File 422

LETECK
zpravodaj
ISSN 1211877X
In this issue:
Axis-Symmetrical 3D
Viscous Turbulent
Flow with (k,)
Standard Turbulent
Model Written to the
Cylindrical System
VUT 001 Marabu
Fuselage Design
Methodology
FE-Modelling
Methodology of
Riveted Joints
Incremental
STAIRCASE Test
Method
Modal Parameters
Identification of
a Structure under
Operational
Conditions
Calibration Methods
of Force and
Displacement
Transducers
Wind Tunnel Tests
of Ice Accretion
CZECH
CZECH
AEROSPACE
P r o c e e d i n g s
No. 2 / 2010
J
u
n
e

2
0
1
0
J
O
U
R
N
A
L

F
O
R

C
Z
E
C
H

A
E
R
O
S
P
A
C
E
R
E
S
E
A
R
C
H
VZKUMN A ZKUEBN LETECK STAV, a.s.
Editorial address: VZL / Aeronautical Research and Test Institute
Beranovch 130, 199 05 Prague - Letany
Czech Republic
Phone.: +420-225 115 223, Fax: +420-286 920 518
Editor-in-Chief: Ladislav Vymtal (e-mail: vymetal@vzlu.cz)
Editor & Litho: Stanislav Dudek (e-mail: dudek@vzlu.cz)
Editorial Board:
Chairman: Josef Kapar, Vice-president ALV, General Director VZL
Vice-Chairman Jan Barto, ALV
Members: Tom Blohradsk, Vladimr Dank, Lubo Janko,
Pavel Kuera, Petr Kudrna, Zdenk Ptek, Antonn Ptk
Publisher: ALV / Association of the Aviation Manufacturers
Beranovch 130, 199 05 Prague 9 - Letany
Czech Republic
ID. No. 65991303
in cooperation with the Czech Technological Platform for the Aviation
and Space.
Printing: Studio Winter Ltd. Prague
Periodicity: Three times per year. Press Reg. No. MK R E 18312.
Subscription and ordering information available at the editorial address. Legal liability for published
manuscripts originality holds the author. Manuscripts contributed are not returned automatically to
authors unless otherwise agreed. Notes and rules for the authors are publi shed at http://www.vzlu.cz/.
Czech AEROSPACE Proceedings
Leteck zpravodaj
2/2010
2010 ALV /Association of the Aviation Manufacturers, All rights reserved. No part of this publication
may be translated, reproduced, stored in a retrieval system or transmitted in any form or by any other means,
electronic, mechanical, photocopying, recording or otherwise without prior permission of the publisher.
ISSN 1211 - 877X
J O U R N A L F O R C Z E C H A E R O S P A C E R E S E A R C H
CZECH
CZECH
A E R O S P A C E
P r o c e e d i n g s
EUROPEAN UNION
EUROPEAN REGIONAL DEVELOPMENT FUND
INVESTMENT IN YOUR FUTURE
The Czech Aerospace Proceedings reflects the achievements in research and development
projects and areas in compliance with the Strategic Research Agenda prepared
and published by the Czech Technological Platform for the Aviation and Space.
Contents / Obsah
VUT 001 Marabu Fuselage Design Methodology
Metodologie nvrhu trupu letounu VUT 100 Marabu
Ing. Michal Mali, PhD, Ing. Tom Urk, Ing. Petr Doupnk, Ing. Ji Zblatzk / Institute of Aerospace
Engineering, Brno University of Technology
Axis-Symmetrical 3D Viscous Turbulent Flow with (k , ) Standard Turbulent Model Written to the
Cylindrical System
Formulace (k , ) standard turbulentnho modelu pro 3D osov symetrick vazk turbulentn proudn
v cylindrickm systmu
RNDr. Jaroslav Pelant, CSc., RNDr. Martin Kyncl / VZL, Plc., Prague
FE-Modelling Methodology of Riveted Joints
Metodika tvorby MKP modelu ntovho spoje
Ing. Jaromr Frek, Evektor, spol. s r.o., Kunovice
NTA/G-update: An Integrated Approach for the Analysis of Complete Hybrid Metal-composite
Wing Structures
Pevnostn analza kdla z hybridnho materilu kov-kompozit
Emanuela Pietropaoli / The Second University of Naples, Dipartimento di Ingegneria Aerospaziale
e Meccanica, Aversa, Italy
Incremental STAIRCASE Test Method
Stupovit metoda zkouen
Ing. Tom Blaina / VZL, Plc., Prague
Wind Tunnel Tests of Ice Accretion
Zkouky tvorby nmrazy v aerodynamickm tunelu
Ing. Nikita ikovsk / VZL, Plc., Prague
Development and Improvement of Calibration Methods of Force and Displacement Transducers
Vvoj a zdokonalovn metod kalibrace snma sly a snma deformac
Ing. Jan Heman / VZL, Plc., Prague
Measurement of Buffet Condition on 2D Profile Within AVERT Project at VZLU
Men podmnek buffetu na 2D profilu ve VZLU v rmci projektu AVERT
Ing. Tom Jelnek, Ing. Martin Nmec / VZL, Plc., Prague
Modal Parameters Identification of a Structure Under Operational Conditions
Identifikace modlnch parametr konstrukce v provoznch podmnkch
Ing. Petr Malnek / VZL, Plc., Prague
Grid Generation in ICEM CFD Software for CFD Calculation of Wing with High Lift Devices
Tvorba st pro vpoty kdla dopravnho letounu s vysokovztlakovou mechanizac
Ing. Armand Drbek, Ing. Petr Vrchota / VZL, Plc., Prague
1 L E T E C K Z P R AV O D A J 2 / 2 0 1 0
2
8
12
17
22
25
29
32
36
39
C Z E C H AE R O S PA C E P R O C E E D I N G S 2
1 Introduction
The VUT 001 Marabu (Figure 1) is a two-seater, central-wing
with glass/carbon-fiber composite fuselage and a metal wing
and horizontal tail unit. The aircraft is meant for flight measu-
ring and for autonomous controlli ng research. The aircraft is
equipped with two engines. A piston engine with rear mounted
propeller and a small jet engine located on the wing [1].
The whole project w as supported by Czech Mi nistry of
Industry and Trade under FI-IM3/041 grant (Design and rea-
lization of VUT 001 Mar abu aircraft for UAV applications in
civil spher e). Industri al partners Prvn brnnsk strojrna
Velk Bte, Jihlavan-Airplanes and PlastServis are also enga-
ged in the project [2], [3].
VUT 001 Marab u aircraft is one of fe w practical projects
fully coordinated, designed and also lar gely manufactured at
university thanks to chi ef desi gner of the Marab u project,
Prof. Ptk, who has a courage and pati ence to reali ze this
project at the uni versity environment. Insti tute of Aerospace
Engineering (IAE) has a unique opportunity to test and to veri-
fy some modern approaches, methods and soluti ons duri ng
designing of prototype. In doi ng so, very valuable feedback
can be achi eved, which allows real e valuation of potenti al of
these methods for practical applications.
One of the si gnificant parti al tasks of I AE i s desi gn and
manufacturing of fusela ge. An i mportant part of the project
was to focus as much as possible on usi ng computer -aided
designing during fuselage development. After successful pro-
ject finish, it should allow us to effectively reconsider benefits
of computer-aided design (CAD), computer-aided engineering
(CAE) and computer-aided manufacturing (CAM) implemen-
tation to development process. Fusela ge design methodology
is described by block diagram in Figure 2.
2 Conceptual and preliminary design
The fuselage, as one of the bi ggest structure parts, i s develo-
ped already in early stage of project. It is subjected to all great
changes during defining of aircraft concept and its specificati-
on. VUT 001 Marab u passed through three mai n concept
changes (Figure 3). CAD software proved to be efficient in this
project stage. It allowed immediate and relati vely fast access
to all information needed about aircraft concepts.
Concept No. 3 (Figure 3) entered to next stage of designing
process. Its preliminary design contained the first real data and
parameters, which were used as i nputs to basic aerodynamic,
ergonomic and stress analyses. Last b ut not least mass analy-
sis was ref ined. This process w as interactive and suf ficiently
fast because 3D parametric model was used.
VUT 001 Marabu Fuselage Design
Methodology
Metodologie postupu nvrhu trupu letounu VUT 001
Ing. Michal Mali, PhD, Ing. Tom Urk, Ing. Petr Doupnk, Ing. Ji Zblatzk /
Institute of Aerospace Engineering, Faculty of Mechanical Engineering,
Brno University of Technology
The article describes the design methodology of an experimental aircraft VUT 001 Marabu. The project is coordina-
ted and partially manufactured by Institute of Aerospace Engineering of Brno University of Technology. The article
is focused on design of composite fuselage and describes all phases of realization: initial design, aerodynamic design
of aircraft configuration, detailed support structure design, manufacturing technology, and static test. The emphasis
is placed on effective modern computer aided design methods used during the project.
lnek popisuje postup pi nvrhu experimentlnho letounu VUT 001 Marabu, kter je koordinovn a z sti
vyrbn na Leteckm stavu VUT v Brn. Podrobnji se lnek vnuje nvrhu kompozitnho trupu letounu a popi-
suje cel postup od zkladnho ideovho nvrhu, pes aerodynamick nvrh konfigurace letounu, nvrh nosn
struktury, vrobu a k pevnostnm zkoukm trupu. Draz je kladen zejmna na efektivn vyuit modernch metod
potaov podpory konstruovn.
Fig. 1 VUT 001 Marabu
Fig. 2 Design methodology block diagram
3 Aerodynamic analyses
Modern CFD tools (CA E system) were used i n aerodynamic
design of aircraft at two levels overall aerodynamic concept
evaluation and particular details fine tuning. The aerodynamic
properties of design were evaluated from qualitative and quan-
titative point of v iew using Navier-Stokes steady, fully turb u-
lent finite volume based solution of flow field at different flight
regimes. CFD approach w as already used duri ng conceptual
design, and especi ally duri ng selecti on of tail uni t arrange-
ment. Two different arrangements, V- and T-tail surfaces, was
studied with emphasis on i nterference with propeller i nduced
flow (Figure 4).
As next step, an exhaustive CFD analysis of isolated fusela-
ge and whole 3D configuration were performed. Main goal of
isolated fusela ge analysi s w as to f ind dra g characteri stics
(Figure 5) and to choose the best wi ng incidence angle.
During analysis there was found flow separation under engi-
ne nacelle behind cockpit at higher angles of attack. The shape
of critical area was modified in a few steps and separation was
delayed as far as internal structure design limitations and ove-
rall aircraft concept allo wed. Predicted improvement in flow
pattern at critical area can be seen in Figure 5.
At later stage of design computations of whole 3D configu-
ration were used as basi s for Pi tot-static system locati on pre-
diction. The target of analysis was to find a place on the fuse-
lage where static and total pressures are not af fected by chan-
ge in angles of attack and sideslip. The location with minimum
static pressure dependence on angle of attack is shown in Figu-
re 7.
Another example of extensive use of CFD method for solu-
tion of a particular practical problem is engine cooling analysis
and nacelle shape design. Upon engine cooling system charac-
teristics, aerodynamic design of engine nacelle shape was per-
formed. Optimal position of coolers, air inlet and outlet shapes
Fig. 3 Evolution in preliminary design
Fig. 4 Complex CFD analysis of whole configuration
Fig. 5 Fuselage drag characteristics
Fig. 6 Cabin and engine nacelle shape modification
with regard to air supply and air exhaust from engine area with
minimum drag was estimated. The cooling and cowling of pis-
ton engine is a very complex task, which is still solved mainly
by experimental approach during test flights. CFD analysis of
such a complex problem is still not very precise. Nevertheless
it gave us at least early warning about critical problems and i t
directed us i nto detailed desi gn of outer and i nner shape of
nacelle and ducting.
4 Detail design
4.1 Master model
The main documentation creation started after completi on of
preliminary aerodynamic and ergonomic analyses. In this case
it was 3D master model of fuselage (Figure 8). Fuselage mas-
ter model is created by system outer surf aces, by system sur-
faces of walls and stringers and by hi nge and leveling points.
As such i t contains all fundamental parameters and i t is main
documentation basi s for all other structural parts whi ch are
somehow connected to fuselage. And also data for manufactu-
ring and for next designing and refining analyses are based on
fuselage master model geometry.
4.2 Regulation requirements and manufacturing limits
The aircraft is designed according to CS-VLA regulation requ-
irements. The development was limited by some manuf actu-
ring constrains and regulations restrictions:
G Fuselage manufacturing in two-pieces female mould with
the mould joint in the longitudinal symmetry plane
G Minimization number of moulds for ti me and f inancial
reasons
G Factor of safety for composi te material supporting struc-
tures has to be i ncreased to 2.25. Standard safety f actor
1.5 is multiplied by 1.25 speci al safety f actor of tem-
perature influence and 1.2 special safety factor of moi-
sture influence
4.3 Structure design and stress analysis
The fusela ge structure w as desi gned as mai nly glass f iber
composite lami nate wi th sandwich core. Carbon f ibers were
used i n local rei nforcement. Pure glass f iber rei nforcement
plastic (GFRP) wi thout sandwich core has proporti onal stiff-
ness (Young modulus/density) 2.5 times worse than aluminum
alloy. Thus GFRP wi thout sandwi ch core i s not ef fective i n
term of mi nimum structure wei ght. Sandwi ch core materi al
was used for overall composite fuselage including flat bulkhe-
ads. Using of sandwich material led to vacuum bagging curing.
The complete 3-D model created before manuf acturing and
used in design phase of the project reduces the role of standard
drawing documentati on. Dra wing documentati on w as used
only i n communi cation wi th e xternal suppli ers. Duri ng the
manufacturing of the fusela ge composite parts drawing docu-
mentation was not used at all.
We must consi der that the comple x shape of the fusela ge
does not ha ve any strai ght edge necessary for planar length
measuring. Therefore longitudinal dimension was measured as
3D dimensions from leveling points on the mould surface.
4.3.1 Lay-up design
All materi als used i n the fusela ge were certif ied. Inter glass
92110 and 92125 twill f abric, Interglass 92145 uni directional
fabric and unidirectional carbon tape were used. L-285 epoxy
resin with L-287 hardener was also used.
The layer mechanical properties were considered according
to conservative values accepted by LB A (The German Ci vil
Aviation Authority) [4] (Table 1).
The composite material mechanical properties used at the fuse-
lage design (especially
1tm
,
1c
) are consi derably lower than
we are able to reach i n laboratory condi tions. It i s caused by
imperfections in fabric structure which arise during the manu-
facturing. The imperfections decrease those strength characte-
ristics considerably. If we want to use hi gher strength charac-
teristics we w ould have to carry out analysi s of i nfluence of
manufacturing on strength characteristics.
The composi te plies i n small ai rcraft fusela ge are usually
laid-up under two directions. Longitudinally -fibers are along
Fig. 7 Static pressure distribution - pitot-static system design
Fig. 8 The basic fuselage
documentation - 3D master model
Tab. 1 Used composite material characteristics
Index x mechanical layer properties laid up under angle 45 Deg to
the angle of loading
the longi tudinal fusela ge ax is and di agonally -f ibers are
under angle 45 to the longi tudinal fuselage axis. The lon-
gitudinal pli es are subjected pri marily to normal loadi ng
(bending, tensi on, compressi on) support. The di agonal
plies subjected to shear forces and torque moment. The
normal loadings are distributed by diagonal plies very non-
effectively. It causes de greased Young modulus of materi-
al wi th the f iber angle 45 to the angle of loadi ng. The
same principle arises during shear loading of a longitudinal
laid plies.
The lami nate desi gn be gan wi th a survey of stack ing
sequences of si milar aircrafts. Information about stack ing
sequences of the composite plies is available in maintenan-
ce manuals [5], [6] (necessary i n case of repai ring the
structures). The manuals are accessible on the Internet.
The lami nate desi gn strate gy w as based on mi nimal
weight requi rement. Therefore the number of la yers over
the whole fusela ge length w as mi nimized. All the more
local reinforcements were used i n places of the load con-
centrations. Load i ntensity of the small ai rcraft structures
is relatively low. Therefore only 1-2 la yers are needed for
each load di rection (longitudinal and under angle 45) for
achieving structure load capacity, if the layers are designed
to the maximum strength accordi ng to table1. In thi s way,
a thin walled structure susceptible to buckling will be obta-
ined. For that reason all available features against buckling
were used: especi ally sandwi ch core, ribs and stri ngers.
The stringers were consisted only from di agonally layered
plies. They were not desi gned for normal loadi ng b ut as
a prevention of a fuselage cross section collapse.
Local rei nforcements were used at the hori zontal tail
unit hinges, in the connecti on between fusela ge cone and
keel, and i n place of centre-wi ng. Uni directional carbon
tape was used along the whole fusela ge (two tapes around
the fuselage cone), as a keel spar flanges, flanges of the
power unit box and reinforcement along cockpit edge [9].
All desi gn strength analyses were carri ed out before
manufacturing usi ng softw are packa ge MSC
Patran/Nastran (CA E) wi th consi dering of geometri cal
nonlinearities. The li near 2D orthotropi c materi al model
was used for simulating of the composite material and line-
ar isotropic material model for sandwich cores. The whole
fuselage i ncluding sandwi ch core w as modeled by shell
elements. The elements properties were set down by Lami-
nate Modeler module [7] whi ch is implemented in prepro-
cessor MSC. Patran.
The finite element model used in simulation was structu-
rally simplified in order to reduce the si mulation time con-
sumption. One nonlinear simulation took approximately 40
minutes. The manual method of pro gressive plies modify-
ing was applied. The final version was achieved in 9 iterati-
on steps.
The main goal of the si mulation was to specify the la y-
ers amount in structure, number of layers working as local
reinforcements. It also sho wed that sandwi ch material has
to be appli ed all o ver the structure of the fusela ge. (The
table of Herex sandwich core is rigid and not easily appli-
cable into curved surfaces; that is why we considered to do
not apply the sandwich in fuselage cone.)
4.4 Manufacturing technology
Chosen concept and usa ge of f iber reinforced plastics led
to selecti on of manuf acturing technolo gy of fusela ge
monocoque i n tw o-piece ne gative mold. Mai n structural
walls were placed and adhesi ve bonded to manuf actured
fuselage monocoque. Most of all composi te parts were
manufactured by hand la y-up wi th consequent v acuum
bagging.
It was needed to speed up manuf acturing of monocoque
mold after f inishing of outer shape of master model. The
mold was manuf actured i n fe w steps. The f irst step w as
creating of 3D model of mold, whi ch was based on master
model and i t also contai ned marks of le veling points and
some marks of system planes (Fi gure 9a). It was not pos-
sible to manufacture mold with all important marks of sys-
tem planes and technological stops due to lack of time. The
milling of tw o-pieced negative mold w as the second step
(Fig. 8b). Polyurethane was chosen as stock for mold mil-
Fig. 9 Manufacturing procedures
ling mainly due to economi c reasons. Mold was milled by
6-axis CNC milling machine JOBS with max stock dimen-
sions 7.5x3x3m (CAM). In ne xt step, composi te posi tive
model of fuselage was manufactured in polyurethane nega-
tive mold. In consequence, surf ace finishing was done on
positive model. Surf ace finishing of negative polyurethane
mold w ould be v ery compli cated and durabili ty of thi s
mold was also very weak. In last step, two-pieced negative
composite mold was manufactured with the aid of positive
model and one half of polyurethane ne gative mold. Dura-
bility of composi te mold i s approximately 10 manuf actu-
ring cycles.
Fuselage monocoque manuf acturing started after com-
pletion of design and stress analysis. All necessary geome-
trical data were adopted from master model. F or example
information about surf ace area of parti cular layers, which
were lay-up to mold, allowed to determine exact amount of
matrix for achieving required fiber volume ratio 35%. And
it was also possible to determi ne final weight of monoco-
que. Dif ference between wei ght assumpti on determi ned
before manufacturing and weight of monocoque after rele-
asing from mold i s 5%. Also preci se unfolded patterns
were manufactured for all sandwich cores.
Monocoque manuf acturing i tself li ed i n la y-up and
impregnation of particular structural layers and subsequent
vacuum bagging of ri ght and left half of mold (Fi g. 9d).
Some walls and stri ngers were bonded i nto ri ght and left
half of mold after curi ng and before monocoque bondi ng.
The rest of cross-secti onal and longi tudinal system w as
installed to manufactured fuselage monocoque.
5 Results and evaluation
5.1 Static test and comparison with analysis
The stati c test w as carri ed out i n I AE testi ng f acilities
(Figure 11). The tested load cases were [8]:
1. Keel and hori zontal tail uni t loaded to gether bend
and torsion of the fuselage cone.
2. Maneuver on horizontal tail uni t bend of the fuse-
lage cone.
Evaluated data during the test were deformation in selected
points on the fusela ge top surf aces, di splacement of the
structure and acti ng force. The deformations were measu-
red by 27 strai n gages located especi ally on the fusela ge
cone and k eel. The HBM 10/120L Y13 strai n ga ges were
used for tensi on/compression measuri ng and strai n ga ges
HBM 6/120XY12 for shear forces measuri ng. The strai n
gages were adhesive bonded on the structure usi ng certifi-
ed glue HBM Z80. The structure displacements were mea-
sured by aripots.
The data during the test were stored using by data logger
ESAM Static. The loaded force w as created by tw o hyd-
raulic cylinders.
Both load cases were appli ed on the structure to 100 %
of limit load level. During the test of the f irst load case to
ultimate load level the test was stopped at 140% due to big
deformation between k eel and fusela ge cone. Duri ng the
Fig. 10 FEM model with highlighted
deformation test points
Fig. 11 Static test
Fig. 12 Displacement results at the end of fuselage Fig. 13 Deformation results
second load case (maneuver on the horizontal tail unit) the
structure withstood to 225% of limit load level without fai-
lure. On the basis of the first load case test result, the rein-
forcement of the connecti on between k eel and fusela ge
cone was designed.
Figures 12 and 13 sho w the comparison of the simulati-
on result wi th the stati c test result. The f irst represents
dependence between limit loads from the second load case
on di splacement whi ch w as measured at the end of the
fuselage. The dif ference between max imal v alues of the
test and simulation is 12.5%. The second is limit loads ver-
sus deformation in selected point on the fuselage according
to Figure 10.
5.2 Benefits of CAD, CAM, CAE application
At this moment, benefits of computer-aided designing appli-
cation can be e valuated from manuf acturing poi nt of v iew
(interaction of CAD and CAM) and from desi gn and stress
analysis point of view (CAD, FEM). Benefits from aerodyna-
mic point of v iew (CAD, CFD) will be reconsi dered during
flight tests.
Usage of 3D fuselage master model is crucial from manu-
facturing point of v iew, mostly when manuf acturing is con-
nected wi th CNC milli ng of fusela ge monocoque mold.
Fuselage monocoque creates approx imately 80% of whole
fuselage structure. Classi c 2D dra wing documentati on i s
insufficient and i t becomes useless. Only 3D documentati on
contains all parameters about structure and also all data for
manufacturing and for analysis. Main deficit during manufac-
turing process w as that i t was not possible to manuf acture
mold with all important marks of system planes and techno-
logical stops for cross-sectional and longitudinal systems and
for hinges.
The practical application of CAD and CAM led to suf fici-
ently lightweight, technologically effective and dimensional-
ly precise structure on the first attempt. Failure did not occur
in the structure duri ng the test. Weak parts of the structure
were detected and thei r rei nforcement w as desi gned. The
reduction of some supports features will be reconsi dered in
the feature. The reserves can be i n some ribs, stri ngers and
local reinforcements.
The finite element model used at the analysi s was created
for desi gn of the lami nate structure before manuf acturing
was begun. And so, this model contains number of inaccura-
cies created by partial changes of structure during the manu-
facturing process and also inaccuracies created thanks to sim-
plifying of the model for time consumption reduction. In near
future a modification of the model i s planning according to
final fuselage structure and accordi ng to enhanced materi al
properties. Presently the full aeroelastic model is being com-
pleted.
6 Conclusion
Computer-aided desi gn enters i nto each sta ge of project. In
order to effectively use its potential some of the basic require-
ments should be fulfilled:
- Complete superi ority of 3D v irtual model. 2D dra wing
documentation i s mi nimized and has only i nformative
character.
- Clear and transparent structure of virtual prototype
- Creation of master models, or surf ace based system
models of particular structural units, which can be easily
and effectively used for manufacturing (CAM) process as
well as for design and verifying (CAE) analyses.
- Ability of data transfer between parti cular software sys-
tems of computer-aided designing (CAD, CAM, CAE).
- Transparent and effective control of data flow.
Recently, PLM (Product Lifec ycle Management) systems are
able to fulfill most of all above mentioned requirements. PLM
systems integrate CAD, CAM, CAE software and data mana-
gement and so the y allo w to w atch and to control product
through its whole life-cycle. Good and effective implementati-
on of PLM system is dependent on precise knowledge of many
factors (design process, manufacturing process, data mana ge-
ment, company communication, etc.). One of the goals of VUT
001 Marabu project is also to gather, to describe and to quali-
tatively e valuate as man y parameters as possible whi ch are
important for effective use of PLM system in aerospace indust-
ry.
In the future, we see a great challenge in implementation of
aircraft certif ication process i nto PLM system. F or e xample
Civil Aviation Authority could be one of cooperating members
inside PLM system wi th exactly defined access to all project
data. It could lead to si gnificant simplification of communica-
tion and eventually to acceleration of certification process.
References
[1] Ptk, A.: Technick specifikace - zadn letounu
VUT 001 Marabu (Technical Specification of VUT
001 Marabu Airplane); Internal report LU53-2006-
001.AS, 2006
[2] Ptk, A., Hlinka, J.; Urk, T.; plchal, M.:
VUT 001 Marabu: Design philosophy and realization
of an experimental aircraft supporting UAV
applications in civil sector; 24
th
International
Conference Unmanned Air Vehicle Systems, Bristol,
United Kingdom, 2009
[3] Hlinka, J.; plchal, M.; Finda, J.: Equipment,
systems and installations for experimental aircraft
supporting UAV applications; International
Symposium Light Weight Unmanned Aerial Vehicle
Systems and Subsystems, Oostende, Belgium, 2009
[4] Interessengemeinschaft Deutscher Akademischer
Fliegergrupen, Dimensionierungsrichwerte fr den
Segel und Motorsegelflugzeugbau, Mrz 1988
[5] DA 40 Series Maintenance Manual, Diamond
Aircraft Industries, Wiener Neustadt, Austria, 2003
[6] H 36 Dimona, Instructions For Continued
Airworthiness; Hoffman Aircraft Corp., Vienna,
Austria, 1985
[7] MSC. MD Patran Laminate Modeler User's Manual,
Santa Ana CA, November 2007
[8] Vank F.: Zaten ocasnch ploch a zadn sti trupu
VUT 001 Marabu; Internal Report LU26-2007-
001.L, 2007
[9] Mali, M.: Podklady pro vrobu trupu VUT 001
Marabu; Internal Report LU04-2009-001.DE, 2009
1 Formulation of the Navier-Stokes equations
for turbulent flow
We will consi der the Na vier-Stokes equations in conservation
form with dimensions. We will apply the la w of conservation
of mass-continuity, momentum, and ener gy for a volume ele-
ment through which fluid is flowing. For the three-dimensional
case this system of the Navier-Stokes equations has the form
(1)
where
and
with p the pressure, the density, (u, v, w) the average value
vector of velocity, and x, y, z the space coordinates, and t the
time. Further, k is the turbulent kinetic energy of flux compo-
nents of the velocity, is the specific turbulent dissipation, P
r
is laminar and P
r
T
is turbulent Prandtl constant number, is the
dynamic v iscosity coef ficient dependent on temperature,
T
= ld is the eddy-v iscosity coef ficient. In the ener gy
equation, e denotes the total energy
where = p/(-1) is the internal energy of a unit mass of the
fluid where the constant >1.
This system of equations (1) is an open system for turbulent
flow. If the turbulent kinetic energy k = 0 , then the system of
equations (1) is a closed system of the Navier-Stokes equations
for a laminar flow. If k = 0 and = 0, then (1) are the Euler
equations. The system studied (1) can be rewritten into the dif-
ferential symbolic form
and in the integral form it reads
where i = 1,2,3,4,5, is from the space R
3
(x,y,z). ( , ) denotes
the scalar product. n is a normal vector to the surface . The
positive orientation is given by the outward direction. Here s is
Axis-Symmetrical 3D Viscous Turbulent Flow
with (k , ) Standard Turbulent Model
Written to the Cylindrical System
Formulace (k , ) standard turbulentnho modelu pro 3D osov
symetrick vazk turbulentn proudn v cylindrickm systmu
RNDr. Jaroslav Pelant, CSc., RNDr. Martin Kyncl / VZL, Prague
This paper deals with a fluid flow through a symmetrical channel in 3D space and time, described by the Navier-
Stokes equations with k-omega turbulence model. System of equations for axis-symmetrical flow is rewritten into
cylindrical coordinates. Computations are done in meridian plane. A one-side modification of the Riemann problem
is used for the boundary conditions as the original result. Total pressure and total density are to be used preferably
at inlet whereas pressure or temperature or velocity or mass flow should be used at the outlet. The finite volume
method with explicit time marching was used. For higher accuracy in space we used Van Leer or Van Albada sche-
me. Described method can be used for flow simulation in symmetrical channels of arbitrary apparatuses.
Pspvek se zabv osov symetrickm proudnm tekutiny ve 3D. Problm je zde popisovn Navier-Stokesovmi
rovnicemi s k- standard modelem turbulence, a tento systm rovnic je pro osov-symetrick problmy pepsn do
cylindrickho systmu. Vpoty jsou pak provdny v meridiln rovin. K diskretizaci vzniklho systmu je pouito
metody konench objem s explicitn asovou diskretizac. Je vyuit originln pstup k een okrajovch podm-
nek zaloen na jednostrann modifikaci Riemannova problmu. Pro zven pesnosti v prostoru je pouito limite-
r Van Albada, Van Leer.
Keywords: Axis-symmetrical Flow, the Navier-Stokes Equations, k- Turbulent Model, Finite
Volume Method, Boundary Conditions, 3D, Riemann Problem.

0 = ) ( ) ( ) ( ) ( ) ( ) (
J
J
J
`
'
'
(
|
q d
z
q s
y
q r
x
q h
z
q g
y
q f
x
q
t
) , , , , ( = e w v u q

) ) ( , , , , ( = ) (
2
u p e uw uv p u u q f + +

) ) ( , , , , ( = ) (
2
v p e vw p v vu v q g + +

) ) ( , , , , ( = ) (
2
w p e p w wv wu w q h + +

J
J
J
J
`
'
'
'
(
|
J
J
J
J
`
'
'
'
(
|
+ + + +
x P P
w v u q r
T
r
T
r
xz xy xx xz xy xx

, , , 0, = ) (

J
J
J
J
`
'
'
'
(
|
J
J
J
J
`
'
'
'
(
|
+ + + +
y P P
w v u q s
T
r
T
r
zy yy xy zy yy xy

, , , 0, = ) (

J
J
J
J
`
'
'
'
(
|
J
J
J
J
`
'
'
'
(
|
+ + + +
z P P
w v u q d
T
r
T
r
zz yz xz zz yz xz

, , , 0, = ) (

( )
3
2
3
2
3
2
3
4
=
k
z
w
y
v
x
u
T
xx

J
J
J
`
'
'
(
|
+ +

( )
3
2
3
2
3
4
3
2
=
k
z
w
y
v
x
u
T
yy

J
J
J
`
'
'
(
|
+

( )
3
2
3
4
3
2
3
2
=
k
z
w
y
v
x
u
T
zz

J
J
J
`
'
'
(
|
+

( )
J
J
J
`
'
'
(
|
+
y
u
x
v
T
yx xy
= =

( )
J
J
J
`
'
'
(
|
+
z
v
y
w
T
zy yz
= =

( ) , = = J
J
`
'
(
|
+
x
w
z
u
T
xz zx

), (
2
1
=
2 2 2
w v u e + + +

0 =
z y x t
i i i i

, d d d d = d ) ), , , (( d z y x
t
t s n t
i
t
i i i
t

the integral measure in the surface . Using the integral form
we can study a flow with shock waves, too. For example we
can apply the one-di mensional system of the same equati ons
for the predictor of the numerical method at the speci al points
of a chosen grid in the region .
2 Formulation of the equations
for an axis-symmetrical flow
For a symmetrical three dimensional flow we use the transfor-
mation of the equati ons to the c ylindric system
(t,x,y,z t,x,r,) , where = r cos and z = r sin . The for-
mulae for components of velocity are
The axis of rotation is x, the radius is r, and the angle of rotati-
on is . The components v
r
, w
are radial and circle coordina-

tes.
Now we can change denotati on r y, z, v
r
v, and
w
w. The new system of equati ons i n meridian plane for

axis-symmetrical flow has the form
(2)
where
q = (,u,v,w,e)
f(q) = (u,u
2
+ p,uv,uw(e+p)u)
g(q) = (v,vu,v
2
+ p,vw,(e+p)v)
F(q) = (v,uv,(v
2
-w
2
)2vw,(e+p)v)
with
The system studi ed (2) can be re written into the dif ferential
symbolic form
and in the integral form it reads
(3)
where i = 1,2,3,4,5, is from the space R
3
(t,x,y). ( , ) denotes
the scalar product. n is a normal vector to . The positive ori-
entation is given by the outward direction. Here s is the integral
measure i n the surf ace . Usi ng the i ntegral form we can
study a flow with shock waves, too. For example we can apply
the one-dimensional system of the same equations for the pre-
dictor of the numeri cal method at the speci al points of a cho-
sen grid in the region .
3 Modification of the k - standard turbulence
model for axis-symmetrical flow
Turbulent model for ax is-symmetrical flow can be described
by following equations in meridian plane
(4)
(5)
where k the turbulent kinetic energy and the turbulent dissi-
pation are functi ons of ti me t and space coordi nates x,y. The
production terms P
k
and P
are given by formulas

where functions are defined in Chapter 1. for = 0.
This turbulent model k- (4), (5 ) with equations (2) presented
the closed system of equations.
4 Numerical Time Step Method
Let a quadrangle gri d be gi ven by poi nts ( x
j,k
,y
j,k
), j = 1,...,J
and k = 1,..., K in a meridian plane of arbi trary symmetri cal
three-dimensional channel. This grid is independent of ti me t.
To show the pri nciple of the method we choose an arbi trary
cell of the grid. For the sake of simplicity this cell is denoted
This cell is a side of the respective cell from space R
3
(t,x,y).
The cell is defined for time interval . Now we want to use
equation (3) for this special cell . So, let
d
be the lower side
of at time t,
u
be the upper side at time t+ and
f
,
r
,
l
,
h
are other si des called walls. Integral equation (3) has the
special form
(6)
where
Q
r
=
i
r
(y
2
-y
1
)-
i
r
(x
2
-x
1
)
Q
f
=
i
f
(y
1
-y
4
)-
i
f
(x
1
-x
4
)
Q
h
=
i
h
(y
3
-y
2
)-
i
h
(x
3
-x
2
)
Q
l
=
i
l
(y
4
-y
3
)-
i
l
(x
4
-x
3
)
and the upper index at
i
,
i
,
i
means the value on a side of the
same notation. Using equation (6) for
i
u
it is possible to obta-
in state values p, , u, v at the centre of
u
if other values are
known at other centres of six sides. The state values for
d
are
known. They are i nitial condi tions at the f irst time step. The
main problem is to obtain state values on the sides
f
,
h
,
l
,
r
called walls. Let us choose one w all. If thi s wall is inside
the grid, then we can define state values from either side of the
wall in the middle of the segment at time t. Any definition can-

sin cos = w v v
r

cos sin = w v w
r
+

) (
1
) (
1
= ) ( ) ( ) ( ) ( q G
y
q F
y
q s
y
q r
x
q g
y
q f
x
q
t
+
J
J
J
`
'
'
(
|
J
J
`
'
(
|
+ + + +
x Pr Pr
w v u q r
T
T
xz xy xx xz xy xx

) ( , , , 0, = ) (

J
J
J
`
'
'
(
|
+ + + +
y Pr Pr
w v u q s
T
T
zy yy xy zy yy xy

) ( , , , 0, = ) (

'
'
(
|
J
J
J
`
'
'
(
|
+
J
J
J
`
'
'
(
|
+
J
J
J
`
'
'
(
|
+ , ) ( ,
3
4
) ( ,
3
1
) ( 0, = ) (
y
w
y
w
y
v
y
v
y
u
x
v
q G
T T T

J
J
J
`
+ +
J
J
J
`
'
'
(
|
+ v
k
y Pr Pr y
w
w
y
u
u
x
v
u
x
u
v
T
T
T
3
2
) (
3
1
3
4
) (

3
2
) (
3
2
3
4
=
k
y
v
x
u
T xx
+
J
J
`
'
(
|
+

3
2
) (
3
4
3
2
=
k
y
v
x
u
T yy
+
J
J
J
`
'
'
(
|

) ( = =
T yx xy
y
u
x
v
+
J
J
J
`
'
'
(
|

). ( = ), ( =
T yz T xz
y
w
x
w
+
J
J
J
`
'
'
(
|
+ J
J
`
'
(
|

i
i i i
f
y x t
=

, d d d ) , , ( = d ) ), , , (( t y x t y x f s n
i i i i

( ) + J
J
`
'
(
|
x
k
x
k P
y
kv
x
ku
t
k
T
k k

*
=

( ) ( )
J
J
J
`
'
'
(
|
+ +
J
J
J
`
'
'
(
|
+ kv
y
k
y y
k
y
T
k
T
k

1

( ) + J
J
`
'
(
|
x x
P
y
v
x
u
t
T

2 *
=

( ) ( )
J
J
J
`
'
'
(
|
+ +
J
J
J
`
'
'
(
|
+ v
y y y y
T T

1

+
y
v
x
v
y
u
x
u
P
yy yx xy xx k
=
y
kv
x
w
y
v
x
u
y
w
y
v
y
w
y
w
y
v
y
v
y
w
T
3
2
3
4
3
4
2
3
4
2 2 2 2

]
]
]
]
J
J
`
'
(
|
J
J
J
`
'
'
(
|
+
J
J
J
`
'
'
(
|
+
J
J
J
`
'
'
(
|
+
. , a a , , , a = w , =
*
*
2
*
seeKok constants re nd here
k
P
P
k
k

)) , ( ), , ( ), , ( ), , (( =
4 4 3 3 2 2 1 1
y x y x y x y x
d

y
y x
f Q Q Q Q
d
d
i h l r f d
d
i u
u
i
d d
=

+ + + + P P P P

)) )( ( ) )( ((
2
1
= =
1 3 2 4 2 4 1 3
y y x x y y x x
d u
P P P P
not keep the laws of conservation. For this case it is possible to
apply the results shown in [4], Chapter 2. If thi s wall is outsi-
de side of the grid, then we can apply [3], Chapters 4-8 for i nlet
and outlet flo w. If the menti oned wall i s on the surf ace of
a body, then we can apply [4], Chapter 2, too . This technique
can be used because of properti es of equati ons i n the space
R
3
(x,y,z), which is invariant wi th respect to arbi trary rotation
matrix transformation. It i s sufficient to choose the the same
plane grid i n some dif ferent meridian planes and so to con-
struct a special grid in R
3
(t,x,y) for the symmetrical flow. Time
step for each w all i s bounded by elementary shocks or
expansion waves coming from opposite walls. We know velo-
cities of those waves using [4], Chapter 2 or [3], Chapters 4-8.
The last problem whi ch i s necessary to e xplain i s how to
define state v alues at ei ther side at the centre of a common
boundary of two cells when state values are known at the cent-
res of all the cells at the same ti me t. For this case it is possib-
le to apply some dif ferent schemes. But those schemes must
lead to a stable process of convergence to a solution of steady-
state flow.
To show some schemes we will use j,k indices for a cell of
vertex ( x
j,k
,y
j,k
), ( x
j+1,k
,y
j+1,k
), ( x
j+1,k+1
,y
j+1,k+1
), ( x
j,k+1
,y
j,k+1
).
State values at the centre have the same index as the cell. State
values as the li mits from the ri ght and left si de of poi nts on
common boundary ha ve the mi ddle-value index of the cells.
The upper indices r and l mean the right and the left limits. The
first-order scheme is
For simplicity the follo wing schemes are wri tten for square-
grid. Another important scheme is
VAN LEER
where -1,1. We can use = 1 and = -1 at the same time
and choose lower difference and if the both dif ferences are of
the opposi te si gn, then we choose zero -difference. Value
1/3 yields a third-order truncation error.
Higher-order schemes can be written as
VAN ALBADA
with
and a function () comes from a region of monotonicity. For
example we can choose the function
The convergence of the ti me-step process i s checked by resi-
due of the solved system of equations.
5 Examples
Here we present computational simulation of 3D axis-symme-
trical turbulent gas flow for given geometries. Axis x is the axis
of symmetry . Adiabatic constant = 1.4, gas constant
R = 287.04 [ Jkg
-1
K
-1
], dynami cal coef ficient of
viscosity = 0.1697 [kgm
-1
s
-1
], heat conducti vity coefficient
k = 0.0211 [Wm
-1
K
-1
]. Coefficients
k
,
*
, ,
, forming cho-
sen turbulence model, were given from [5].
Fig. 1 shows 3D axis-symmetrical geometry, mach number
isolines and streamlines of the following example. Computati-
onal mesh in 2D crosscut consi sted of 244x46 quadrilaterals.
At the i nlet we chose boundary condi tion conserv ing total
pressure p
o
= 101325 [Pa], total temperature T
o
= 273.15 [K],
and zero tangenti al velocity ( see [3]). At the outlet we used
boundary condition from [3] for given pressure p = 45722.531
[Pa]. This yielded transonic flow. 600000 iterations were com-
puted. Pressure, density, mach and turbulent kinetic energy iso-
lines are illustrated at Fig. 2 in 2D geometry crosscut. Results
are normed to critical values at the inlet. In this case the velo-
city is normed to critical speed of sound c
a
= 302.44163441783
[ms
-1
], pressure is normed to c
a
2
a
= 74939.4129966198 [Pa],
density to
a
= 0.819256453423711 [kgm
-3
].
Another example was made for geometry i n Fi g. 3. Total
pressure at the i nlet p
o
= 101905 [ Pa], total temperature
T
o
= 875 [K], zero tangential velocity (condition based on [3]).
Outlet average pressure was given as p = 100000 [Pa]. Mesh
consisted of 86x46 quadrilaterals, 5400000 i terations compu-
ted. Fi g. 3 demonstrates mach number i solines and v elocity
dstreamlines. Fig. 4 shows pressure, density, mach, and turbu-
lent kinetic energy isolines.
Outlet boundary condition changed for the computation dis-
played in Figs. 5 and 6. We chose p = 41325 [Pa]. Computati-
onal geometry di dn't change (Fi g. 5). Mesh consi sted of
244x46 quadrilaterals, 2300000 iterations computed. Pressure,
C Z E C H AE R O S PA C E P R O C E E D I N G S 1 0
Fig. 1

k j
r
k j
k j
l
k j
q mmq q q
1,
,
2
1 ,
,
2
1
= 5 , =
+
+ +

), (
4
1
) (
4
1
=
) (
4
1
) (
4
1
=
2, 1, 1, , 1,
,
2
1
1, , , 1, ,
,
2
1
k j k j k j k j k j
r
k j
k j k j k j k j k j
l
k j
q q q q q q
q q q q q q
+ + + +
+
+
+
+
+
+
+
+
+

) )( ) ((
2
1
=
) )( (
2
1
=
2, 1,
1
1, 1, 1,
1, , , ,
,
2
1
k j k j k j k j
r
k j
k j k j k j k j
l
k j
q q q q
q q q q
+ +
+ + +
+
+
+

k j k j
k j k j
k j
q q
q q
1, ,
, 1,
,
=

1
= ) (
2
2
+
+

Fig. 3
density, mach, and turbulent kinetic energy isolines are shown
in Fig. 6. Fig 5 is with mach isolines and velocity streamlines.
6 Conclusion
We presented numerical method to solv e the Navier-Sto-
kes equations with k- turbulence model for 3D axis-sym-
metrical flo w. Described method can be used for flo w
simulation in symmetrical channels of arbi trary apparatu-
ses. Computational results were presented.
Acknowledgment
This work was supported by the Grant MSM 0001066902
of the Mi nistry of Education of the Czech Republi c. The
authors acknowledge this support.
1 1 L E T E C K Z P R AV O D A J 2 / 2 0 1 0
References
[1] M. Feistauer and J. Felcman
and I. Strakraba:
Mathematical and
Computational Methods for
Compressible Flow; Oxford
University Press, Oxford,
2003
[2] J. Pelant: ARTI Reports
VZL, Z-65, Z-67 to Z-73.
Prague, 1996-2000
[3] J. Pelant, M. Kyncl:
Applications of the Navier-
Stokes Equations for 3d
Viscous Laminar Flow for
Symmetric Inlet and Outlet
Parts of Turbine Engines with
the Use of Various Boundary
Conditions; Report VZL
R3998, Prague, 2006 (in
English)
[4] J. Pelant, M. Kyncl:
Applications of the Navier-
Stokes Equations for 2d
Viscous, Compressible
Turbulent Flow on Steady
Grids with the (EARSM)
Turbulent Model; Report
VZL R4300, Prague, 2008
(in English)
[5] C. Johan Kok: Resolving the
Dependence on Free-stream
Values for k- Turbulence
Model; AIAA Journal, Vol. 38,
No. 7, July 2000 (i n English)
Fig. 4
Fig. 5
Fig. 6
Fig. 2
Introduction
Evektor, spol. s r.o. company is involved in design and certifi-
cation of se veral types of ai rplanes. The production is almost
entirely related to all-metal ai rplanes, where ri veting i s the
most common type of joi ning. The rivets are very different in
comparison wi th e.g. welds, whi ch are toda y commonly
modelled in finite element analyses. The stiffness of a riveted
joint is substantially lower. As the most of the f inite element
models of aircraft structures are built from 1D and 2D types of
elements, a suitable way of riveted joints representation has to
be looked for. The model has to be built quickly, should not be
too big, but the load-carrying capacity as well as the joint stiff-
ness should be as close to observed values as possible.
The simplest solution of direct joining of two sheets together
by gluing the mesh to gether is far from reali ty and could be
better related to seam welds. Another way of solution is the use
of simple 1D elements (such as beams) with a tuned cross-sec-
tion in order to si mulate the joi nt stiffness adequately. There
are quite a lot of v arious stiffness estimates given for ri veted
structures [3], b ut the f inal values dispersion and poor docu-
mentation of different rivet types is obvious.
Evektor's interest in better understandi ng the phenomenon
and its intention to improve the design practice were the impe-
tus to joi n the consorti um around the IMPERJ A project (an
acronym for Improving the Fatigue Performance of Ri veted
Joints i n Airframes) under the EUREKA scheme. The full
name of the project unveils another focus of the works the
finite element models and achieved results should serve as the
input for further fatigue analyses.
The timetable for the appli ed research i n Evektor was thus
set. Two selected types of specimens will be evaluated experi-
mentally to get the full S-N curv e. These (quite complicated)
specimens will be i nput for stati c evaluation by the FEA and
fatigue life prediction based on the FEA. The necessity of pro-
ducing the FEA-based f atigue life esti mates lead i nto the full
3D solid models of riveted joints as the first outcome, while the
search for the simplified 1D-2D solution will be looked at just
afterwards, so that it could be compared with adequately tuned
solid models.
The focus on the experiments involving only the fatigue tes-
ting had sho wn to be too opti mistic and soon another set of
simpler specimens described in the next chapter was proposed
for static experiments. This experimental evidence concluded
the preparatory part that allo wed proper tuni ng of the FEA-
models. The paper presented here shows the most of the expe-
rience gained during the participation in the IMPERJA project
in the section of the FEA.
Small-scale test specimen of the riveted joint
The basic static characteristics of the ri veted joint were evalu-
ated on the basi s of stati c experiments made on si mple two-
rivet lap-joint specimens (see Fig. 1). They consist of two she-
ets from D16C aluminium alloy, which were connected by two
NAS1097AD-4 ri vets [1]. The thi ckness of the sheets i s
0.8 mm. This combination of sheet thickness, the rivet type and
diameter is not very common, because the depth of the coun-
tersunk is almost the same, as the sheet thi ckness. This may
cause some problems, because only a small cylindrical bearing
area remains between the sheet and the ri vet to transfer loads
from one to another, while the load transfer through the coni-
cal part is quite inefficient.
FE-Modelling Methodology of Riveted
Joints
Metodika tvorby MKP modelu ntovho spoje
Ing. Jaromr Frek / Evektor, spol. s r.o., Kunovice
Riveting is the most common way of connecting parts together in the aircraft industry. This document deals with
finite element (FE) simulations of riveted joints of various complexity. The riveted structures are modelled realisti-
cally including first the phase of forming a riveted joint and then also the shearing of the joint. Some parameters
influencing behaviour of such joints are evaluated in further FE-analyses. Properly tuned detailed solid FE-models
were then used for a definition of their suitable replacement by simpler modelling structures. The replacement
allows engineering application of rivets to more complex models of aircraft structures while keeping the computati-
on time acceptably short.
Ntovn je nejastj zpsob zhotoven spoje v leteckm prmyslu. Tento dokument se zabv analzou rznch
ntovch spoj pomoc metody konench prvk (MKP). Ntovan spoje jsou modelovny realisticky s uvaovnm
prvotn fze tvorby ntovho spoje i nslednho zaten ve smyku. Nkter parametry, kter mohou ovlivnit chov-
n takovch spoj, jsou ovovny v dalch MKP analzch. Vhodn naladn podrobn objemov MKP modely
jsou poslze pouity pro definici vhodn nhrady ntovho spoje pomoc jednoduch konen-prvkovch entit.
Takov nhrada umon relnou inenrskou MKP analzu i sloitjch leteckch konstrukc pi pimenm
vpoetnm asu.
Keywords: rivets, FEM, riveted joints.
Fig. 1 Rivet NAS1097AD geometry and a scheme of the small-scale
specimen
The use of two rivets, instead of only one had to lo wer the
effect of a secondary bending of the speci men during a shear
strength test. A dependency of the squeezing force on the rive-
ting tool di splacement was captured duri ng the ri veting pro-
cess. This dependenc y was used later for v erification of the
FEA-model and tuning of the rivet material properties.
FEM model of the riveted joint
Due to requi red precision of the si mulation, the mesh of the
rivet and the surroundi ng area has to be f ine enough (Fi g. 2).
However, such a fine mesh can cause unacceptable ti me con-
sumption of the simulation. The FE-model was finally created
as a half of the real joi nt thanks to the longi tudinal symmetry
of both halves. It enabled us to lower the number of nodes and
elements and helped to speed up the simulation. The FE-model
has about 70000 nodes, and a similar number of elements.
There i s a high number of elements i n the contact, whi ch
makes another b urden i ncreasing the computati on ti me. The
full solution of the described problem took almost 2 da ys on
a single processor machine.
The FE-model w as created from li near 8-noded 6-si ded ele-
ments. MSC.PATRAN was used for pre- and post-processing,
MSC.MARC served as a solver.
The FEM simulation consists of three sta ges (Fig. 3). First,
the riveted joint is created by a prescribed displacement of the
riveting tools. The driven head i s formed duri ng the process,
while the ri vet shank i s expanded. Second, the ri veting tools
are released. The third stage in the si mulation corresponds to
the shear strength test of the ri veted specimen.
The FE-model comprises of several parts with different mate-
rial properties and potential interactions among them.
The sheets and the ri vets are defined as deformable bodies.
Appropriate materi al characteri stics were set. The boundary
conditions are defined by rigid surfaces (Fig. 3). One rigid sur-
face above and one bellow each rivet were modelled for ri ve-
ted joint creation (driven head formi ng), realized by enforced
displacement of the bottom surface. The contact between these
surfaces and rivets was defined as a TOUCHED contact, which
enables both sliding and separation of contacting surfaces. The
contacts between tw o sheets and between sheets and ri vets
were defined also as the TOUCHED contact. The friction coef-
ficient was set to 0.12 on the basi s of Mller's experimental
work [5]. Another surfaces were created at the ends of the she-
ets. They act as a boundary condition during the shear strength
test. In the thi rd phase of the si mulation, one surf ace is fixed
while the second one moves. The surfaces are glued to the she-
ets and cannot separate.
A true stress-strain curve is a suitable solution for the use in
FE-analyses wi th lar ge strai ns. The si mplest w ay of such
a curve definition is the use of apower law equation (Hollomon
law). Hollomon law is defined by coefficients k and n (A and m
respectively in MSC.AFEA):
as a constitutive equati on relati ng the plasti c strai n
pl
and
stress . At the very start, the coefficients were obtained from
[2] and [4]. However, the response of the model based on these
coefficients seemed to be too conserv ative. Results from the
FEM si mulation sho wed much lo wer stif fness and shear
strength of the joint as well. Thus the experimental evidence of
the squeezing force v. riveting tool di splacement dependency
based on the measurements mentioned in the start of this cha-
pter was utilized. The elasto-plastic behaviour was defined by
a table on its basis.
FEM simulation of the riveting process
High plastic straining of the rivet shank is observed during the
riveting process. The dri ven head i s bei ng formed and the
shank expands in the hole. The riveting tool displacement was
set to result in a driven head height 1.7 mm in conformity with
[1]. A very good correlati on between e xperimental data and
data from simulation was reached as is depicted in Fig. 4.
FEM simulation of the shear strength test
A comparison of the e xperiment and the FE-analysi s can be
seen in Fig. 5. The correlation between the experiment and the
simulation is perfect, if one bears in mind that the input data for
elastic-plastic behaviour were deri ved from a simple uniaxial
experiment. The simulation does not reach the e xperimentally
set load-carrying capacity of the joi nt because the calculati on

n
pl
k
Fig. 2 FEM model of the riveted specimen
Fig. 3 Main stages of the simulation
Fig. 4 The dependency of squeezing force on riveting tool
displacement
0
2000
4000
6000
8000
10000
12000
14000
0 0,5 1 1,5 2 2,5 3 3,5
Riveting tool displacement [mm]
S
q
u
e
e
z
i
n
g

f
o
r
c
e

[
N
]
Experiment
MSC.MARC
stops less than 5% belo w it. However, it should be noted that
both lines exceed the theoretical empirics-based load-carrying
capacity [3]. The calculation stop is caused by the fact that the
finite element mesh is at the end of the si mulation so distorted
(especially in the area under countersunk head) that the solver
is not able to f ind a converged solution for further i terations.
The problem can be solved by the use of adapti ve remeshing,
nevertheless this feature was not available in the MSC.MARC
license during the project soluti on. Anyway, the mode of the
deformation near the break (Fi g. 6) and the load le vel similar
to the real load-carryi ng capacity, show that the si militude of
the modelling and the reality is acceptable.
Sensitivity study
The original FEM simulation was performed with friction coef-
ficient set to 0.12 and the riveting tool displacement set to obta-
in the f inal driven head hei ght 1.7 mm. To evaluate the influ-
ence of these parameters, another tw o simulations were per-
formed.
For the first one, the friction coefficient between sheets was
set to 1.0. Ho wever this value is hi gh and ma ybe unrealistic,
the results can sho w, how the beha viour of the ri veted joi nt
changes. Mller i n [5] states, that almost 25% of the load i s
carried by friction between connected sheets.
The second variation of the original model deals with driven
head hei ght. By i ncreasing its hei ght, the shank e xpansion in
the hole will be lower, thus a small decreasing of the joint stiff-
ness can be expected.
The influence of fri ction coefficient can be seen i n Fi g. 7.
The stiffness of the joint is slightly higher, but the increase of
the joint load-carrying capacity is significant. The higher fricti-
on coefficient leads to a better transfer of the carri ed load and
prevents the early establi shment of local plasti c deformation.
However, a sharp edge i n the curve can be seen. The contact
areas separated at thi s step due to deformati on of the sheets
and almost no load can be later carri ed by friction.
The change of the dri ven head height has only small ef fect
on stiffness and carrying capacity of the joint. The plastic stra-
in in the critical area is lower here and the element shape chan-
ges are not so pronounced. The FEA simulation is thus able to
continue up to a final stage of the full load application.
Another types of rivets
The analysis of the ri veted specimen was performed also for
other rivet types. The goal of these analyses is to get informa-
tion about joint behaviour with different geometry and materi-
al of the rivet. FE-models for other rivet types are very similar
to the FE-model for the NAS1097AD-4 rivets.
MS20470AD-4 rivets (ADV) are solid rivets with a protruding
head. The materi al of these ri vets i s the same as for
NAS1097AD-4 rivets - 2117T4 aluminium alloy.
Rivets 1624-0411 (AVEX) are blind rivets with a flat head.
After riveting process, the stem i s pulled from the ri vet shank
and the shank sta ys hollo w. These ri vets are made from
AlMg2,5 aluminium alloy.
Fig. 5 Comparison of a load curve from the experiment
and the simulation
Fig. 7 Comparison of behaviour of the joints with
different parameters
0
500
1000
1500
2000
2500
3000
3500
0 0,5 1 1,5
v [mm]
F

[
N
]
Theoretical carrying capacity
of the joint
Experiment
MSC.MARC
Fig. 6 Stress distribution in the rivet and the surrounding area
at the last converged increment
Fig. 8 FE-models of MS20470AD-4 rivet on the left
and AVEX-4 rivet on right
0
500
1000
1500
2000
2500
3000
3500
4000
0 0,5 1 1,5
v [mm]
F

[
N
]
Theoretical carrying capacity of the joint
Experiment
Original FEM analysis
FEM analysis with high friction
FEM analysis with high driven head
0
500
1000
1500
2000
2500
3000
3500
4000
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8
v [mm]
F

[
N
]
ADV-4
AVEX-4
NAS1097AD-4
Fig. 9 Comparison of different rivet types
The comparison of behaviour of all three types of the rivets can
be seen in Figure 9.
Differences in the behaviour of dif ferent rivet types can be
analysed from Fi g. 9. The joi nt based on ADV-4 ri vets has
negligibly higher elastic stiffness, but the carrying capacity of
the joint is much hi gher than for N AS1097AD-4 ri vets. The
AVEX-4 ri vets ha ve the lo west stif fness and also carryi ng
capacity. This is due to the hollow shank and also material with
lower mechanical properties can be blamed. Stress distribution
in a hollow ri vet and the especi ally hi gh deformation of the
rivet shank are depicted in Fig. 10.
FEM models for fatigue assessment
On the basis of the simple riveted joint model described above,
two more comple x models were created. They correspond to
the e xperimental speci mens e valuated duri ng the IMPERJ A
project related f atigue tests. The load levels for these models
were specified from the experiments.
The first specimen named EVE1 consi sts of two sheets of
thickness 0.8mm from D16C materi al, which are joi ned by 7
NAS1097AD-4 ri vets. Thanks to the symmetry only a half
model could be solv ed. The speci men conf iguration i s v ery
similar to the small-scale specimens described before. In f act,
the small-scale specimens were prepared just for tuning of the
FEA-models static characteristics, so that the response could
be usable also for a detailed fatigue analysis.
The second specimen named EVE2 consists of three sheets
(each of 2.5 mm thickness) from 2024T3 Clad materi al. The
sheets were joined by 7 MS20426AD-6 ri vets. Thanks to two
planes of symmetry , only quarter model can be solv ed (see
Fig. 11).
The initial setup of these analyses i s the same as for pre vious
analysis and the boundary conditions were defined in a similar
way as by previous models.
During the analysi s of EVE1 model, the si milar problem
with too pronounced mesh di stortion occurs as for the si mple
specimen. This is due to the same combination of the rivet type,
rivet diameters and sheet thi cknesses. The analysis of EVE2
model ran without problems. Although the rivet MS20426AD-
6 is also countersunk, the hei ght of the countersunk and the
remaining sheet thickness provide much larger cylindrical area
for the load transfer during shearing than for the EVE1 model.
The stress distribution and deformation of the rivet can be seen
in Fig. 12.
The load v . displacement dependenc y i n i ndividual sheets
can be observ ed in Fi g. 13. A difference in loads carri ed by
upper and lower sheets is obvious. The resulting unsymmetri-
cal load distribution is induced by the countersunk head of the
rivet with worse load-transfer capability.
Simple rivet replacement for more complex
FE-analyses
Common structures in the aircraft industry consist of thin she-
ets connected together by a large number of rivets. Practically,
it is impossible to perform the FE-analyses wi th similar solid
models described before. The models are usually b uilt from
shell elements. This is the reason for our search for a simple
rivet replacement i n such analyses. The results from soli d
models of riveted joints still can serve for determination of the
joint stiffness.
The stiffness of the ri veted joint can also be determi ned by
some of the analyti cal equations mentioned in [3]. The inputs
are usually materi al properti es, thi cknesses of joi ned sheets
and rivet diameter. Nevertheless, these equati ons do not tak e
into account the i nfluence of the ri vet type. Also the v alues
obtained from these equations can differ a lot, see Fig. 14. And
at last, only the li near behaviour is described by these formu-
las, so the w orking di agram part, where the non-li nearities
occur, is not included and the real load-carryi ng capacity can-
not be used, while the ri vets are too stiff there.
Fig. 13 Load/displacement dependency for EVE2 specimen
0
10000
20000
30000
40000
50000
60000
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8
v [mm]
F

[
N
]
total load - middle sheet
upper sheet
lower sheet
Theoretical carrying capacity of
the joint
Fig. 10 Stress distribution and deformation in the AVEX rivet
Fig. 12 Stress distribution and deformation of MS20426AD-6 rivet
in EVE2 specimen
Fig. 11 FEM model of EVE2 specimen
Due to these reasons the stif fness of the ri vet was determined
from the solid model. The model consisting of plates defined by
2D elements (PLA TE) and ri vets replaced by 1D elements
(BUSH) was created for EVE2 speci men. BUSH element i s
a general spring element that can have different stiffness in each
axis of i ts coordinate system. Stif fness can be def ined even as
a non-linear dependenc y between force and di splacement.
Thanks to it, the ri vet behaviour close to reali ty can be si mula-
ted. The bush elements were connected to the surrounding mesh
by RBE (RIGID) elements. The external load was applied by
enforced displacement. The stiffness of BUSH elements is defi-
ned by a table. The joint is asymmetric due to the countersunk
head in one sheet. Different stiffness is thus defined for the joint
between the mi ddle and the upper plates and dif ferent for the
joint between the mi ddle and lo wer plates, see Fi g. 15. The
simulation was performed using MAS.NASTRAN as the solver.
A particular disadvantage of the proposed soluti on may be
the dependency of the non-linear behaviour of BUSH element
on the chosen coordi nate system. There will be no dif ference
for linear behaviour. But for non-li near behaviour of the ele-
ment, if there i s a shear force, which is not parallel to an y of
the axes, an error can be i nduced in the f inal load carri ed by
this element. The coordinate system of the element should be
chosen carefully with respect to the presumed load di rection.
In our case, the correlati on between 3D and si mple model i s
very good, as can be pro ven from Fi g. 16. Ob viously, such
a model cannot describe the stress distribution in the close vicini-
ty to the rivet, but the load transfer and related joi nt behaviour is
close to the reality.
Conclusion
The FE-simulations of various riveted specimens show that it is
possible to simulate behaviour of these joints with an adequate
accuracy. The problem occurring with the great mesh distortion
is noted, but it could be solved by careful definition of the initi-
al element shapes or with the help of adaptive remeshing.
Although the results presented in the paper are very good if
compared with experiments, it should be emphasi zed that the
influence of properly set i nput data, such as correct geometry
or proper material properties, is enormous.
The bi ggest disadvantage of the full soli d FE-models i s the
time required to build the model and to run the analyses. Also the
tuning of the model beha viour to f it to the kno wn experimental
data i s very ti me consumi ng. The use of B USH elements for
more complex models seems to be a promising way of obtaining
the overall load di stribution in the ri veted joint. One of the bi g-
gest advantages is the possibility to define a non-linear behaviour
of the riveted joint.
References
[1] [1] Vlachynsk, J.: Riveted joints; [AEN 5561, zm. B], Evektor,
spol. s r.o., Kunovice, 2006
[2] Vlachynsk, J.: Metal materials; [AEN 2003, zm. B], Evektor,
spol. s r.o., Kunovice, 2007
[3] Chybk, L.: nosnost ntovch spoj; [EVE-VY-20-001-ST, zm.
B], Evektor, spol. s r.o., Kunovice, 2008
[4] Rice, R. C.; Jackson, J. L.; Bakuckas, J.; Thompson, S.: MMPDS-
Metallic Materials Properties Development and Standardization;
Washington, D.C., U.S. Department of Transportation, January
2003
[5] Mller, R. P. G.: An Experimental and Analytical Investigation on
the Fatigue Behavior of Fuselage Riveted Lap Joints; Delft
University of Technology, Delft, 1995
0,00E+00
1,00E-04
2,00E-04
3,00E-04
4,00E-04
5,00E-04
6,00E-04
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
t/D0 [-]
c

[
m
m
/
N
]
Boeing 1
Boeing 2
Grumman
Douglas
Huth
Fig. 14 Comparison of rivet flexibility calculated from different
formulas [3] for the sheet thickness / rivet diameter ratio
-5000
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
-0,6 -0,4 -0,2 0 0,2 0,4 0,6
u [mm]
F

[
N
]
lower joint
upper joint
0
10000
20000
30000
40000
50000
60000
70000
0 0,2 0,4 0,6 0,8 1
L [mm]
F

[
N
]
FEM 3D
FEM 2D
Theoretical carrying capacity of
the joint
Fig. 15 Above Simple FE-model built from 1D and 2D elements
on the left side and the definition of the stiffnesses of BUSH
elements (right)
Fig. 16 Comparison of the 3D and simple FEM model
I. Introduction
During the design and development of a non-conventional pro-
totypical wi ng structure, se veral analyti cal approaches are
used. These i nclude techni ques a vailable i n the li terature to
perform preliminary estimations, followed by a more realistic
theoretical approach that is strictly connected with the specific
structure and has to be constructed and v alidated with experi-
mental/numerical results.
Different theories have been proposed to predict the behavi-
or of a wing structure under certain prescribed loads.
1-9
These
models range from si mplified (lami nated beam) to compre-
hensive (complete wi ng structure),
1
and aim to reduce the
wing to an equivalent element in terms of macroscopic behavi-
or, which then allows conventional analysis methods to be app-
lied. Hwo and Tsai
2
have investigated whole composi te wing
structures with structural characteristics that are constant in the
chordwise direction: these structures are determinate and with-
stand loads that can be e xpressed analytically. In their model,
proposed for the study of a stiffened composite multicell wing,
the wing is treated as a composite sandwich plate; the wi ng
skins, stringers, and spar flanges are modeled as the sandwi ch
faces while the wing spar webs and the ribs are consi dered as
a sandwich honeycomb core. Nevertheless, the f ield of appli-
cation of this method is restricted to full-composi te structures
particularly constrained and wi th no materi al hybridism. This
paper describes a new analytical-numerical approach, whi ch
has a basis in the methodolo gy proposed by Hw o and Tsai2.
Despite the initial close analogy between the two formulations
(consisting in the reduction of the whole wing to an equivalent
element by means of the def inition of a pseudolamina
2
) some
significant differences have been introduced. These differen-
ces and the ne w capabilities allowed by them, consti tute the
core of the NTA/G-update approach. This approach has
been developed and applied to the analysi s of the static struc-
tural behavior of a wing for an aircraft with composite leading
edge and metallic wing box.
In order to comply wi th the EASA/F AA requi rements i n
terms of proof of structure and proof of strength, se veral tests
must be performed on the structure. In parti cular, the cri tical
design loads need to be e valuated usi ng aerodynami c data
from gallery tests and real i nertial loads di stributions. The
shear, bendi ng moment, and torque moment i nternal loads
distributions need to be determi ned and opportunely di screti-
zed, the entity and distribution of the static test load have to be
evaluated and the load application system must be defined and
designed. It should be noted that prior to executing a static test
on a real structure, the de veloper must be able to predi ct the
behavior of the structure i n terms of di splacements and stres-
ses to avoid cracks during the tests, to dimension the levers for
loads application and to correctly posi tion the strain gauges in
the critical areas. Conventional FEM codes may provide realis-
tic results for stresses distributions (even that they are not cost-
effective), but do not properly account for the stiffness and the
deformation of the structure.
The design of a hybrid wing requires a large effort, especi-
ally since methodologies must be de veloped that account for
the above mentioned aspects and can handle the beha vior of
real structures. Secti on II describes the ne w approach and i ts
hypotheses for determining the stiffness matrices of a compo-
site laminate panel equivalent to the hybrid wing and also int-
roduces the G-update procedure. The numerical aspects of the
NTA/G-update approach are presented i n Section III while
comparisons of the obtained results with the ones of the nume-
rical models developed with PATRAN/NASTRAN are descri-
bed in Section IV.
II. STEP 1: Identification of an equivalent element,
hypotheses and methods
This section begins with an analysis of the wing structure that
has been consi dered. The hypotheses that form the basi s for
the novel approach are def ined and then an equi valent set of
stiffness matrices are i dentified. It i s also sho wn the w ay by
which the outcomes of the NTA/G-update approach can be
used to correct and update the shear modulus di stribution
(here follows G-distribution).
NTA/G-update: An Integrated Approach for the
Analysis of Complete Hybrid Metal-composite
Wing Structures
Pevnostn analza kdla z hybridnho materilu kov-kompozit
Emanuela Pietropaoli / The Second University of Naples, Dipartimento di
Ingegneria Aerospaziale e Meccanica, Aversa, Italy
The static behavior of hybrid metal-composite wing structure can be predicted using an innovative integrated ap-
proach named NTA/G-update constituted by the NTA-approach and the G-update procedure. The developed
NTA/G-update approach has been validated comparing its results with the outputs of numerical models used to
simulate static tests upon wings with material hybridism. In the NTA-approach the generic hybrid wing is reduced to
an equivalent composite laminate panel collapsing its structural characteristics into some stiffness matrices and con-
densing the effective chord-wise and span-wise loads into some concentrated forces acting upon the equivalent panel.
This panel is subsequently analyzed using a finite element commercial code, without any additional effort required in
terms of numerical discretization. Taking into account outcomes of the NTA-approach, the classic method for the
shear flows calculation can be updated (G-update procedure): the stresses redistribution among the structural ele-
ments of the wing can be determined and the twist angle at the wing tip can be calculated.
Keywords: hybrid metal composite wing, strength analysis, prediction.
The wi ng analyzed i s desi gned as a conventional double-
spar architecture (see Fig. 1). The main spar defines the boun-
dary of the composite leading edge and, with the rear spar, sus-
tains the bending and shear loads.
The metallic wing box uses a suitable arrangement of stringers
(axial members) and ribs (planar members). It has been realis-
tic assumed that the number of stringers on the upper skin vari-
es along the span due to the presence of the trapdoors for the
wing-integrated fuel tank and the ribs act as anti-sloshi ng ele-
ments (see Fi g. 1). The engi nes are posi tioned close to the
fuselage (after the inner flaps) in pushing configuration, on the
wing trailing edge. The movable surfaces (ailerons and flaps)
have not been consi dered i n the def inition of the equi valent
panel for the wi ng (EASA re gulations request separate stati c
tests for these surf aces, wi th loads specif ically determi ned).
Assuming that the wing is structurally symmetric and that the
acting loads are symmetric, only the semi-wing has been ana-
lyzed in this paper.
The wing has been di vided into structurally homo geneous
zones, which are identified with some key parameters (see Fig.
2): the number of stringers, the thickness and the staking sequ-
ences of the sk ins, the areas of the caps of the spars, and the
thickness of the spar webs. In this way the real structural con-
figuration of the wi ng has been consi dered and some of the
hypothesis of Hwo and Tsai's model
2
have been relaxed.
The number of homo geneous zones depends on the chord-
wise and span-wi se variability of the number and di mensions
of the structural elements, and on the type of material used for
each of them. It should be noted that the term homogeneous-
zone implies that: the number of the stringers, the thicknesses
of the involved elements and the materials are constant in that
zone. Because of thi s local homo genization, an equi valent
plate element can be used to model each zone. However, some
additional considerations are needed in order to reduce a struc-
tural zone i n an element consti tuting the wi ng-equivalent
panel, since both composi te and metalli c skins are i nvolved.
Indeed, if in a zone the skin is made of metallic material, a vir-
tual composite laminate whose material is a metal needs to be
introduced. Alternatively, if the sk in i s made of composi te
materials, no addi tional considerations or hypotheses are nee-
ded, as the ef fective staking sequence is accounted for, within
classical lamination theory
10
. Besides, the ef fects of the stri n-
gers and of the caps of the spars ha ve been introduced adding
to the staking sequence of the plies a pseudolamina
2
.
The above zonal representation allows the hypothesis that the
entire wing is made of composi te materials to be relaxed with
respect to the Hwo-Tsai
2
method. Consistent with this homo-
genization procedure the wing mid-plane may be assumed to
be covered by laminate elements (panels) adjacent to one anot-
her in both the span-wise and chord-wise directions (Figs. 4/5)
Fig. 1 The macroscopical architecture of a hybrid metal composite
wing structure
Fig. 3 Connections between structural elements and layers of the
equivalent panels
Fig. 2 Parameters used to identify each homogeneous zone that
composes the wing
Fig. 4 Visualization of the relationship between the homogeneous
zones and the equivalent laminate panels
Fig. 5 Scheme for the shifting from the middle plane
After thi s, each panel has to be mo ved upward (for the
upper-skin) and downward (for the lower-skin), relative to the
middle plane of the wing (see Fig. 6) with a gap defined by the
averaged local thi ckness of the prof ile. As a matter of f act,
a step-wise profile-thickness distribution along the chord has
been used instead of the polynomial one proposed by Hwo and
Tsai
2
. The size of the gap for each panel (shown as h in Fig. 5)
is taken to be the weighted-average value of the local distribu-
tion of prof ile- thi cknesses consi dering as wei ght the area
underlined by the skin in the homogeneous zone (Fig. 6).
The structural characteri stics of the equi valent panels ha ve
been mathemati cally collapsed i nto three stif fness matri ces
(extensional, coupli ng and bendi ng stif fness, called the
[A,B,D] matrices)
2,10
for each zone. On each semi-wi ng, 33
homogeneous zones that completely i dentify the structural
behaviour of the wing have been individuated; this results in 99
stiffness matrices, which have been calculated using a custom
EXCEL
spreadsheet. The matrices calculation is the ultimate

step needed to i ndividuate a unique global equi valent structu-
ral element (Fig. 7).
Since the structural characteri stics of the whole wi ng ha ve
been collapsed i nto the stif fness matri ces of an equi valent
composite laminate panel (Fig. 7) by using the NTA-approach,
these matrices can be used to e valuate the di stribution of the
shear modulus ( G) i n chord-wi se and span-wi se di rections.
These values of G can be taken as inputs for the application of
the shear flows method ( G-update procedure) avoiding in this
way to consider a unique and unrealistic local value of G as far
as the stresses calculation and twist rotations are concerned.
Since the 99 [A], [B], [D] matrices for the 33 homogeneous
zones are kno wn from the NTA-approach, the correspondi ng
compliance matrices can be simply calculated (see Ref. [12]).
Referring to classical theory (see Ref. [12]) for the closed sec-
tion orthotropic beams (single cell), the shear modulus G can
be expressed as:
(eq. 1)
where
66
,
66
,
66
are the terms of the compli ance matrices
estimated for each zone (including the ones covering the wing
box) and t is the local thickness of the skin.
It is important to note that, as consequence of the steps of
the NTA, the formula in the (eq. 1) becomes applicable to the
calculation of the shear modulus ( G) not only for the zones
corresponding to the composi te leading-edge but also for the
ones of the metallic wing box (in which the G value could be
different from the one of the aluminium) (see Fig. 9). With the
formula def ined in the (eq.1), 33 dif ferent values for the G
modulus have been calculated upon the wi ng: the shear flows
evaluation (see Ref. [13]) and the twi st angle calculation have
been performed consi dering three dif ferent v alues for the
shear modulus in each section (G1,G2 and G3 in Fig. 9), (see
SECTION IV).
III. STEP 2: Numerical model, external constraints
and loads
As menti oned abo ve, the outcomes of STEP 1 are the 99
[A,B,D] matri ces and 33 v alues def ining the shear modulus
distribution. While the 99 [A,B,D] matrices are directly impor-
ted into a finite element commercial code, the G-update valu-
es are separately utili zed into the shear flo ws calculation per-
formed using a custom EXCEL w orksheet. Whereas for the
numerical i mplications of the NT A-approach details will be
given in this section, the results of the G-update procedure will
be shown in the SECTION IV.
Starting from the equi valent flat plate, a finite element
model has been b uilt i n ANSYS
by usi ng the ei ght nodes

layered structural shell element SHELL99. This element type
makes use of the materi al matrices [A,B,D] computed outside
of the ANSYS program,
11
rather than property def initions for
all of the la yers (material, fibre orientation, and thickness). In
this case, the materi al matrices components are establi shed as
real constants associ ated with one or more elements. The 33
zones i ndividuated for the materi al matri x computati ons i n
EXCEL become an equi valent number of geometri c enti ties
(areas) in ANSYS (see Fig. 9). A correspondence is established
between all the finite elements belonging to each of these areas
and the defined real constants sets.
The static scheme for a simply supported flat plate has been
adopted. Since the structure i s symmetric with respect to the
plane YZ (A
0
) (Fig. 10), only a semi-wing has been considered
in the numeri c model, and symmetry condi tions ha ve been
imposed. At section A
1
of Fig. 10, two cylindrical hinges exist
at the connection between the wing and fuselage structure (all
A
b
h=A/b
X
Fig. 6 Stepwise distribution of the thickness along the chord
Fig. 7 Representation of a real wing as equivalent composite
laminate panel composing the equivalent elements
Fig. 8 Scheme for the identification of the Gi corresponding
to each section

66
2
66
66 66 2
66
1
zone i
zone i
zone i
zone i v
v zone i
zone i
zone i
G
t

the displacements have been f ixed equal to zero , so that only
rotation around the Y axis is permitted).
IV. STEP 3 Comparison with the results
of a detailed FEM model
In order to esti mate the potenti alities of the no vel i ntegrated
approach (NTA/G-update approach) the results respecti vely
of the NTA approach and of the G-update procedure have been
compared wi th the ones of a detailed PATRAN/NASTRAN
finite element model. The distribution of loads consi dered for
the comparison is the most cri tical loading condi tion chosen
among the ones def ined by the appli cable regulation require-
ments (FAA/EASA) . It i s worth noting that the loads appli ed
onto the models, reproduce the e xecution of a static test on
a full scale component. A symmetrical un-flapped positive gust
was chosen since it was critical in terms of shear and bending.
The stati c test load i s composed of concentrated forces
acting on each of the patches posi tions (see Fi g. 11). For the
load application on the engi ne zone, a dedicated system of
levers has been used (see Fi g. 11).
The stati c test loads ha ve been i mported as concentrated
loads and applied on the equivalent flat plate as defined in the
SECTION II; the effect of the concentrated loads in the engi-
ne zone has been tak en into account by means of multi poi nt
constraints (MPC) rigid elements (see Fig. 12).
The maximum displacements at the free end of the wing and
the twist angle outcomes of the NTA/G-update model (see
Fig. 13), have been compared with the displacements predicted
by the results from a PATRAN/NASTRAN
detailed f inite
element model.
The compari son among the numeri cal results from the tw o
models and the outcomes of a full scale test upon the hybri d
structure permits to state that the NTA model is able to predict
the real beha viour of the structure wi th li mited di screpancy.
Also, it has been found that the time requested for the analysis
and meshing in the NTA/G-update approach is strongly redu-
ced with respect to the one in the conventional numerical app-
roaches.
STEP 4 Shear flows calculation and comparison
Utilizing the classical shear flows calculati on (see Ref. [13])
and taking into account the G-update defined in the Section II,
the di stributions of Fi g. 14 for the shear flo ws i n the mai n
structural elements have been obtained.
In order to esti mate the effect of the G variability upon the
shear flows distributions, the shear flows calculation has been
repeated assuming a constant value for shear modulus of the
whole leading edge precisely equal to the G
12
of the composi-
te material: The outcomes shown in Fig. 15 have been deter-
mined.
The comparison between the outcomes of the shear flows
calculation (see Figs. 14 and 15) highlights that the choice of
the G-values has a not ne gligible ef fect on the soluti on.
Fig. 9 Real constant set i associated to the area Ai

A0 Ux=0; Uy=0; ROTY=0; ROTZ=0

A1 Ux=0; Uy=0; Uz=0; ROTX=0; ROTZ=0

Fig. 13 Points for the displacement evaluation in the tip end section
at limit load
A
B
C
Fig. 11 External loads
distribution
Fig. 12 Applied loads
Table 1 Boundary conditions
Fig. 10 Graphical representation of the chosen boundary conditions
Those outcomes, e xpressed i n term of stresses, ha ve been
also compared wi th the stresses di stribution obtained from
the detailed FEM model of the whole wi ng: the comparison
among those data sho ws di screpancies between the NTA-
predicted and PATRAN/NASTRAN-calculated values b ut i t
confirms the effectiveness of the G update procedure as far
as the predi ction of the stresses redi stributions among the
structural elements i s concerned. The application of the G
update pr ocedure inside the classi cal theory for the twi st
angle evaluation (see Ref. [13]) also permits to estimate the
twist angle distribution along the span without any additional
computational efforts.
It can be seen that the NT A/G-updated approach approx i-
mate well the real twi st angle v alue giving the boundari es of
the interval in which it can be found.
V. Conclusions
Structurally homogeneous zones of an hybri d metal-com-
posite wing have been identified using the arrangement and
type of i ts structural elements. After performi ng thi s
homogenization procedure, the wi ng mid-plane has been
treated as consti tuted of equi valent lami nate panels adja-
cent to one another in the span-wise and chord-wise directi-
ons.
The structural characteri stics of the equi valent panels
have been mathemati cally collapsed i nto three stif fness
matrices, representi ng e xtensional, coupli ng, and bendi ng
stiffnesses
2,10
for each zone; 99 matrices have been calcula-
ted in all using an EXCEL
spreadsheet. Starting from the

obtained equi valent flat plate geometry , a finite element
model was built in ANSYS
by using the eight nodes laye-

red structural shell element SHELL99. The most cri tical
loads were applied to the wing as concentrated forces on the
equivalent plate, and the deflecti on value at the ti p of the
wing and the twi st angle were calculated and compared
with the ones coming from a static test on a full scale wing.
The outputs from the NT A have been utili zed to correct
the results of the classi cal approaches for the shear flo ws
calculation and for the e valuation of the twi st angle. Inde-
ed, it has been recognized that the main cause of the lack of
precision of the results provided by the classical shear flows
calculation was in the assumptions conventionally made for
the shear modulus calculati on. Hence the NT A approach
has been used to determi ne a more realistic distribution for
the G modulus inside the whole wing (G-update procedure):
33 dif ferent v alues for the G modulus were determi ned
upon the wing and the shear flows estimation was repeated
considering three di stinct values for the shear modulus i n
each section. These values of G have been considered not
only for an accurate esti mation of the angle of twi st at the
tip section but also to update the stresses calculati ons.
It has been shown that the NTA/G-update approach is able
to predict the real beha vior of the structure wi th a limited
discrepancy and also the ti me requests for the analysi s and
meshing is strongly reduced. The approach presented here is
more general and applicable to the prediction of a wing struc-
ture under certain prescribed loads compared wi th conventi-
onal theory, without requiring any limitations on the type of
structural arrangement. This method allows the determinati-
on, with negligible computational costs, of the macroscopi c
behavior of a full scale hybrid structure with an applied gene-
ric load, using any boundary condi tions. Furthermore i t also
allows a partial improving of the effectiveness of the analyti-

Shear flows [kg/mm]
-2
-1
0
1
2
3
4
5
6
7
0 1 2 3 4 5 6 7
span [m]
Main spar
Rear spar
Upper skin 1-2
Upper skin 2-3
Upper skin 3-4
Upper skin 4-5
Lower skin 6-7
Lower skin 7-8
Lower skin 8-9
Lower skin 9-10
Leading edge
Fig. 14 Shear flows updated with variable values for the G
Shear flows [kg/mm]
-2
0
2
4
6
8
10
0 1 2 3 4 5 6 7
span [m]
Main spar
Rear spar
Upper skin 1-2
Upper skin 2-3
Upper skin 3-4
Upper skin 4-5
Lower skin 6-7
Lower skin 7-8
Lower skin 8-9
Lower skin 9-10
Leading edge
Fig. 15 Shear flows distribution with G = G
12
Angle of twist (deg)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 1000 2000 3000 4000 5000 6000
span(mm)
Fig. 16 Twist angle distribution along the span updated
with variable values for the G

Angle of twist (deg)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1000 2000 3000 4000 5000 6000
span(mm)
Fig. 17 Angle of twist calculated with G = G
12
cal shear flows calculation when the twist angle estimation is
concerned (G-update procedure).
Acknowledgments
This w ork has been carri ed out i n collaborati on wi th the
Second University of Naples (Aversa, Italy) wi thin the Ph.D.
work of Emanuela Pietropaoli.
1 Introduction
The project Research of Material Properties in Terms of Frac-
ture Mechanics Approaches and its subproject Development
of Testing and Data Evaluati on Methodology with a View to
Determining Non-standar d Mater ial Char acteristics include
the preparation of methods for non-standard or uni que testing
and assessment of materi al properties. In the area of the f ati-
gue properties of materi als and structural nodes, the f atigue
properties are investigated throughout the enti re life spectrum
as a standard. Procedures defined in the past have been estab-
lished for such tests. Ho wever, it is necessary in certain cases
to prove the life of a structural component using a predefined
number of load oscillati ons. Different procedures ha ve to be
used in such cases. Such testing procedures and methods have
not been used at VZLU (Aeronautical Research and Testing
Institute).
Table 2 Comparison between the NTA and Updated shear flows
theory in terms of twist angle calculation
References
1
Chyanbin Hwu and H. S. Gai: Vibration Analysis of
Composite Wing Structures by a Matrix Form Comprehensive
Model; AIAA Journal, Vol. 41, No. 11, 2003,pp. 2261-2273
2
Hwu, C., and Tsai, Z. S.: Aeroelastic Divergence of Stiffened
Composite Multicell Wing Structures; Journal of Aircraft,
Vol. 39, No. 2, 2002, pp. 242-251
3
Librescu, L., and Khdeir, A. A.: Aeroelastic Divergence of
Swept-Forward Composite Wings Including Warping Restraint
Effect; AIAA Journal, Vol. 26, No. 11, 1988, pp. 1373-1377
4
Chandra, R., Stemple, A. D., and Chopra, I.: Thin-Walled
Composite Beams Under Bending, Torsional, and Extensional
Loads, Journal of Aircraft, Vol. 27, 1990, pp. 619-626
5
Volovoi,V. V., and Hodges, D. H.: Single- and Multicelled
Composite Thin-Walled Beams, AIAA Journal, Vol. 40, No.
5, 2002, pp. 960-965
6
Diederich, F. W., and Budiansky, B.: Divergence of Swept
Wings; NASA TN 1680, Aug. 1948
7
Karpouzian, G., and Librescu, L.: Comprehensive Model of
Anisotropic Composite Aircraft Wings Suitable for Aeroelastic
Analyses; Journal of Aircraft, Vol. 31, No. 3, 1994, pp. 703-
712
8
Librescu, L., and Thangjitham, S.: Analytical Studies on
Static Aeroelastic Behavior of Forward-Swept Composite Wing
Structures, Journal of Aircraft, Vol. 28, No. 2, 1991, pp. 151-
157
9
Hu, J. S., and Hwu, C.: Free Vibration of Delaminated
Composite Sandwich Beams; AIAA Journal, Vol. 33, No. 9,
1995, pp. 1-8
10
Jones R. M.: Mechanics of Composite Materials; Taylor &
Francis
11
ANSYS Manual ver. 10
12
Kollar L. P., Springer G.S.: Mechanics of composite
structures, Cambridge University Press, 2003
13
Bruhn E. F.: Analysis and Design of Flight Vehicle
Structures; Jacobs Pub, June 1973
Incremental STAIRCASE Test Method
Stupovit metoda zkouen
Ing. Tom Blaina / VZL, Plc., Prague
This study provides an analysis and suggests a methodology for conducting probative fatigue tests for the required
life. The methodology has been verified in fatigue tests using aluminium casting parts made of the AlSi9Cu3 alloy. The
obtained data and test specimens showing signs of breakage may be used for further analyses. The study aims at pro-
posing a methodology for tests which have to show a certain level of load for the defined life, and to verify the metho-
dology by experiments.
V pedloen prci je proveden rozbor a vypracovna metodika pro realizaci prkazov navov zkouky na poado-
vanou ivotnost. Metodika byla ovena pi navovch zkoukch s hlinkovmi odlitky ze slitiny AlSi9Cu3.
Namen data a poruen zkuebn tlesa lze vyut pr o dal analzy. Clem prce bylo zpracovat metodiku pro
zkouky, kdy je nutn prokzat hladinu zaten pro definovanou ivotnost a nsledn tuto metodiku experimentln
ovit.
2 Marks and symbols used
Fa Total load range corresponding with reference load Fn
[kN]
Fd Fatigue limit for NC oscillations [kN]
Fd
0
Initial load [kN]
Fm Mean load corresponding with reference load Fn [kN]
Fn Required reference load (Fn = Fm Fa) [kN]
k Coefficient to obtain the mean life value (K=Fd
0
*) [ ]
Ne Test specimen group (number of parts tested) [ ]
Nc Number of load oscillations [ ]
m Mean life [ ]
q Relative variance [ ]
s Standard deviation [ ]
F Load increment [kN]
* Value added to initial load Fn [ ]
This study provides an analysi s and suggests the methodo-
logy for conducting probative fatigue tests for the required life.
The methodology has been v erified in fatigue tests using alu-
minium casting parts made of the AlSi9Cu3 alloy. The obtai-
ned data and test specimens showing signs of breakage may be
used for further analyses.
The study aims at proposing a methodology for tests which
have to show a certain level of load for the defined life, and to
verify the methodology by experiments.
3 Definition of fatigue test methodology
3.1 Test Method
The method comprises of three phases:
- Preliminary phase
- First phase
- Second phase
The objective of the preliminary phase of the e xperiment is to
obtain an initial estimate of the expected and required life, res-
pectively. The objective of the first phase of the experiment is,
provided that the load presumed in the preliminary stage is cor-
rect, to determine the actual level of load for the required life.
In practice, it is achieved by gradually testing specimens to the
required number of oscillations. Depending on the result of this
test, the load for another speci men is determined. The load is
increased or decreased usi ng predef ined i ncrements. The
objective is to obtai n the i dentical number of speci mens with
and without breakage. In the last phase of the e xperiment, all
the speci mens wi thout breakage are tested until breaka ge i n
order to determine the variance characteristics.
3.1.1 Preliminary Phase
The objective of the preliminary stage is to define the initial test
level Fd
0
*. This decision can be made in two methods:
1) Calculation the a verage load le vel for the requi red
number of load oscillati ons can be esti mated by calcula-
tion. From the result obtained by calculation for load Fn,
for which the speci men has been desi gned, the requi red
load Fd
0
= KxFn corresponding with the mean value m*
of the same part. The i nitial i ncrement i s calculated,
F*=Fd
0
*=K. This method requi res basi c knowledge of
the material characteristics and knowledge of the specific
stress concentration values in the part.
2) Test the Locati test. This method i s used when there
are no detailed materi al characteristics or stress concent-
ration v alues a vailable. In thi s method, the i nitial load
level i s selected based on e xperience, i nformation from
literature, or by estimate. The fatigue test is initiated using
the selected load level. After a predefined number of load
oscillations Nc (200,000 to 300,000 oscillati ons are
recommended), the load is increased using the F* incre-
ments (the recommended value is 10 per cent of the initi-
al load value) until breakage. The initial load level in the
test F*=F/Fn is usually selected as a value 20% lo wer
than the e xpected f atigue limit the ti me life, if we
can define such a load level. Based on the course of the
test (load, number of oscillati ons), the equi valent or
expected level of load for the requi red life is determined.
The load le vel for the ne xt phase of the e xperiment i s
determined usi ng the Whler curve and Mi ner's rule of
cumulative dama ge. Thus we determi ne the i nitial load
level (degree) Fd
0
* .for the f irst phase of the test. This
value is automatically calculated from the e xperimental
data in the excel spreadsheet see Appendix. The ideal
basis for the determination of the load level is the known
fatigue curve, which adds and deducts the life of the spe-
cimen load at the individual levels.
3.1.2 First Phase
The objective of the first phase is to determine the preliminary
life values m*, variance q, equivalent load F by testing speci-
men using a single load level. The procedure of the incremen-
tal test is based on the knowledge of:
- the initial status Fd
0
* = Fd
0
/Fn.
- the knowledge of the number of Nc_st oscillations, during
which the specimen is tested at each load level.
The recommended change in the load of the specimens is 10%
of Fd
0
*. This 10% change i s not def ined as f ixed, and can be
modified depending on the requirements and objectives of the
experiment. Ho wever, the i ncremental change must remai n
constant after the experiment is launched.
For e xample, the automoti ve i ndustry def ines a standard
requirements for the certification of life at a level of load using
Nc_st = 1,000,000. By a greement with the client or calculati-
on officer, this number may be reduced to 500,000 oscillations.
The tested life is always related to a specific structural element
or part.
The i ncremental procedure i s conducted accordi ng to the
flow chart provided in Fig. 3-1.
After the determination of the regular test result using this met-
hodology, it is necessary to test at least 6 speci mens, and the
test results (test completed with breakage or without breakage)
must meet the following criteria:
If 6 specimens are tested:
3 without breakage and 3 with breakage the incremen-
tal test has been completed.
In any other case the test continues.
Fig. 3.1 Staircase test procedure
The test has been completed i n case of an y result. If the
test result rati o between speci mens wi th breaka ge and
without breakage i s less than or equal to 3/8 or hi gher
than or equal to 8/3, the result must be discussed with the
client, process engineers or calculation officers. This situ-
ation ma y occur ei ther i n case of si gnificant v ariance
(which may point, for e xample, to an i ncorrect or non-
observed manufacturing technology) or if the i nitial load
level has been calculated i ncorrectly (the calculati on i s
not reliable).
It is a general requirement that the number of speci mens with
and without breakage be approximately identical were 6 to 10
specimens are tested. Such test results will then pro vide an
adequate approximation to the i deal load and a verage life m*
values.
It may be generally stated that si x specimens are suf ficient
to conduct the incremental test if the proper technological pro-
cedures and a correct calculation method are followed.
In specific cases, such as if three consecuti ve breakages or
non-breakages occur duri ng the test after the load has been
changed (this situation may occur if the i nitial load level Fd
0
*
has been incorrectly defined in the first stage of the experiment
based on the Locati test i n the preli minary phase), the Locati
test must be changed to an incremental test see the prelimi-
nary phase, wi th an i ntact speci men. In an y case, such an
occurrence shows certain non-homogeneity of the deli very or
series, and the results must be subject to further analysis, or, if
applicable, the test must be repeated with a new value of Fd
0
*.
3.1.3 Second Phase
The values of mean life m*, variance q and equivalent load F
are determined in this phase. This process of obtai ning more
accurate results is performed using the specimens tested in the
first stage of the i ncremental test and sho wing no breaka ge.
The second phase of the test usually follo ws the full comple-
tion of the first phase, after the number of specimens tested has
met the requirements described in the previous section. It may
be, however practical to launch the second phase of the test i n
the course of the first stage (e.g. to avoid the repeated removal
and re-installation of specimens without breakage from/in the
testing equipment. The test of each undamaged specimen con-
tinues in a manner similar to the Locati test described abo ve
(preliminary phase), using the following new parameters:
If the requi red number of oscillati ons in the f irst phase di d
not result i n the breaka ge of the speci men at the gi ven load
level, the specimen test continues using a load level increased
by 10 per cent of Fd
0
*. The same number of oscillations is usu-
ally performed at this increased load level (e.g. Nc=1,000,000)
as that used i n the first stage of the e xperiment. Alternatively,
the number of oscillations (Nc) may be reduced as much as to
200,000 by a greement wi th the cli ent. If the i ncreased level
does not result i n breakage, the load i s again increased by 10
per cent of Fd
0
* and the test continues. This procedure is furt-
her followed until specimen breakage.
3.2 Test Evaluation
The values of the mean life m* and v ariance q (or standard
deviation, respectively) are obtained from the tests conducted
with a given number of specimens. The test results are accep-
ted if the values of m* and q are below the influential curve of
the relati ve mean life a gainst the relati ve standard de viation
corresponding to the actual number of speci mens tested, or
below the stabili sed curve defined by the manuf acturer if the
manufacturer has a vailable the v ariance characteristics of the
production or if the number of specimens tested is higher than
50.
The results of the Locati test performed i n the preli minary
phase (speci men 0) are di sregarded for the purposes of the
determination of the mean v alues and the standard de viation.
The value of m* is determined only based on the results of the
incremental test. The standard deviation s* is therefore calcu-
lated from the equal loads Fd*. The v ariance i s calculated
using the following relation: q = s*/m*.
If the relative variance is unknown, then:
A reduced number of tests (e.g. between 6 and 15) results
in a value wi th a low-reliability relati ve variance q and
requires a higher mean v alue m* of the correspondi ng
parts.
If the relative variance is known, then:
The designer may optimise the dimensioning of the parts
based on thi s relati ve v ariance q. For that reason, i t i s
important to create and re gularly update databases of q,
and to check the products with a view to feedback to pro-
duction.
The graph i n Fig. 3-2 i s automatically generated i n the Excel
spreadsheet using the data measured during the test. Following
is the description of the graph. The test was conducted using 6
specimens, which resulted in m*=1.72 and qp=0.10 (point 1 in
Fig. 3-2). In order for the test of these 6 speci mens to be accep-
ted, the value of m* for qp=0.10 must be at least 2 (poi nt 2 in
Fig. 3-2). This i s not the case here, hence thi s result i s not
accepted. On the other hand, if the materi al supplied has a vari-
ance of less than qs=0.117 the result i s accepted because the
point [1.72; 0.117] (poi nt 3 i n Fi g 3-2) i s situated below the
stabilised curv e (product curv e). In thi s case, the result i s
dimensioned to the minimum.
Introduction
The serviceability in the ice condition is one of important eva-
luation criteria of small commercial airplanes. Problems of ice
accretion are under research of InICE project.
The aim of project is creation of analytical and experimental
surrounding to integrate influence of ice accretion on airplane
aerodynamic characteristics into all sta ges of de velopment of
small commercial airplane, certification process including. This
background will allow design an airplane with regard to duty
in expected ice conditions since start of design. It leads finally
to increasing safety of air transport and serviceability of an air-
plane.
The results of tunnel tests performed i n the frame work of
InICE project were used to v erify results of calculati on met-
hod and as input data for other calculations.
Experiment description
The experiments were performed on the e xperimental model
of a small commerci al aircraft (Fig. 1). It i s twin-engine high
wing airplane with T-tail surfaces.
The measurements were performed i n 3 m-diameter lo w-
speed wi nd tunnel at VZL, the achi eved Re number w as
0.710
6
at reference dimension c
MAC
= 0.2 m.
In report [3] the form, dimensions and positions of ice imita-
tion were set by selecti on of the most adv erse cases from cal-
culated results done parti cularly at AVR [1] and i n EVEK-
TOR [2] for crui se, takeoff and landi ng configurations of ai r-
plane with flaps deflected to 0, 20 and 38 in glaze and rime
ice. For experiment were chosen cases of glaze ice accretion as
the most unfavourable.
The measurements were aimed to evaluate effect of ice acc-
retion on lateral and longitudinal characteristics of the airplane
model with zero sideslip angle.
The programme of experiment involves choice of ice accre-
tion created in cruise configuration flight and in flight with the
flaps deflected.
Fig. 3-2 Acceptance graph
If the v ariance of the materi al suppli ed i s lo wer than
qp=0.10, the result i s accepted because the poi nt [1.72;
0.10] i s clearly belo w the curv e This poi nt ma y be
moved back to the value of 1.5 (point 4 in Fig. 3-2). This
may be used to save weight.
3.3 Test Discontinuation Criteria
The test must be stopped upon the occurrence of any of
the following events:
Detection of the first visible fatigue crack on the spe-
cimen;
Loss of stiffness of the specimen indicated by the test
equipment,
Clear breakage of the specimen;
F ailure of contact between at least one attachment
point on the test equipment and the specimen.
4 Conclusion
The incremental test under discussion is used to determine the
time life of structures and thei r components. The result is the
value of fatigue load to which the structure or part i s resistant
for a defined number of oscillations.
Zkouky tvorby nmrazy v aerodynamickm tunelu
Ing. Nikita ikovsk / VZL, Plc., Prague
In frame of cooperation on project InICE, the wind tunnel measurements were realized at VZLU to test influence of
ice accretion on aerodynamic characteristics of a small commercial airplane. The presence of ice is simulated on the
model by the ice accretion models placed on wing and tail surfaces.
Keywords: InICE, simulation of ice accretion, wind tunnel test..
Nomenclature
c
L
lift coefficient
c
D
drag coefficient
c
D0
drag coefficient at c
L
= 0
c
m
pitching moment coefficient
c
MAC
mean aerodynamic chord
Re Reynolds number
angle of attack related to fuselage base plane
FL
angle of lift flaps deflection
The ice accretion of cruise configuration corresponds to fli ght
at high speed (Fig. 2).
The results of ice accretion calculations showed that the ice
accretion gre w up at lo wer speed of fli ght on both leadi ng
edges of wi ng and i n the slot of deflected flaps on the upper
side of nose of flap (Fi g. 3).
As the time of landing flaps exposition through the landi ng
manoeuvre is too short for growing up of the ice accretion, the
creation of ice on landing configuration is practically impossib-
le. That's the reason why the iced landing configuration wasn't
included in test programme. In contrary, iced takeoff configu-
ration with flaps deflected not only i n takeoff position (Fig. 3)
but also in landing position was included in measurement pro-
gramme.
In the same way the iced cruise configuration with the flaps
deflected into the tak eoff posi tion was included into the test
programme.
That way the experiment programme includes even the prac-
tical duty cases where after ice accretion the flaps are removed
to other position.
Results
The cruise configuration ice accretion (Fig. 6) markedly lowers
the maximum lift coefficient (to 43%) and i ncrease the c
D0
by
60%, but changes of lift curv e slope and zero lift angle of
attack are negligible whether at cruise flap position or at take-
off position.
The fundamental porti on of the change of c
D0
corresponds
to the installation of tail ice accretion model (Fig. 8). It is pos-
sible to f ind i t from dif ference between curv es TAILS ICE
ONLY, NO ICE and ICE ON, whi ch represents i ce accretion
on tail surfaces and on wing including slot ice.
The ef fect of i ce accreti on of tak eoff conf iguration at
KL
= 20 (Fig. 7) is similar to the ice effect of cruise configu-
ration at
KL
= 0 (Fig. 6), but loss of maximum lift coefficient
is relatively lower. As the presence of slot accreti on at takeoff
flaps position has no effect on characteristics (Fig. 8), it is clear
that the lower loss of max imum lift coef ficient due to tak eoff
configuration relates to the less unfavourable form dimensions
and position of ice accretion on leading edge and to later flo w
separation.
On the contrary, the measurement of tak eoff configuration
accretion wi th flaps mo ved to landi ng posi tion demonstrates
the considerable change of lift coef ficient not in maximum lift
coefficient zone only b ut in whole range of angles of attack,
and decrement of lift curve slope (Fig. 7).
The disturbance of flow field over configuration with flaps
Fig. 2 (Left) Ice accretion of
cruise configuration
Fig. 3 (Right) Ice accretion
of the takeoff configuration on
leading edge and in the flap
slot
Fig. 5 The model of ice on
tail surfaces of the airplane
model
with takeoff accretion deflected on landi ng posi tion leads to
decrease of the lift curve under the lift curve corresponding to
the flaps wi th takeoff accretion deflected on tak eoff position
(Fig. 7).
To prove portion of slot i ce accretion on total changes of
characteristics due to ice, the slot ice accretion was demounted
and measurements of both tak eoff and landing flaps positions
were performed.
Results of measurements i ndicate that the ef fect of ice acc-
retion i n slot of tak eoff conf iguration (Fi g. 8) i s relati vely
small, but in the case of iced flaps removed to landing position
the effect of slot ice accretion is strong (Fig. 9).
A question remains, if the ice creation on the nose of flap in
reality after reachi ng of landi ng flaps posi tion would stay on
and would continue growing, or if the ice creation would be in
new flow field broken down.
Conclusions
The results confirmed strong reduction of maximum lift coef-
ficient due to ice accretion effect. Simultaneously it was found
that in case of tak eoff icing at lo wer air speed the mo ving of
flaps to landing position can lead to hi gh and unexpected loss
of lift. It could be especially in landing maneuver fatal. In that
case it seems better to recommend k eeping the tak eoff flaps
position for landing.
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-15 -10 -5 0 5 10 15 20
[ ]
c
L

[
1
]
NO ICE, FL0
NO ICE, FL20
ICE ON, FL0
ICE ON, FL20
-1
-0,5
0
0,5
1
1,5
2
2,5
3
0 0,1 0,2 0,3 0,4
c
D
[1]
c
L

[
1
]
NO ICE, FL0
NO ICE, FL20
ICE ON, FL0
ICE ON, FL20
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-0,4 -0,2 0 0,2 0,4
c
m
[1]
c
L

[
1
]
NO ICE, FL0
NO ICE, FL20
ICE ON, FL0
ICE ON, FL20
Fig. 6 The ice accretion of cruise configuration effect at cruise/ takeoff flaps position
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-15 -10 -5 0 5 10 15 20
[ ]
c
L

[
1
]
NO ICE, FL20
NO ICE, FL38
ICE ON, FL20
ICE ON, FL38
-1
-0,5
0
0,5
1
1,5
2
2,5
3
0 0,1 0,2 0,3 0,4
cD [1]
c
L

[
1
]
NO ICE, FL20
NO ICE, FL38
ICE ON, FL20
ICE ON, FL38
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-0,4 -0,2 0 0,2 0,4
cm [1]
c
L

[
1
]
NO ICE, FL20
NO ICE, FL38
ICE ON, FL20
ICE ON, FL38
Fig. 7 The effect of takeoff configuration ice accretion at takeoff/landing position of flaps
Acknowledgement
The measurements are part of FT-TA4/044, Integration of ice
accretion influence on flight characteristics and performance in
all stages of development of small transport aircraft including
certification process according to CS-23 and FAR 23 - InICE
project supported by the Ministry of Industry and Trade of the
Czech Republic.
References:
[1] Hoen, B.: Stanoven tvar kr itick nmr azy na no snch
a stabilizanch plo chch e xperimentlnho letounu ;
(DVZ_E21_08); H AV R, Praha, 2008
[2] Piteck, P.: Stanoven tvaru nmrazy pro podmnky "SLD"
na vybr anch ezec h kdla e xperimentlnho letounu ;
(DVZ_E210_08); EVEKTOR, Uhersk Hradit, 2008
[3] ikovsk, N.: prava model pro men v a erodynamic-
km tunelu; (DVZ_E24_08), VZL: R-4302), Praha, 2008
[4] ikovsk, N., Jandov, K: Men vlivu nmrazy na aero-
dynamick charakteristiky letounu na modelu celho letounu
a mo delu o casnch plo ch v a erodynamickm tunelu ;
(DVZ_E25_08), VZL: R-4486, Praha, 2008
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-15 -10 -5 0 5 10 15 20
[ ]
c
L

[
1
]
NO ICE
TAILS ICE ONLY
ICE ON
SLOT ICE OFF
-1
-0,5
0
0,5
1
1,5
2
2,5
3
0 0,1 0,2 0,3 0,4
c
D
[1]
c
L

[
1
]
NO ICE
TAILS ICE ONLY
ICE ON
SLOT ICE OFF
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-0,4 -0,2 0 0,2 0,4
c
m
[1]
c
L

[
1
]
NO ICE
TAILS ICE ONLY
ICE ON
SLOT ICE OFF
Fig. 9 The effect of slot ice accretion, takeoff configuration ice accretion,
FL
= 38
Editorial Note: Some formerly mentioned pictures are
printed on the outer ba ck page of thi s issue in colour f or
better clarity.
-1
-0,5
0
0,5
1
1,5
2
2,5
3
0 0,1 0,2 0,3 0,4
c
D
[1]
c
L

[
1
]
NO ICE
TAILS ICE ONLY
ICE ON
SLOT ICE OFF
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-0,4 -0,2 0 0,2 0,4
cm [1]
c
L

[
1
]
NO ICE
TAILS ICE ONLY
ICE ON
SLOT ICE OFF
-1
-0,5
0
0,5
1
1,5
2
2,5
3
-15 -10 -5 0 5 10 15 20
[ ]
c
L

[
1
]
NO ICE
TAILS ICE ONLY
ICE ON
SLOT ICE OFF
Fig. 8 The effect of slot ice accretion, takeoff configuration ice accretion,
FL
= 20
1 Introduction
Electronic sensors of physical quantities are main and the only
link between the real world and the electronic control system.
These sensors convert a real physical quantities into electrical
signals in a readable form for control systems. Due to the i nc-
reasing time requirements of test preparation and reducing the
cost of i ts preparation it was necessary to de velop a new sys-
tem for calibrati on of force and di splacement sensors, whi ch
are the main parts of control loops of ai rcraft structures tests.
Naturally they meet the requi rement of metrolo gy continuity
of the sensor on calibration standard.
This arti cle deals wi th the de velopment and i mprovement
calibration methods of force and di splacement transducers at
Strength of structures division of the VZL.
2 General calibration overview
The sensor calibration is a critical step in experimental techni-
ques. The used sensors are requi red to be calibrated so that
their electrical signals relate to actual values of physical quan-
tities. In the case of long-ti me tests v erification of calibrati on
after the test i s also performed. The result of calibrati on i s
a calibration protocol and control system conf iguration using
the acquired calibration constants. Location of sensors on the
test i s usually on hydrauli c actuator . Force transducer (load
cell) between the rod end and the test arti cle and the displace-
ment transducer (usually LVDT) is installed inside the actua-
tor. Fig. 1 shows a typical setup.
3 Sensors and principles of calibration
Range def ines the domai n of appli cability of a sensor. F or
instance, a force sensor wi th a 10kN range can only measure
a maximum force of 10kN. When the sensor is beyond range,
the sensor provides no reliable information on physical values.
Sensitivity is the minimum physical value that a sensor can
detect. Sensitivity is usually expressed as percent of the sensor
range.
The calibration of a sensor relates A/D signals (i.e., numbers
displayed on computer screen) and physi cal quanti ties (e.g.,
Development and Improvement of
Calibration Methods of Force and
Displacement Transducers
Vvoj a zdokonalovn metod kalibrace snma sly a snma
deformac
Ing. Jan Heman / VZL, Plc., Prague
Static and fatigue tests of aircraft structures require calibrated force and displacement transducers. Within the
research project, the task of the VZL Strength of structures division was to ensure an unambiguous capability to
measure forces and deformations as required for testing, certification companies and supervisory authorities. In the
context of ever increasing demands on the time of calibration process a comprehensive system has been developed to
calibrate force and displacement sensors under mandatory requirements of metrology. This system is based on the
resources used in the laboratory of the VZL Strength of structures division and newly developed products for
efficient calibration of sensors.
Statick a navov zkouky leteckch konstrukc vyaduj kalibrovan snmae sil a deformac. V rmci
vzkumnho zmru bylo kolem zkuebny na divizi Pevnost konstrukc VZL na poadovan a nezpochybniteln
rovni zajistit schopnost kalibrace men sil a deformac pro zkouky, dohldac ady a certifikujc spolenosti.
V souvislosti pedevm se stle vymi asovmi nroky na proveden kalibrac byla vyvinuta ucelen soustava pro
kalibrovn snma sil a deformac podle poadavk obligatorn metrologie. Tento systm vychz z prostedk
pouvanch v laboratoch tvaru pevnosti konstrukc a nov vyvinutch ppravk pro efektivn kalibrovn
snma.
Keywords: Static, fatigue, test, aircraft, calibration, force, displacement, deformation, transducer,
metrology, efficient calibration.
Fig. 1
forces). With modern data acqui sition systems and electroni c
sensors, only mere numbers are obtai ned on the computer
screen. These numbers become meani ngful only after the y
have been related to physical quantities by means of a calibra-
tion. Calibrations are al ways carried out before starti ng a test.
They do not need to be repeated for each load case. It is howe-
ver a good experimental practice to check that the sensors are
properly calibrated before starting any experiment.
Typical force calibrati on setup i s shown in Fi g. 2. Kno wn
values of load standard are compared wi th force transducer
output. In practi ce, it is very difficult to obtai n a perfect cor-
respondance between physical value and A/D signal. Therefo-
re, the A/D data should be corrected by usi ng the calibrati on
results. For example, if the A/D signal value obtained from the
experiment i s y, the physi cal force x can be calculated as
x = y/c. Where c is determined from the calibration results.
Fig. 3 shows typical setup of actuators LVDT calibration. Dis-
placement standard represents acalibrated steel ruler. Everything
else is analogous to previous force calibration techniques.
4 The improved method of force transducer
calibration
The improved method of force transducer calibrati on is based
on e xperience and recent kno wledge from thi s f ield i n con-
junction wi th modern testi ng systems, especi ally the MTS
company.
The most accurate method of calibrati on is a physical cali-
bration. The pri nciple li es i n the physi cal calibrati on of the
force transducer inside the control loop of specific test control
system (including specific cables and plugs). This setup inclu-
des any negative effects that ma y occur duri ng the test. The
calibration standard included in this control loop gi ves unam-
biguously value of real force. Settings the calibration parame-
ters of the control system will achi eve correspondence betwe-
en the data displayed on the screen and real force.
On the other side, individual physical calibration of all force
transducers on the specific test and test control system is more
time-consuming and costly than other methods of calibrati on.
The Strength of structures division PK 3200 usually uses met-
hod of sensitivity calibration, where data of transducer sensiti-
vity comes from an accredi ted calibrati on laboratory of the
VZL. Improved method of force transducers calibrati on lies
in v erifying the sensi tivity cali-
bration by means of performi ng
physical calibration of one rando-
mly selected channel. Mutual
comparison of selected physi cal
calibration result of the transdu-
cer with its sensitivity calibration
result provides the real joi nt bet-
ween the sensi tivity calibrati on
provided by a separate calibration
laboratory and a specific test con-
trol system and the real forces.
This method also v erifies the
functionality of the test control
system before each test and pro-
vides a high safety and correct
loading of the test specimen of all
control channels. Of course there
is still need to verify all used con-
ditioners i n the control system.
This method bri ngs a reli able
physical calibration of all sensors
with a significant savings of ti me
and costs.
Within the scope of this method
we prepared methodological pro-
cedures for calibration of force transducers on the selected test
control systems used at the Strength of structures division. For
physical calibration two developed mobile calibrati on stands
are used allowing calibration of force transducers up to 50kN
and 660kN (Fig. 4).
Fig. 2
Fig. 3
Fig. 4
5 Improved method of displacement transducer
calibration
Displacement transducers data are usually used for moni toring
(measuring of deformations, safety system deformation limits
initiates unloading and system shutdo wn), some tests ha ve dis-
placement transducer as feedback i n the control loop. In an y
case, it is necessary to calibrate the transducer before the test.
Previous calibration method of actuators displacement trans-
ducers (LVDT) was calibrating using calibrated steel ruler (Fig.
3). Reachable resoluti on of measurements w as approximately
0.5 mm. The disadvantage lay in the need for two and more skil-
led workers and dif ficult stoking and setti ng of the steel ruler .
Next, the worker has to recalculate each measured value to real
displacement from zero point.
In order to increase efficiency of calibration electronic digital
calibration equipment was developed. For calibration procedure
one skilled worker is enough. This equipment is based on the
wire sensor wi th incremental encoder and reaches range 0 to
1500mm, linearity 0.05% FSO and resoluti on up to 0.025 mm.
With the wi re principle (Fig. 5), a linear motion is transformed
into a change in pulses by rotati on. A measuring wire made of
highly flexible stainless steel wires is wound onto a drum with
the aid of a long life spring motor. The winding drum is coupled
axially with an incremental encoder.
The sensor i s mounted on actuator body (usi ng de veloped
universal holder for most actuators, machined by waterjet tech-
nology) and the end of wire is usually coupled to the piston rod
using steel holder wi th bolt (Fi g. 6). Wire direction has to be
parallel to the pi ston rod. The equipment allows you to choose
a reference point zero. The measured v alue from zero
(negative or posi tive) is displayed on the di gital display device
(Fig. 7). Display is also possible via the serial interface RS232 in
combination wi th the de veloped appli cation KZP-PC for MS
Windows operating system (Fig. 8). Fig. 9 shows the real setup
of calibration equipment on the vertical MTS 244.21 series actu-
ator on the stand. Calibrating is possible with mounted load cell.
This equipment is currently in the verification operation.
Fig. 9
Fig. 5 Fig. 6
Fig. 7
Fig. 8
6 Conclusion
This article presents new and improved calibration methods of
force and displacement transducers used for static and fatigue
aircraft structures test. These methods are based on the resour-
ces used in the laboratory of the VZL Strength of structures
division, newly developed products for the efficient calibration,
experiences and recent knowledges from this field in conjunc-
tion with modern multi channel testing systems.
1 Introduction
Within the European project AVERT (Aerodynamic Validati-
on of Emi ssion Reduci ng Technologies) VZLU performed
tests to help to better understand b uffet beha viour on 2D
profiles.
Subject of this experiment was to verify the functionality of
the fluidic device on the traili ng edge on the model and the
influence of blowing on change of shock wave movement. The
blowing slot was located at the pressure si de near the traili ng
edge. The direction of blowing was perpendicular to the chord
of the profile. As a pressurised air the dry nitrogen was chosen.
The test at VZLU was performed under the transoni c regi-
mes around M=0.736 with several and several constant pres-
sures in the cavity of fluidic TED.
2 Nomenclature
. . . angle of incidence of the model in the wind tunnel
c
p
. . . pressure coefficient
c
L
. . . lift coefficient
c
M
. . moment coefficient
c
D
. . drag coefficient
G . . . mass flow
M . . Mach number
p . . . pressure
p
o
. . . total pressure
p
s
. . . static pressure
q . . . dynamic pressure
T . . . temperature
Indexes:
ref . . reference value
tun . . tunnel value
corr . corrected value
n . . . normal value
w . . . wake rake
0 . . . total
s . . . static
slot . value inside the slot
cavity value inside the cavity
Abbreviations:
TE - trailind edge
TED - trailing edge (fluidic) device
ONERA - The French Aerospace Lab
LEA - Aerodynamics Lab, University of Poitiers
3 Wind tunnel
The atmospheric intermittent suck do wn wind tunnel mark ed
as A4 at VZLU was used for these tests. The test section of
this WT was 0.4 x 0.8 m, which is normally used for 2D pro-
file measurement. The velocity in this wind tunnel is control-
led by setting the throttle behi nd test section and it can be set
continuously from M = 0.15 up to M = 1.0.
The sucked air is modified in the drying plant at the inlet of
the wind tunnel. The air is firstly cooled do wn to reduce the
humidity and then again warmed up to the desi red temperatu-
re. The temperature in the wind tunnel is 20 - 30 C and rela-
tive humidity is less than 20%. The drying prevents the two-
References
[1] MTS System Corporation: Model 493.10/793.00 Controller
Service, MTS, 2003
[2] MICRO-EPSILON MESSTECHNIK: Betriebsanleitung
wireSENSOR P60/96/115, Micro-Epsilon, 2009
[3] MICRO-EPSILON MESSTECHNIK: Betriebsanleitung
DD214NE, Micro-Epsilon, 2009
[4] Heman, J.: Funkn vzorek kalibran zazen
polohomr; VZL R-4599, 2009
Measurement of Buffet Condition
on 2D Profile Within AVERT Project at VZLU
Men podmnek buffetu na 2D profilu ve VZLU v rmci projektu
AVERT
Ing. Tom Jelnek, Ing. Martin Nmec / VZL, Plc., Prague
The article contains the description of unsteady pressure measurement on 2D profile, the preparation and
the analysis of the measurement. The experiment was performed by acquisition and analysis of pressure
transducers data. Transducers were located on the surface of the model. The measurement was performed
within a several phases, where the original model was measured at first without any other equipment and
then with the fluidic trailing edge equipment. The effect of blowing on change of shock wave movement
was studied and analyzed. The experiment was performed within the AVERT project under 6
th
FP.
lnek pedstavuje popis nestacionrnho tlakovho men na 2D profilu, jeho ppravy a vyhodnocen. Experiment
spoval ve sbru dat ze snma umstnch na po vrchu profilu a vyhodnocen pohybu rzov vlny. Men probha-
lo v nkolika etapch, kdy byl jednak men standardn model bez dalch pdavnch zazen, nsledn pak model
vybaven zazenm vyfukovn na odtokov hran. Byl studovn a vyhodnocen vliv tohoto vyfukovn na zmnu
pohybu rzov vlny. Experiment byl provdn v rmci mezinrodnho projektu AVERT een v rmci 6. RP.
Keywords: unsteady measurement, buffet, fluidic control, AVERT.
phase flow especially at transonic regimes.
There are slotted upper and lo wer walls in the test secti on.
The gap for this measurement was set to 10%.
The test secti on has two positions which can be used to f it
the model (see Fig. 1).
The first window which is normally used for this configura-
tion of measurement (test secti on A4 and measurement of 2D
airfoils) could not be used because of planed PI V application.
The first window cant be used due to technical reasons - opti-
cal access of laser sheet and cameras from si dewall.
To provide the drag measurement the wi nd tunnel is equip-
ped with a wake rake probe with 100 pressure taps. The inlets
of pressure taps are located 1.5 of chord (i t means 300mm)
behind the TE of the model.
Wind tunnel data:
G Test section dimensions: 0.4x0.8m
G Re ~3.106
G M = 0.15 - 1.0
G Tu = 0.2%
Wind tunnel calibration:
Calibration of used test secti on was performed for purpose of
AVERT measurement. Calibrati on w as performed
for M= 0.37 to 1.12 for se veral sets of slots wi dth on upper
and lower walls by axis probe of 15mm di ameter with 6 pres-
sure taps. The characteristic of boundary layer was also deter-
mined within the calibration.
4 Model
OAT 15A model was chosen and w as k indly borro wed by
ONERA for this test.
The model has chord length of 200mm and span of 390mm.
The span was 10mm less than the wi dth of test secti on. The
difference was filed with special manufactured piece of 10mm.
The model was equipped with:
84 pressure taps (67 at line A under angle 10 with 18 on
removable TE and 15 at line B)
19 Kulites transducers (16 on upper si de, 3 on lower side
and 4 on TE)
The model was equipped with transition tripping lines at 8%
of chord on both sides to ensure fully turbulent flow.
The new trailing edge with fluidic device:
The new trailing edge (desi gned and manuf actured by LEA)
equipped with a special cavity for pressurized air and a blowing
slot on pressure si de was used in place of the ori ginal trailing
edge of the model.
The new TED was equipped with:
Five pressure taps (at 160mm and 197mm on both si des
of the model and at 200mm ri ght on TE)
0 Kulites
One pressure tap i nside the cavity to control the pressure
in the cavity.
Two inputs to fill the cavity with the pressurized air from
both sides.
Because of expansion of the pressurized air from the cavity
to outsi de there w as a strong requi rement to use v ery lo w
humidity of this air. According to this requirement the tank of
50 li tters of Ni trogen at 200Bars w as chosen to supply the
cavity.
5 Measurement equipment
Measurement equi pment was based on po werful PC. It w as
equipped with RS232 and GPIB interfaces for communication
with tunnel controller and other peri pheries. Data acqui sition
Fig. 1 Drawing of wind tunnel test section
Fig. 3 ONERA OAT15A model
Fig. 2 Ratio of pressure and velocities according to distance
from the wall (mm)
Fig. 4 Trailing edge with fluidic equipment
was ensured by four si multaneously sampled A/D cards Nati-
onal Instruments - PCI-6123 (8 channels each) - sampling fre-
quency up to 500 kHz. Signal conditioning was done by Dewe-
tron system - each channel was configured separately by pro-
grammable module D AQP-BRIDGE-A, bandwi dth of the
module was 20 kHz.
Stationary measurement was done by two five-channel Sca-
nivalve multi plexers. It could acqui re 120 pressures (model
and wake rake probe) during one measurement run.
The calibration of angle of incidence
The angle of i ncidence was calibrated by the le vel and i t was
calibrated again when the model w as removed from the wi nd
tunnel to be sure that the zero angle i s still the same. The align-
ment with the upper and lower wall with the profile was chec-
ked as well. The angle of i ncidence for the measurement w as
set by hand along the scales on both wi ndows.
The measurement set-up
During each measurement when the Mach number in the wind
tunnel was stabilised (within 0.5 sec.) the data acquisition starts
to collect:
Pressure di stribution from pressure taps on the prof ile
(line A and B),
Static and total pressure from wake rake probe,
Static pressure from upper and lower wall of the wind tun-
nel and
Signals from Kulite transducers (collected at 20kHz).
6 Results and Analysis
With help of 232 runs of the wind tunnel there is a huge file of
data which could help to understand the b uffet phenomenon
and the influence of blowing on trailing edge.
A lot of points in the polar were measured within this expe-
riment. The study w as provided on one v alue of M and the
and the pressure i n the ca vity were v ariables. There were
observed the i nfluence of blo wing on pressure si de near the
TE. From charts on Fig. 7 we can assume these general points:
G Increase of pressure in the cavity generally caused incre-
ase of lift coef ficient - the pressure di stribution is getting
wider. Also the shock w ave on the sucti on si de mo ves
more downstream.
G On constant we can observe that the buffet starts earli-
er with increase of cavity pressure and shock wave moti-
on is more strong according the analysis of RMS value.
G If we look at the data i n terms of constant cL the increa-
se of cavity pressure caused the retreat of buffet.
G These general points are in conformity with the results of
ONERA mechanical mini-flap tests.
The RMS value:
Kulites data processing contains two procedures. The first was
averaging to obtain steady pressure distribution. The second
was focused on appreciation of buffeting. For this, RMS valu-
es of cp are computed for each K ulite (with note that no wi n-
dowing was applied):
Profile distribution of cp RMS w as provided for each measu-
rement. This distribution was affected by strength of buffeting.
Therefore some parameters of the curv e were chosen to app-
raise conditions on the prof ile. The situation is showed on the
Fig. 8.
Fig. 5 The measurement scheme
Fig. 6 Example of Kulites data
Fig. 8 RMS value
Fig. 7 Influence of pressure increase on the polar

( )
n
c c
c
n
pmean pi
pRMS
Clear i nformation i s max imum value of cp RMS and the

position of this value on the profile. Next information about the
process is distance on profile, where pressure fluctuati ons are
significant. This is well described by range where the cp RMS
is higher than selected value.
The results show that the influence of blowing is very simi-
lar to the i nfluence of steady mechanical mini-flaps what was
previously designed and performed by ONERA. Thus there is
a good presumption that the controlled pulsed jet on traili ng
edge should more retread the buffet or even remove them.
The map of RMS range v alue (see Fig. 9) shows that there
is some threshold whi ch di vides this area for b uffet and non
buffet region. The Fig. 10 shows the change of shock frequen-
cy on the investigated plane of polar.
7 Visualisation
For better understanding of the condition on the model the sur-
face oil v isualization was performed wi th oil and the whi te
powder of ma gnesium. The v isualisation w as performed
for M= 0.730 and = 2, 2.5, 3. On these pi ctures we can
see that the flow remains 2D up to = 3. The lateral extensi-
on has no si gnificant influence and that transition tripping line
to turbulent flow is sufficient.
Acknowledgement
Authors would lik e to thank to Mr Norman Wood and the
European Community for the sci entific and f inancial support
under the contract AVERT, contract No .: R TD
REG/H.3(2006)A/14212.
Another thank i s dedi cated to Mr . Eri c Coustols from
ONERA who ensured the model for the measurement wi thin
this project and to Mr . Jean-Bernard Dor from ONERA for
useful consultancy.
Fig. 9 The map of RMS range value
Fig. 10 The map of shock frequency
Fig. 11 Oil visualization for M=0.730 and = 2
Fig. 12 Oil visualization for M=0.730 and = 2.5
Fig. 13 Oil visualization for M=0.730 and = 3
References
[1] Pankhurst, R.C., Holder, D.W.: Wind Tunnel Technique; London,
1952
[2] Randall, R.B., Tech, B.: Frequency Analysis, 1987
[3] Nmec, M., Benetka J.: Calibration of A4 Wind Tunnel Test
Section; VZLU Report, 2008
1 Introduction
Recent developments of e xperimental techniques have intro-
duced operati onal modal analysi s (OMA) methods i nto the
standard use. Their main advance consists in the possibility to
identify di rectly modal parameters of a structure under real
operational boundary conditions and excitation forces. Finally
it enables us to achieve a higher complexity stage of numerical
models validation.
The classical experimental modal analysi s (EMA) compri-
ses a set of methods for modal parameters i dentification of
a real physi cal structure. Their standard scheme i s: known
input - unknown (second order vibratory) system - known
output. Li nearity and determi nism of a tested structure are
general and common assumpti ons. As a result of EMA we
expect a set of modal parameters, whi ch forms the base for
dynamic behavior description of the structure i n a frequency
range of i nterest. The term kno wn input means: 1) Uni que
and completely measurable set of e xcitation forces. Practi cal
realization of these forces represents se veral physical exciters
with a possibility to introduce correctly dynamic forces into the
structure and measure them onli ne. 2) No other forces act
during the experiment. It is evident that as logical consequence
of conditions 1) and 2) results laboratory i solation of a tes-
ted structure in course of the identification experiment.
Unknown system i nformation inherent in vibration move-
ment of a structure under laboratory test or real operati onal
conditions is derived from a set of measured signals of discre-
te points distributed on the structure. Some mi nimum spatial
density of measuri ng points is necessary to achi eve sufficient
mode shapes resolution in the frequency range of interest.
As a result of e xperimental i dentification process we get
a compact set of modal parameters (mode frequenc y, mode
shape, mode damping factor, modal mass). This set represents
dynamic properties of a structure in a frequency range of inter-
est. The k ey to the max imally conf ident descri ption of the
structures dynamics is the best possible quality of its identified
modal parameters. Modal parameters then come as the i nput to
the numerical model updati ng process. Verified, updated and
validated numeri cal model of a structure i s f inally used for
computational simulations of dynamic loads variants including
those which are experimentally not feasible. Especially, a valid
dynamic model of an ai rcraft structure serv es as the i nput to
the flutter analysis procedure.
The main goal of the EMA process (aircraft structures rela-
ted special nomenclature: Ground Vibration Test - GVT) i s to
provide complete and conf ident information into the dynami c
and flutter analysis procedures during the critical aircraft deve-
lopment phase. Questi on ne vertheless remai ns: What e xtent
Modal Parameters Identification
of a Structure Under Operational Conditions
Identifikace modlnch parametr konstrukce v provoznch
podmnkch
Ing. Petr Malnek / VZL, Plc., Prague
The classical experimental modal analysis EMA comprises a set of methods for modal parameters identification of
a real physical structure. Their standard scheme is known input - unknown system - known output . Linearity and deter-
minism of the tested structure are general and common assumptions. As a result of EMA we expect a set of modal
parameters, which forms the base for a dynamic behavior description of the structure in a frequency range of interest.
The term a known input means: 1) Unique and completely measurable set of excitation forces. Practical realization
of these forces represents several physical exciters with a possibility to introduce correctly dynamic forces to the
structure and measure them online. 2) No other forces act during the experiment. It is evident that a logical consequ-
ence of conditions 1) - 2) is laboratory isolation of a tested structure in course of the identification experiment.
The operational modal analysis, OMA techniques permit to use unknown operational forces as an only excitation and
to identify directly modal parameters of the tested structure under real operational boundary conditions. This article
presents some experience and results of introducing OMA at VZLU Strength of structures departments modal test
laboratory.
Klasick experimentln modln analza EMA zahrnuje soubor metodik identifikace modlnch parametr kmita-
vch dynamickch systm podle schmatu: znm vstup - neznm systm - znm vstup . Uvaovny jsou pedpokla-
dy linearity a determinismu systmu. Vsledkem identifikace je soubor modlnch parametr charakterizujcch kmi-
tav systm ve zvolenm frekvennm rozsahu. EMA vyaduje jednoznan definovanou a zcela mitelnou soustavu
psobcch sil. Praktick realizace takov silov soustavy pedstavuje: 1) Fyzickou realizaci silovch budi s monos-
t dostaten pesnho men psobc sly a monost zaveden tto sly zvolenm smrem do pedem vybranho
bodu konstrukce. 2) Striktn eliminaci vech ostatnch neznmch a nemitelnch silovch ink mezi okolm
a identifikovanou konstrukc. Tento poadavek logicky vede k nutnosti laboratorn izolace systmu v prbhu pro-
cesu identifikace.
Provozn modln analza OMA dovoluje vyut k buzen identifikovan konstrukce neznmch provoznch budcch
ink a umouje zjiovat modln parametry konstrukce s provoznmi okrajovmi podmnkami. lnek prezentu-
je dosavadn zkuenosti se zavdnm metodiky provozn modln analzy OMA na pracoviti Pevnost konstrukc
VZL.
Keywords: modal parameters, operational modal analysis (OMA), real operational excitation,
operational boundary conditions.
discrepancy e xists between modal parameters i dentified i n
laboratory conditions and those really e xisting within in flight
conditions? The theoretical linearity assumption guarantees the
independence of structural parameters on the load pattern.
Indeed for real ai rcraft structures the li nearity preserves only
inside a small region around the structures working center -
the relati ve zero posi tion of the structure for an y v ibration
movement under gi ven boundary and e xcitation condi tions.
Working centers of an ai rcraft structure staying on the ground
contrary to the structure i n flight conditions differ significant-
ly when taking into account a typically high modal density of
aircraft structures.
Classical EMA (GVT) i n laboratory condi tions produces
detailed information concerning all vibration modes including
their nonli nearity characteri zation. But li mited di splacements
of testing vibrations and zero air stream velocity predetermine
that, for e xample, identified damping factors include material
and frictional components only. What distribution of damping
exists duri ng the fli ght? The theoreti cal and computati onal
aeroelastic analyses give the answer and at present an e xperi-
mentally OMA deri ved in fli ght modal parameters set of an
aircraft structure helps as a higher complexity stage of nume-
rical models validation instrument.
2 Aspects and limitations of Operational modal
analysis methodology
At VZLU Strength of structures department - structural testing
group the system LMS TestLab was installed with a software
package Operational Modal Analysis PolyMax. The system
enables to get some appli cation experience of the output only
modal analysis.
Operational modal analysis techniques can use natural ope-
rational excitation of a structure as the only source of structu-
res vibratory movement, and then derive the system informati-
on from measured v ibration si gnals. An assumpti on there
exists: Operati onal e xciting forces are broadband stati onary
random forces. System identification scheme seems now: unk-
nown i nput (br oadband stati onary r andom) - unkno wn
(second order vibratory) system - known output. Realizati-
on of a controlled and measured artif icial excitation and labo-
ratory isolation are unnecessary for the method. Indeed opera-
tional forces should feed suf ficient excitation power over the
frequency band of interest into the structure. Thus the method
has a potential to i dentify modal parameters of a structure in
real operational and boundary conditions.
An anticipated input data format for OMA method i s a cor-
relation function of ti me domain record. Auto- and cross-cor-
relation functi ons are calculated di rectly from sampled ra w
time data or the y are deri ved from measured auto- and cross-
power spectra usi ng inverse FFT. Over recent years, se veral
modal parameter esti mation techniques have been de veloped
for a modal parameter extraction from output-only data. Let us
mention among others Natural Exci tation Technique, Stochas-
tic Subspace Methods and Frequenc y Domain Decomposition
Technique [6], [7], [8]. The Mathemati cal background i n
review, see [9].
As real operati onal excitation forces are not kno wn, OMA
method does not i dentify modal mass parameters di rectly.
Only indirect means exist [5]. Other limitations of the method
originate from limited possibilities to realize in operation vibra-
tion measurements on the large aircraft structure. Observabili-
ty of mode shapes is reduced due to a limited number of mea-
suring poi nts. Repeatabili ty of measurement records seems
Fig. 1 Nolwing Model
Fig. 2 Nolwing Model Geometry
Fig. 3 MAC correlation between OMA and PRM on Nolwing Model
poor as operational conditions show high order volatility com-
pared to the laboratory conditions. Frequency resolution beco-
mes limited, when a record length i s too short due to li mited
time i nterval of constant condi tions. Other dif ficulties ari se,
when the structure under test is excited by deterministic perio-
dic forces, as they are generated by engines, propellers, rotors,
gears and transmi ssions. OMA methods i dentify these frequ-
ency components as f alse modes. Experi menter's evaluation
should come into account to select structural only modes.
Practical working using OMA method compri ses stages as
a tested object descri ption, its geometry def inition, a possible
configurations list, boundary conditions characterization, mea-
surements methods, data stora ge and processi ng methods
description, mai n seri es of v ibration measurements and
records, measurement and record of all relevant parameters of
external conditions. Then stages continue as an e valuation of
data usi ng OMA methods, results presentati on, correlati on
with numeri cal soluti on results and classi cal EMA results,
interpretations of discrepancies and final evaluation.
As an e xample let us suppose an ai rcraft structure i n well
defined flight conditions, stable crui se constant ai r speed and
height of flight, when none of the relevant parameters of exter-
nal condi tions changes si gnificantly. The elementary OMA
method experiment is denoted R
i
, with experiment order i. At
the be gining the conti nuous record of e xternal parameters
starts into the database f ile PR
it
(M), where M is a number of
parameters. Then the record of the structures vibration respon-
se to a natural operational excitation (turbulence, gusts, acous-
tic noi se, propulsion systems generated v ibrations,) starts.
Let the record is stored as database file XR
i
(NxK) of time seri-
es of samples. N is a number of measured channels of vibrati-
on responses, K is a total number of samples, the constant
sampling frequency f
s
is supposed. Finally when vibration res-
ponses measurements are f inished, the e xternal parameters
record is stopped. Thereby the data record to the elementary
experiment R
i
is f inished. R
i
could conti nue i mmediately by
data analysis and evaluation using OMA algorithms. The result
of R
i
represents a set of modal parameters MOR
i
. On the basis
of R
1
- R
i
results evaluation a new elementary experiment R
i
+1
is defined and executed. Or R
i
data are simply stored for offli-
ne processi ng and e valuation and the ne xt e xperiment R
i
+1
continues according to a plan of experimental works.
3 Comparison of OMA and EMA on the dynamic
model NollWing under the laboratory conditions
To get the basi c experience with the OMA, the f irst compara-
tive experiments at the laboratory condi tions have been done
on the NollWing, the wing dynamic model of a commuter. Pre-
vious experimental works on the NollWing model using classi-
cal EMA methods phase resonance (PRM) and phase separa-
tion (PSM) serve as a benchmark [1], [2], [3], [4]. One half of
the wing model was as a cantilever fastened to a rigid frame-
work structure, see Figs. 1, 2. The wing model structure has 8
modes in the frequenc y range 0 - 25 [Hz], see. Tab 1, whi ch
contains frequencies and modal damping factors.
Preliminary tests i ndicated that the OMA method wi th an
electromagnetic exciter and broadband e xcitation force gi ves
nearly identical results as classical PSM.
Fig. 4a Strip mode shape representation, Mode 1, 1
st
vertical
bending, OMA-PRM results overlapping
Fig. 4c Strip mode shape representation, Mode7, 1
st
torsion,
OMA-PRM results overlapping
Fig. 4b Mode 1, 1
st
vertical bending, OMA-PRM results overlapping Fig. 4d Mode 7, 1
st
torsion, OMA-PRM results overlapping
Therefore another mean of e xcitation was used. Soft ti p
impact hammer i mpulses randomly di stributed i n ti me and
space around the NollW ing structure duri ng the elementary
experiment record. Thus the structure has really been forced
by an unkno wn e xcitation, moreo ver si gnificantly dif fering
from a broadband stationary signal assumption.
The resulting modal parameters set see in Tab. 1 OMATest-
Lab column. For comparison PRM and PSM wi th OMA see.
Tab.1. The comparable frequencies are not corrected ones. For
MAC correlation results of all eight modes coming from PRM
and OMA see Fi g.3. Mode shapes ha ve been compared i n
detail by the pro gram MoPar too. As an example see parallel
displays of PRM and OMA deri ved modes 1 and 7, see Fi g.4
a,b,c,d, OMA-gray, PRM-black lines.
4 Conclusions
The starti ng seri es of e xperimental w ork wi th operati onal
modal analysis OMA have been realized. The method is syste-
matically i ntroduced among standard modal testi ng i nstru-
ments portfolio at VZLU, Strength of structures department.
Introduction
Generating a good-quality grid is currently the single most impor-
tant user -controlled aspect of CFD, and i s the most dif ficult
aspect of CFD modelling. For very complex geometry, it can take
days to weeks (or even months) to create a quality grid.
An average CFD model for DNS simulation, such as airplane,
would contai n billi ons of gri d cells and i t would tak e man y
months if not years to solv e a si ngle model. That is the reason
why RANS i s still i n use for CFD calculati on i n engi neering
application.
The tetrahedral gri d has some adv antages and on the other
hand some disadvantages in comparison with the hexahedral grid.
Each of them needs specif ic procedure and the ti me consumpti-
on, rate of user-controlled time specifically, is different. Some of
the problems appear during creation of tetrahedral and hexahed-
ral grid are described and "hopefully" sorted out.
Tab. 1
References
[1] Hlavat, V., ern, O.: Modln zkouka modelu NOLWING kdla
malho dopravnho letadla systmem PRODERA; Zprva VZL R-
4215
[2] Malnek, P.: Vyeten modlnch charakteristik modelu NOLWING
kdla malho dopravnho letadla systmem TestLab; Zprva
VZL R-4221
[3] Hlavat, V., ern, O.: Zhodnocen metod pro zpracovn dat
modlnch a letovch flatrovch men malch dopravnch letadel;
Zprva VZL R-4216
[4] ern, O.: Teorie a praxe modlnch zkouek konstrukc letadel;
Zprva VZL R-2689/92
[5] Coppotelli, G.: On the estimate of FRFs from operational data;
Mechanical System and Signal Processing, 2008
[6] Brincker, R., Zhang, L., Andersen, P.: Modal identification from
ambient responses using Frequency Domain Decomposition; Proc.
of the IMAC18, San Antonio, Texas, 2000
[7] Coppotelli, G., Di Vincenzo, F. G., Mastroddi, F.: Aeroelastic
system identification of a flying UAV in operatative conditions;
IFASD, 2008
[8] Chiang, D. Y., Lin, Ch. S.: Identification of Modal Parameters from
Nonstationary Ambient Vibration Data Using Correlation
Technique; AIAA Journal, Vol. 46, No. 11, November 2008
[9] The LMS Theory and Background Book, 2005
The paper shortly presents laboratory benchmark of EMA and
OMA tests results and discuses special aspects of a modal test
under real operati onal condi tions. The OMA method seems
promising not only i n ai rcraft structures testi ng area . Other
application fields could encompass a laboratory vibration tes-
ting, advanced monitoring and power plant subsystems opera-
tional dynamic properties identification.
Acknowledgment
This work was supported by the Czech Ministry of Education,
Project Number MSM 0001066903.
Grid Generation in ICEM CFD Software
for CFD Calculation of Wing with High Lift
Devices
Tvorba st pro vpoty kdla dopravnho letounu
s vysokovztlakovou mechanizac
Ing. Armand Drbek, Ing. Petr Vrchota / VZL, Plc., Prague
This article deals with the grid creation for Euler and RANS computation of wing of airliner with high-lift devices in
ANSYS ICEM CFD program. The pros and cons usage of tetrahedral and hexahedral grid are brought out. The pro-
blems which can appear during this process are mentioned and the possible solutions are presented.
Tento lnek se zabv s tvorbou st pro vpoty kdla dopravnho letounu s vysokovztlakou mechanizac pomoc
Eulerovch a RANS rovnic v programu ANSYS ICEM CFD. V lnku jsou uvedeny vhody a nevhody pouit tetra-
hedrln a hexahedrln st. Dle jsou zde popsny problmy, kter se mohou objevit bhem procesu tvorby st
a naznaeny monosti jak je odstranit.
Keywords: ground tetrahedral grid, hexahedral grid, CFD.
Model of the wing
Geometry was created i n CATIA V5R16 program. It w as the
wing of the ai rliner intended for mi ddle distance. The hi gh-lift
system consi sted of double-slotted flap and deflected aileron.
The landing configuration is depicted in Figure 1.
Because the geometri cal model w as i ntended not only for
CFD calculation, therefore the real shape of the slotted flap w as
created. The flap was separated i nto two parts, i nner and outer
part. The separation was done in the kink of the wing. The other
gaps were between the end of the flap and the wi ng and betwe-
en the wing and the aileron. All gaps are sources of problems and
complicate the meshi ng process. Therefore some si mplification
of the geometry was done.
The flaps was connected and modelled as one body (two bodi-
es, front flap and rear flap, stri ctly speaking). The gap between
outer part of the flap and the wi ng was kept as well as the gaps
between wing and deflected aileron. The simplified geometry is
depicted in Figure 1.
Geometry import to ANSYS CFD
Before import geometry part i nto ICEM the connecti on of the
curves and surf aces should be check ed in CAD pro gram. Any
geometry reparations are more complicated in meshing program,
therefore is recommended do thi s in CAD program. CATIA V4
format seems to be better than IGS or an y other geometry for-
mat.
Geometry was imported with very high accuracy especially if
thin boundary la yer wi th v ery small f irst element i s needed.
ANSYS ICEM provides automatic correction of geometry when
Repair Geometry icon is used. Specific colour (red, green, yellow
or blue) illustrates connectivity to surfaces, which can be used to
determinate any gaps in geometry.
Tetrahedral grid generation
Tetrahedral grid generation is more automati c than he xahedral
grid and i t is prof itably used i n cases when geometry i s more
complicated. Grid generation is friendlier, no need to create com-
plicated blocks. This automatic generation brings some problems
also.
User is losing control above the generated grid. Frequent sour-
ce of errors are very thin trailing edges of wings, where bad ele-
ments or gaps are created. These errors can be eli minated when
is set appropriate value in minimal quality window at define thin
cuts icon at Global mesh setup menu.
Boundary la yer i s si mulated by creati ng pri smatic la yers.
ICEM CFD creates p yramids element as a connecti on between
prismatic layers and tetrahedral gri d. Generated pyramids could
have bad surface orientation as it was happened in this case main-
ly in regions where some parts of geometry are v ery close (for
example between main wing and flap). This error could be elimi-
nated by means of only one prismatic layer is generated which is
consequently divided into required number of layers.
When tetrahedral grid with prism was generated another pro-
blems could be appeared. Spacing between prisms and tetrahed-
ral grids was too high as you can see in Figure 2.
Recommended value of spaci ng between elements i s 1.1 or
1.2 for solver we used i n following part of thi s task. Inadequate
spacing could be repaired if grid along the profile is significantly
smoothed (enough small elements on the surface). High number
of cells was the result of this approach.
Hexahedral grid generation
Creation of hexahedral grid is more time consuming (of course it
is depending on the comple xity of geometry). Gri d is created as
multi block, structured. Some lines of blocks are associated to the
curves or surfaces of the wing. Geometrical complexity of some
regions on wing required special approach on curv e association.
Function Automatic linear in Edit edge menu can be used when
complicated geometry ha ve to be described more di rectly. This
function ensures that associ ated edge di rectly cop y the curv e.
Quality of edge i s influenced of the number of the nodes. The
same association and node di stribution on adjacent o -grid could
be done wi th the Li nk edges functi on in the Edi t Edge menu.
Disadvantage of this function is, when user wants to change loca-
tion of a vertex on Edge, where Automatic linear is used. Associ-
ation must be done a gain, because ICEM CFD doesn't re-com-
pute location of nodes.
A lot of errors hang to gether with boundary layer generating.
Firstly should be controlled quali ty of geometry if i t is sufficient
for generally first element of boundary layer. Elements with bad
volume orientation are usually in regions, where the errors in geo-
metry are (gaps, not connected curv es directly on surf aces etc.)
The second most common error w as too acute angles between
edges were i n regions, where flap w as deflected. Changi ng the
geometry of the blocks and better spacing on lines leads to repa-
ir this error.
The worst place where in connection of the outer part of deploy-
ed flaps and the wing. The reason why was too thin and too skew.
A lot of bad elements were placed i nside these blocks. The bet-
ter spacing and more cells i n these blocks the better cells, b ut
there is a limit of reasonable amount of cells.
The negative volume elements inside the long blocks can be fixed
by splitting of the blocks and modification of the spacing or addi-
tion of the cells.
Conclusion
It managed to create unstructured tetrahedral gri d around wi ng
with hi gh-lift de vices and structured he xahedral gri d, as well.
Both grids were suitable and their quality was sufficient to make
CFD simulation in EDGE program, which is more ambi tious of
grid quality than for example FLUENT program.
The main advantage of unstructured tetrahedral gri d against
structured hexahedral grid (lesser time rate of user control) w as
little bit depreciated by fixing of the mesh errors (ne gative volu-
me cells, surf ace orientation etc.). Creati on of structured he xa-
hedral grid is more time consuming and more ambitious of block
topology, but the final grid has usually better quality.
This task will be continued by CFD simulation in EDGE pro-
gram.
References
[1] Anderson J.D. Jr.: Computational Fluid Dynamics The basics with
applications; McGraw-Hill International Edition, 1995
[2] EDGE manual:
http://www.foi.se/upload/projects/edge/documentation-latest/edge-
quickstart.pdf
Fig. 1 Simplification of geometry (above)
Fig. 2 Inappropriate spacing between tetrahedral grid and
prism layers (right)
Fig. 3 Tetrahedral grid on deflected aileron. It was region where bad
elements were mainly generated (left side) Typical generated bad element.
Edges of the pyramide element is overcrossed here
Fig. 4 Example of
multi-block structure
with use automatic linear
function
Fig. 5 Various views of
hexahedral grid on wing
A S S O C I A T I O N O F T H E
AV I A T I O N MA N U F A C T U R E R S
Fig. 1 The model
of small commercial
airplane (the model
of icing of takeoff
configuration,
FL
= 38)
Fig. 4 The model of
airplane with ice
accretion model of
takeoff configuration,
FL
= 38
Colour illustrations to the ar ticle published on pages 25 to 28.

File 422

Încărcat de

Informații document

Descriere originală:

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

File 422

Încărcat de

Drepturi de autor:

Formate disponibile

LETECK

are radial and circle coordina-

w. The new system of equati ons i n meridian plane for

are given by formulas

spreadsheet. The matrices calculation is the ultimate

by usi ng the ei ght nodes

spreadsheet. Starting from the

by using the eight nodes laye-

Clear i nformation i s max imum value of cp RMS and the

S-ar putea să vă placă și