on Fig. 6.

Fig. 6 Final sailplane wing planform (developed

surface).

Design of Optimized Wing Airfoils

Analyses of load distribution on the wing having
specified planform at various flight conditions (wing CL,
flap settings, side-slip angle) allows to find local sectional
lift coefficients and corresponding Reynolds number at
each of wing station. On the Figure.5 example dependency Fig. 8 Geometry and drag characteristics of the
of local lift coefficient vs. Reynolds number for one of the designed two-dimensional airfoil
wing station (in the wing-winglet intersection) at various
flight conditions (speeds, flap angles and sailplane mass) is Closed funicular polygon express conditions
presented. Filled region between lines express all possible corresponding to filled region on the figure 7. Ends of each
conditions that can occur in the level flight and for which of drag polar correspond to sideslip angle +/- 3O. It is seen
airfoil should be optimized. Upper and lower lines that in entire envelope of the level flight conditions the
correspond to sideslip conditions (+/- 3O). airfoil works at nearly minimum drag. Such a design
procedure is applied for various wing and winglet sections
along the span. Figure 9 presents optimized airfoils family
for subsequent sing and winglet section.

Fig. 9 Geometry of airfoils designed for various span

stations.
Aerodynamic Interference
Fig. 7 Dependency CL vs. Re at various flight conditions Iniotial geometry of the wing is obtained by applying
for one of the wing station. two-dimensional optimized airfoils on the optimized
planform. Due to aerodynamic interference, mainly in the
wing-winglet concave corner and near fuselage junction,
Designing of the two-dimensional airfoil for required pressure distributions (and as a result boundary layer
flow conditions is performed by mixed procedure. Initially development and aerodynamic characteristics) differs from
multipoint optimization procedure [8] is applied that allow the values on the optimized (two-dimensional) airfoils.
to find geometry and corresponding pressure distribution Figure 10 pressents inviscid pressure distribution in one of
that minimize weighted sum of drag at specified design the design points of airfoil presented on the Figure 8. Blue
points. Optimized airfoil contour and pressure distributions line corresponds to two-dimensional flow, red and green
are usually not smooth, so inverse design procedure [9] is after putting airfoil on the three-dimensioanl wing (in the
used to define final profile (including manual checking out
Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

wing-winglet region). Two dimensional flow over the surface at higher lift coefficients. Additionally it can
airfoil have some special features: completely laminar flow produce flow separation in the rear part of wing-fuselage
on the lower surface in all design points, about 75% intersection, significantly increasing drag. At a final design
laminar flow on the upper surface that follows transition stage three-dimensioanl inverse design/optimization
with very week separation bubble. method[15] is used in order to enforce pressure distribution
closesd to the two-dimensioanl design pressures on the
optimized airfoils. This is equivalent to removing adverse
interference effects and leading finally to the so called
design with neutral interference. The design method allows
to find modification of the external geometry that
producesthe pressure distributions closesd to the target at a
few design points (design angles of attack and
configurations: flaps/controls deflections). Additionally,
induced drag can be minimized at given lift and moment
coefficients (trim conditions), and some additional
geometrical constraints can be enforced. The following
objective function is minimized:

NDP
E= ∑ WP
n =1
n ⋅ E P _ n + WC ⋅ E C + WR ⋅ E R +
NDP
Fig. 10 nviscid pressure distribution on the isolated airfoil
and on the wing ∑ WD
n =1
n ⋅ (C D _ ind ) n +
NDP NDP

∑λ
n =1
L
n • (C L − C LD ) n + ∑λ
n =1
M
n • (C M − C MD ) n

where: n is design point number, NDP is the number of

design points, WP, WD, WR, WD, are weight factors for
pressure deviation norm, geometry constraints penalty
function, regularity term and induced drag respectively. EPn,
EC, ER are: pressure distribution deviation norm (target-
actual) at n-th design point, geometry constraints penalty
function and regularity term.
As a result, it is possible to determine the geometry
(surface shape, wing twist, angles of attack, flaps/controls
deflections) that produces pressure distributions closest to
those specified at some angles of attack (or lift coefficients)
and sailplane configurations, minimizes induced drag at
specified lift and moment coefficients (trim conditions) and
anforces some geometrical constraints and regular (smooth)
Fig. 11 Calculated pressure distribution on the airfoil and surface. Flow analysis is performed by a higher order panel
on the wing, boundary layer included. method.
Design pressure distributions are taken from the two-
Flow in the specified section of the three-dimensional wing dimensional airfoils, with some modifications near wing-
produce laminar separation near trailing edge on the lower fuselage intersection (lowering pressure gradients in the
surface and relatively strong laminar separation bubble on pressure recovery region to prevent separation of the
the upper wing surface. Calculated pressure distribution in turbulent boundary layer). Inviscid pressure distribution in
the case of viscous flow is seen on the fig. 11. Expected the section considered earlier (Fig. 10) after designing is
profile drag in this section increases by about 30% due to presented on the Fig.12.
above problems. Geometry is slightly modified but pressure distribution
Even more difficult problems occure in the wing-fuselage (and expected boundary layer development and flow
intersection region[14], that can lead to lost of laminar features) should be similar to that of two-dimensional
flow on the lower surface at high speeds and on the upper optimized airfoil (e.g. strong separation bubbles removed).
Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

nature with subseqent modifications of assumed geometry

parameters and some characteristics and flow features.
Aerodynamic design process of of Diana-2 sailplane had
16 iterative loops. Final result allows to expact maximum
glide ratio over near mythical (for 15-m span racing
sailplanes) value 50, excelent circling and stall
characteristics and very good speed polar. Flight tests
confirm expectations. Maximum lift coefficient obtained in
flight is about 1.7 – unusually high for this class of planes,
stall is very gently with no tendency to wing drop and
sailplane retains full control at stall. Taking into account
additionally very low empty weight (182 kg) and unusually
hevy water ballast (248 kg) it should lead to very good
high-performance sailplane – Fig. 14.

Fig. 12 Inviscid pressure distribution in the wing section

after inverse design.

As example Fig. 13 presents expected extend of laminar

flow (blue color) on the lower wing surface at maximum
sailplane mass (m/S=57.6 kg/m2), performance flaps angle
0O and speed 200 km/h.

Fig. 14 Designed sailplane with nonlinear planform of

optimized wing.

Finally, as an addition the the aerodynamic design,

determination of the best flap position (depending on speed
and mass) and optimization of flight tactical choices for
various weather conditions were performed [5] in order to
help the pilot to take best flight decisions (mainly regards
amount of water balast for specific weather conditions) and
take full advantage of the new plane. The final project
Fig. 13 Expected range of laminar flow (blue surface), quality is best judged by results in competitions. Pilots
m/S=57.6 kg/m2, βFL=0O, V=200 km/h. flying prototype of Diana-2 have took (in totally five
competitions) four times first place (in this: First World
Gliding Championships on Grand-Prix rules –2005 and
CONCLUDING REMARKS World Gliding Championships -2006) and one second
Real design process have took into account much place – that confirm efficiency of the presented design
larger number of design aspects, not only of aerodynamic strategy.
origin. Pure aerodynamic aspects were also considered
much more detailed during design process. As example
winglet size, geometry of wing-winglet junction (radius of REFERENCES
intersection), horizontal tail settings, optimization of center 1. K.H. Horstmann, „Neue Modellaufwindvertei-lungen
of gravity location, removing spanwise pressure gradients und ihr Einfluss auf die Auslegung von Segelflugzeugen“
near trailing edge of the wing, lowering tendency to cross- OSTIV Publication XIV (1976)
flow instability in the fuselage boundary layer by chosing
wing setting angle, etc. Real design process is iterative in 2. F. Thomas, Fundamentals of Sailplane Design, College
Park Press (1999)
 Inverse Problems, Design and Optimization Symposium
Miami, Florida,U.S.A., April 16-18, 2007

3. J.H. Cochrane, “MacCready Theory with Uncertain Lift

and Limited Altitude”, Technical Soaring 23(3) (July
1999), also:
http://www.gsb.uchicago.edu/fac/john.cochrane/research/Soaring

4. R. Almgren, A. Tourin, “The Mathematics of Gliding

Racing,” RCA, Nov. 7, 2004

5. K. Kubrynski, “Aerodynamic design and cross-country

flight performance analysis of Diana-2 sailplane” OSTIV
Congress, Eskilstuna 2006 (to be published in Technical
Soaring).

6. L.M.M. Boermans, A. van Garrel, “Design and

Windtunnel Test Results of a Flapped Laminar Flow
Airfoil for High-Performance Sailplane Applications” –
ICAS 94-5.4.3

7. K. Kubrynski, “Inverse Treatment of Design Problems

in Low Speed Aerodynamics”, Fifth World Congress on
Computational Mechanics, Eds. H.A.Mang,
F.G.Rammerstorfer, J.Eberhardsteiner, Vienna (2002)

8. M. Drela, “Design and Optimization method for

multielement airfoils” AIAA Paper 93-0969, Feb. 1993.

9. M. Drela, “XFOIL: An Analysis and Design System for

Low Reynolds Number Airfoils”, in Low Reynolds
Number Aerodynamics, Ed. T. J. Mueller, Lecture Notes in
Eng. 54 (1989)

10. K. Kubrynski, “Wing-Winglet Design Methodology for

Low Speed Applications”, AIAA Pap. 03-0215, 41’th
Aerospace Science Meeting and Exhibit, Reno (2003)

11. W. Schuemann, “A New Wing Planform with

Improved Low-Speed Performance” – Soaring Vol. 47,
No. 2, Feb. 1983

12. S.F. Hoerner, „Fluid Dynamic Drag”, Washington

1958

13. M.M. Munk, “The Minimum Induced Drag of Airfoils”

NACA R 121, 1921

14. L.M.M. Boermans, K. Kubrynski, F. Nicolosi, “Wing-

Fuselage Design of High-Performance Sailplanes” –
Boundary Layer Separation in Aircraft Aerodynamics, ed.
Henkes R.A.W.M., Bakker P.G. – Delft (1997)

15. K. Kubrynski, “Subsonic Aerodynamic Design Via

Optimization”, Ed. K.Fuji, G.S.Dulikravich, Notes on
Numerical Fluid Mechanics, Vol. 68, Vieweg (1999)

View publication stats