O c t o b e r
2 0 1 0
ISSN 1211—877X
CZECH
AEROSPACE
Proceedings
LETECKÝ
zpravodaj
In this issue:
Static Strength
Verification Procedure
of the Real Composite
Structures
J OU R N A L
F O R
C Z E C H
AE RO S PAC E
R E S E A R C H
Electromagnetic
Threats of Flight
Safety and their
Evaluations Using CEM
CFD Calculation of the
Wing with High-lift
Devices
Hardware in the Loop
Simulation of FBW
Components
Composite Material
Machining and
Manufacturing of
Molds Using the CNC
Machine Tool Unit
Fatigue Prediction
Based on Finite
Element Analysis of
Riveted Joints
Aerodynamic
Optimization
Procedure of Flapped
Sailplane Airfoil
No. 3 / 2010
CZECH
AEROSPACE
P r o c e e d i n g s
J OU R N A L
F O R
C Z E C H
AE RO S PAC E
R E S E A R C H
VÝZKUMNÝ A ZKUŠEBNÍ LETECKÝ ÚSTAV, a.s.
Editorial address:
VZLÚ / Aeronautical Research and Test Institute
Beranových 130, 199 05 Prague - Letňany
Czech Republic
Phone.: +420-225 115 223, Fax: +420-286 920 518
Editor-in-Chief:
Editor & Litho:
Ladislav Vymětal (e-mail: vymetal@vzlu.cz)
Stanislav Dudek (e-mail: dudek@vzlu.cz)
Editorial Board:
Chairman:
Vice-Chairman
Members:
Josef Kašpar, Vice-president ALV, General Director VZLÚ
Jan Bartoň, ALV
Tomáš Bělohradský, Vladimír Daněk, Luboš Janko,
Pavel Kučera, Petr Kudrna, Zdeněk Pátek, Antonín Píštěk
Publisher:
ALV / Association of the Aviation Manufacturers
Beranových 130, 199 05 Prague 9 - Letňany
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 ČR 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 published at http://www.vzlu.cz/.
Czech AEROSPACE Proceedings
Letecký zpravodaj
3/2010
© 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
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.
EUROPEAN UNION
EUROPEAN REGIONAL DEVELOPMENT FUND
INVESTMENT IN YOUR FUTURE
1
LETECKÝ
Z P R AV O D A J
2/2010
Contents / Obsah
2
Probability of Detection and Prediction of Fatigue Crack Growth in Aircraft Structures
Pravděpodobnost zjištění a predikce šíření únavových trhlin leteckých konstrukcí
Ing. Marie Boháčová, Ing. Jakub Šedek, Ing. Jiří Pavlas / VZLÚ, Plc., Prague
6
Electromagnetic Threats of Flight Safety and Their Evaluations Using CEM
Ohrožení letu elektromagnetickým prostředím a jeho vyhodnocení s pomocí CEM
Pavel Tobola, Lubomír Martinák, Zdeněk Řezníček / Evektor, spol. s r. o., Uherské Hradiště - Kunovice
13
CFD Calculation of the Wing with High-lift Devices
CFD výpočet křídla se vztlakovou mechanizací
Ing. Armand Drábek, Ing. Petr Vrchota / VZLÚ, Plc., Prague
16
Modern Design of Avionic Equipment for the L159 Aircraft
Moderní konstrukce leteckých palubních přístrojů pro letoun L159
doc. Ing. Rudolf Jalovecký, CSc. / Department of Aerospace Electrical Systems, Defence University,
Brno
20
Hardware in the Loop Simulation of FBW Components
HIL simulátor FBW komponent
Zdeněk Hanzálek, Pavel Hospodář, Martin Hromčík and Libor Waszniowski / Department of Control
Engineering, Czech Technical University in Prague, Faculty of Electrical Engineering
Jiří Doubrava / AERO Vodochody, Ltd.
27
Fatigue Prediction Based on Finite Element Analysis of Riveted Joints
Únavová predikce na základě MKP analýzy nýtovaného spoje
Ing. Jan Papuga, Ph.D. / Evektor, spol. s r. o., Uherské Hradiště - Kunovice
32
Static Strength Verification Procedure of the Real Composite Structures
Metodika ověření statické pevnosti reálných kompozitových konstrukcí
Ing. Vladimír Snop / VZLÚ, Plc., Prague
35
Composite Material Machining and Manufacturing of Molds Using the CNC Machine Tool Unit
Obrábění kompozitů a výroba forem na CNC obráběcím centru
Ing. Radek Mazal / HPH s.r.o., Kutná Hora
39
Aerodynamic Optimization Procedure of Flapped Sailplane Airfoil
Postup aerodynamické optimalizace profilu větroně s klapkou
RNDr. Jaroslav Hájek, PhD., Doc. Ing. Zdeněk Pátek, CSc., M.S., Mgr. András Szöllös, Ing. Marian Zabloudil
/ VZLÚ, Plc., Prague
C Z E C H A E R O S PA C E P R O C E E D I N G S
2
Probability of Detection and Prediction
of Fatigue Crack Growth in Aircraft
Structures
Pravděpodobnost zjištění a predikce šíření únavových
trhlin leteckých konstrukcí
Ing. Marie Boháčová, Ing. Jakub Šedek, Ing. Jiří Pavlas / VZLÚ, Plc., Prague
This paper describes possibilities of nondestructive inspection and computing procedures for assessings fatigue
damage of commuter aircraft’s principal structural elements. Nondestructive inspection provides crack detection and
limits for minimal detectable crack size for the method studied. Computing provides numerical calculation of fatigue
crack growth.
Článek popisuje možnosti nedestruktivní kontroly a výpočetní postupy pro odhad únavového poškození kritických
míst primární konstrukce malého dopravního letounu. Nedestruktivní kontrola zabezpečuje detekci trhlin a definuje
minimální zjistitelnou velikost trhliny pro danou metodu. Výpočetní část se zabývá numerickým výpočtem šíření
únavových trhlin.
Introduction
Aircraft are designed in accordance with the rules of Safelife philosophy. After the expiration of safe service life it is
possible to prolong the service life of an old aircraft only by
accepting the Damage Tolerance philosophy. This philosophy enables effective utilization of current state of the airframe. Determination of in-service inspection intervals and choice of proper NDI procedures is the task of the manufacturer.
In-service aircraft are checked in order to detect degradation which might lead to premature failure. The importance
of maintaining of structural integrity has to be emphasized,
mainly because of potential danger of fatigue failure and
corrosion. In turn it emphasizes the importance of inspection programs under which an aircraft may attain long, safe
service life. Such programs must be applied to all Principal
Structural Elements (PSE) and must focus on hidden areas
and areas exposed to high strain.
Computer simulation can provide the estimation of the
crack growth curves with sufficient accuracy, even in case of
complicated PSE structures which are crucial for preserving
airframe integrity and which participate in transfer of load.
One of the PSE is a landing gear beam on which the computer simulation procedure is demonstrated.
Probability of Detection and Nondestructive
Inspection of Aircraft by Eddy Current Testing
Factors Influencing Probability of Detection (PoD)
Probability of crack detection or any other damage detection is
a result of many factors. The nondestructive inspection (NDI)
method, though very important, is still only one of them. In
order to make nondestructive evaluation effective and to maintain high level of structural integrity it is necessary to consider
all factors possibly affecting the final PoD curve. The main
factors influencing the eddy current PoD curve are human factors and personnel qualification, access to inspection area,
inspection procedure, equipment parameters drift, measurement repeatability, detectable crack size, 90% reliability at
a 95% confidence factor, signal-to-noise ratio and reference
standards.
NDI capability of crack detection of a specified size is optimal when the procedure is performed by a qualified technician. The term qualification has a special meaning in NDI. It not
only includes knowledge, but practice and manual skill as
well.
The next factor is the access to the inspection area. In aircraft inspection praxis some of the principal structural elements
are difficult to access. Some are even inaccessible. Thus,
a special approach to NDI performance is required. It may
result in changing originally intended NDT method for another
one. When the thickness of inspected PSE is not within the
range of eddy current penetration, then either different NDT
method must be considered, or another solution must be
found. Such situation may be solved by additional openings
which enable the inspection of the given part of airframe.
These openings can provide access to the required area from
another position, where are for example only two layers of
aluminum sheets.
All NDI procedures should be developed and tested first in
laboratory and then verified by in-service aircraft checks.
These tests should provide limits of specific NDT method,
because they are not obvious and cannot be seen from drawings and sketches. The procedure is following: damage tolerance is performed for each PSE, and a marked-up drawing
and crack growth curves (Figures 1 and 12) are submitted to
NDT engineer. Depending on materials and thickness of PSE,
an experienced NDT engineer chooses appropriate NDI methods for PSE inspection. For eddy current testing, a simulated
structure containing artificial defects is fabricated or the specimens with cyclic fatigue cracks in the second layer of rivets
(Figure 4) are used. The artificial defects are usually EDM
(electrical discharge machining) notches of different size.
These simulated structures and specimens are used for elaboration of preliminary procedures and detectable flaw of different size for each PSE. The procedures are finalized and then
verified on aircraft in service [1].
3
LETECKÝ
Fig. 1 — Typical crack growth curve (adet is the minimal detectable
size of defect)
Eddy current equipment is an important factor in NDI. The
parameters of equipment are defined in each inspection procedure. However, inspectors do not need to use the same type of
equipment. It is possible to use equivalent types of equipment.
The problem is how to determine equivalency of two similar
instruments from different manufacturers or two identical
instruments from one manufacturer. This is especially true in
case of similar eddy current probes which do not detect the
same crack size at 90% PoD [1]. The results gained with similar or identical probes may be compared on the reference standards with EDM notches. Usually, the notch size should
demonstrate the detectable crack size.
Despite all efforts to ensure repeatability, eddy current (EC)
flaw-signal amplitudes are never exactly the same. Instead of
it, there is a distribution of signal amplitude values ranging
from a minimum to a maximum and represented by an average value. If ”n“ is the number giving how many times the
flaw-signal of specific amplitude was observed and ”N“ is the
total number of scans, then the n / N as function of signal
amplitude is the probability density function for the flaw-signal amplitudes. Similarly, the curve of noise signal distribution
can be defined. Let's assume that the probability density function of flaw-signal amplitudes lies to the right of the probability density function of the noise. The given signal amplitude can
be unambiguously interpreted either as a flaw signal or as
a noise signal. In this ideal case, most of cracks would be
detected. In practice, such a situation occurs only in case of
large cracks and weak noise signals. However, when looking
for small cracks, the flaw and noise signals overlap to some
extent (Figure 2). Then, the reliability of inspection is given by
the extent of the overlap. Modern instruments can separate the
noise signals (instrument noise, lift-off and no crack) from
flaw signals, as indicated in Figure 3. The main goal of NDI is
to determinate a value for adet and to have an inspection system
with small Type I error (inability to detect crack, when the
crack is present or is smaller than adet) and Type II error (indication of crack, when no crack is present). In situations, where
the probability density function is significantly overlapped by
the flaw-signal and the noise signal, both errors (Type I and II)
occur, as indicated in Figure 2. This is the reason why a threshold ”A“, respectively minimum detectable crack size adet
must be determined (Figure 2).
A limit for the detectable crack size adet represents a crack
length corresponding to high probability of detection. The
Z P R AV O D A J
3/2010
Fig. 2 — Interaction of Signal/Noise Discrimination and the PoD [1]
detectable crack length is different for each inspection method
and PSE. However a chosen method will determine adet and
establish initial point for determination of checks in service.
A primary NDI method and at least one alternative NDI method are developed for most of PSE. The primary NDI method
is the most sensitive one. It is the method, which may detect
minimum crack size and which may be useful for the largest
interval of crack growth ∆N. Successful detection can be achieved only over the threshold determined by a minimal crack size
adet (Figures 1 and 2). Although the errors of Type I and Type
II may occur on this threshold ”A“, the discrimination between flaw-signal and noise signal is sufficient. In addition the
decision criteria (crack vs. no crack) are here clearly defined.
Figure 12 illustrates the crack growth curve resulting from
a computer simulation, but it says nothing about the definition
of the initial point for in-service checks and does not define the
minimum detectable crack size by eddy current.
Fig. 3 — Y-t display of noise vs. cracks
C Z E C H A E R O S PA C E P R O C E E D I N G S
4
(1)
KI
σ
a
β
Stress intensity factor in Mode I [MPa √m]
Nominal stress [MPa]
Crack length [m]
Correction function [-]
Fig. 4 — Typical XY display of crack in the second layer of rivet
As mentioned above, the NDI is reliable in term of signal noise discrimination, if the flaw-signal is standard and if it is
well separated from noise-signal (Figure 3). In case of the
coincidence of noise and flaw-signals, there is no discrimination and the NDI is useless. Signal-to-noise ratio should be three
to one or better. It is obvious from Figure 3 that the crack signals of 2 mm and 4 mm cracks are entirely separated from
noise signal. The small signal on the left is not interpretable
(mixed signal of a crack smaller than adet + hole + rivet).
The probability of detection is defined as the probability that
a trained inspector using a given inspection procedure will
detect a flaw of a certain size if it exists. In practice the flawed
specimens are mixed with the flawless specimens. 100% confidence of eddy current inspection can be achieved only for large
flaws. For a small flaw size the confidence level is reduced.
Crack Growth Propagation Analysis
It is possible to use analytical tools for a crack growth distribution in some cases. The PSE can be substituted for a simple
part but the computation geometry need not correspond with
the original shape. The results from the analytical evaluation
show acceptable accuracy in initial crack growth. During the
crack propagation the stress distribution changes subsequently
and the analytical solution can be unsufficient. The analytical
tools do not allow an accurate description of areas with sudden
geometrical changes. The finite element method can give more
accurate result of the stress distribution around the crack front.
Input data for crack growth:
●
●
●
●
Part geometry
Material data
Load history
Crack geometry
The PSE geometry is converted to a computer model for purpose of a finite element mesh creation and subsequent stress
intensity factor computation around the crack front. These
values are used for the crack increment computation along the
crack front. The increments are smoothed into the curve — the
new crack shape. For next crack shape it is needed to create
new model and make the other steps for increment computation. The new crack increment can be computed in PREDIKCE
[3] and AFGROW [2] software. The inputs for these programs
are recomputed stress intensity factor values into beta correction values through the use of Formula (1), load sequence and
material data.
Fig. 5 — Crack prediction computation scheme
The input data for the computation is the crack shape and its
minimum detectable crack size adet from NDI in service conditions or the crack required by pertinent civil or military regulations.
For the verification of calculation the experimental data
from the landing gear beam test created at VZLU were used.
The beam was bolted to the duralumin dummies as shown in
Figure 6. This laboratory fatigue test simulated the average
running conditions. One load sequence corresponds to one
flight hour. One critical area on each side of the beam was
identified.
Fig. 6 — Landing beam assembly
Material crack growth data used in the computation were obtained from an experiment applied on a semi-product. From
these data it is possible to obtain Paris law constants and define a crack distribution in dependence on the stress intensity
factor amplitude:
(2)
da/dN Crack growth rate [m/cycle]
C,n Paris law constants
dK Stress intensity factor amplitude [MPa √m]
5
LETECKÝ
Z P R AV O D A J
3/2010
Fig. 7 — Critical area
There are two local stress peaks at the edges of the rivet hole.
The position of the hole is shown in Figure 7. It is likely that
there were two cracks at the edge of the corner distributing
from the hole until the outer crack propagated to the upper
flange side. After this moment the final crack was significantly growing into the flange centre direction.
The document JSSG-2006 [4] defines the initial state with
0,1 inch (0,254 mm) edge crack in diameter. This size is used
for the initial model. This crack is applied on the both sides of
the hole. There is a bottom view in Figure 8.
By the gradual calculation of the increments along the crack
front and its smoothing was obtained a new crack shape. All
computational procedure is shown in Figure 5.
The new crack shapes are shown in Figure 9. The computation shows some significant moments. The crack increments
are relatively small until the propagation reaches the flange
corner (39 000 flight hours). In this moment the crack growth
on the left side of the hole is rapidly increasing. The crack
growth rate on the right side is not significantly influenced by
the other side until the left crack propagates through whole
area. The computed crack growth curves are displayed on the
left section in Figures 10a and 10b.
The computation predicts the slower crack propagation than
the test result shows. The deviation is about 24% (considered
crack growth). It is probably caused by used material model
and inaccurately initiation identification ability of the experimental crack growth propagation. With considering of random
crack initiation the computed values are moved to the X-axis to
fit the experimental data as shown on the right section in Figs.
10a and 10b. The minimum detectable crack size is 2 mm. The
inspection interval is determined by the following Formula (3):
∆N I = ( Na1 − Na det ) / n
Fig. 8 — Initial crack 0,1 inch (dark area)
Fig. 9 — Computed crack shapes
Figs. 10a,b
— Crack growth
versus flight hours
(L1, L2 correspond
to Figure 9)
(3)
∆NI Inspection interval in flight hours
Na1 Flight hours to the critical crack length
Nadet Flight hours to the minimal detectable crack size
n
inspection factor n = 3
In accordance with equation (3) the inspection interval is 6600
flight hours for safety aircraft running.
Conclusion
NDI and computing analysis of PSE are inseparable from
Damage Tolerance philosophy for providing the required operating reliability. The main goal of airframe nondestructive
inspection by eddy current is determining the minimum detectable crack size which establishes the initial point for determination of in-service checks. The computing analysis defines the
crack growth curve of each PSE. It serves as the determination of inspection intervals for aircraft.
References
[1]
Hagemaier, D. (1998): Factors Influencing Eddy
Current PoD in the Field Environment; RTO meeting
proceedings 10, RTO-MP-10, ISBN 92-837-1002-9
[2]
Harter James A. (2007): AFGROW Users Guide and
Technical Manual, AFRL-VA-WP-Air Force
Research Laboratory-TR-2008-XXXX
[3]
Růžek R. (1998): Program PREDIKCE pro výpočet
šíření únavových poruch v leteckých konstrukcích
při reálném provozním zatížení, uživatelská příručka
verze 4.00; VZLÚ R-2939/98
[4]
United States of America, Department of Defense
(1998): Joint Service Specification Guide, Aircraft
Structures, JSSG-2006
C Z E C H A E R O S PA C E P R O C E E D I N G S
6
Electromagnetic Threats of Flight Safety and
Their Evaluations Using CEM
Ohrožení letu elektromagnetickým prostředím a jeho vyhodnocení
s pomocí CEM
Pavel Tobola, Lubomír Martinák, Zdeněk Řezníček / Evektor, spol. s r. o.,
Uherské Hradiště - Kunovice
Environmental electromagnetic threats of an aircraft and its vulnerable systems are examined at Evektor by using
methods of Computational Electromagnetics (CEM). As an airframe shall provide an inherent systems' protection
against the threats (indirect effects of lightning and HIRF), the intensity of internal electromagnetic field as well as
responses on cable bundles in this internal environment are to be evaluated to quantify them. The full-wave FDTD
(Finite Difference Time Domain) and MTLN (Multi-conductor Transmission Line Network) models with appropriate
solvers are used for this purpose by Evektor company.
Evektor, spol. s r.o. zkoumá ohrožení letounu a jeho zranitelných soustav okolním elektromagnetickým prostředím
metodami výpočtů elektromagnetického pole (CEM). Protože drak letounu musí poskytovat inherentní ochranu
těchto soustav proti elektromagnetickým hrozbám (nepřímé účinky blesku a HIRF), vyhodnocují se intenzity vnitřního elektromagnetického pole i odezvy kabelových svazků na toto pole. K tomu Evektor využívá celovlnného elektromagnetického modelu FDTD (metoda konečných diferencí v časové oblasti) a MTLN (vícevodičová metoda sítě
přenosových vedení) s návazností na příslušné solvery.
Keywords: Electromagnetic threats, computational electromagnetics, environmental EMC, FDTD
& MTLN solvers.
1 In general
The recent technical progress in civil aviation has brought
a successive replacement of classical flight instruments by
modern avionics. Output signals of necessary sensors are digitally processed and displayed using colour LCD screens giving
an overview about basic parameters of an aircraft and its flight.
Avionic systems become more complex and integrated and
they execute more functions that are critical to the flight safety.
Unfortunately, modern avionics is more vulnerable to electromagnetic interferences than the classical instrumentation.
The interferences are caused partly by unwanted radiations of
other systems installed on the aircraft and partly by environmental effects as lightnings and high intensity radiated fields
(HIRF). In the first case, we talk about an intersystem electromagnetic compatibility (EMC) and in the latter about the environmental electromagnetic compatibility (3E-compatibility) of
the airplane systems.
If an aircraft is struck by a lightning (with the probable frequency of one strike per year) or meets a strong RF transmitter or radar (several times during a flight), hazardous responses (voltages and currents) are forced into the internal wiring
of the aircraft. The cable bundles excited by this way become
unwanted sources of electrical energy that is distributed inside
the airframe. The responses (transients in case of lightning and
modulated RF signals in case of radiated fields) may reach
such energetic levels that input/output circuits of sensitive
instruments connected to the wiring may be destroyed or
a functional upset may occur on them. Consequences of these
incidents may be fatal if the injured avionic system carries out
critical functions with reference to the flight safety.
The above mentioned situation is more complicated by a frequent use of new composite structural materials that are
known not only by their excellent mechanical properties, low
weight and low manufacturing costs, but unfortunately also by
their low shielding effectiveness in a comparison to metallic
materials used formerly. If these composite materials cannot
provide a quite reliable inherent protection to aircraft systems,
they must be appended by special protection layers, especially
when used for an aircraft skin covering critical systems.
Other complication of the situation is a continuous increase
in radiated power on transmitters and radars that may be met
during flight and increasing number of requirements for operation of wideband portable electronic devices (WPEDs) to be
used by passengers inside airplanes during a flight.
An effective protection of an aircraft and its critical systems
against these electromagnetic environmental effects (3E) is
established by aviation regulations and must be demonstrated
when any new aircraft is certified for the business as well as
general aviation. The approach to this demonstration is specified by the standards specified in the Table 1 (in their European as well as U.S. versions) and certification tests are usually
done at several accredited EMC laboratories using the experimental way given by the standards.
Tab. 1 — Electromagnetic Environmental Effects' Protection
Certification Standards
An internal intensity of electromagnetic field as well as wiring
responses are measured when the aircraft is radiated by a high
intensity RF field or excited by conducting of current waves
(simulated lightning transients) through the airframe. This
7
experimental aircraft testing requires a special instrumentation,
skilled personnel and is usually executed inside an anechoic
room of extreme dimensions. If only several laboratories in the
world are accredited with this experimental EMC testing on
airplanes, their pricings are not too competitive and the costs
for these tests may become unacceptable for a majority of
small aircraft manufacturers.
There is a paradox in general aviation that the smaller is the
aircraft the smaller is the consideration paid to these environmental threats, but contrariwise, the greater is the hazard caused by them when considering current densities on a small aircraft connected as part of the lightning channel or airframe skin
thickness of a small aircraft exposed to a strong radar radiation. Taking into consideration a big aircraft (e.g. Airbus or Boeing), the problem seems to be resolved now, because the costs
for demonstration of compliance with regulations using tests
represent only an insignificant part of a budget of a new aircraft development project. In the case of a new small aircraft
certification, we cannot consider any cost reduction for this
demonstration especially when using the airframe of composite materials and installing modern (and more vulnerable and
critical) avionic equipment. The costs for compliance demonstration by experimental testing grow to per-centually significant part of the project budget and this may be a cause of the
project non-success.
Having considered the reasons mentioned above, Evektor
company is looking for inexpensive EMC tests and certification methods for small aircraft using the following concepts:
- replacement of experimental test methods with simulations
using means of computational electromagnetics (CEM).
- use of new structural composite materials with well specified or certified electromagnetic properties (prepregs) to create a predictable airframe internal environment for electromagnetically sensitive installations and systems,
- use of software for electromagnetic field visualization in
time as well as frequency domain and computing of induced response levels in wirings situated in this environment
using an aircraft digital mock-up data for materials and spatial arrangement,
- verification of the maximum error between the data obtained
by simulation and measurement,
- embodiment of an electromagnetic visualization into a software which is normally used for creating of aircraft digital
mock-up as well as cabling inside or outside the digital
model (problem of antennas' layout on a small aircraft
body).
2 Electromagnetic field and wiring response
simulations
The fundamental idea to examine the action of electromagnetic
environmental effects (3E) on an aircraft and its systems using
computational models is not quite new at our company. Finite
element method (FEM) has been used in Evektor especially for
structural analysis (mechanical strength) calculations since
1995. These activities were begun by our hire engagement in
some aircraft projects of Aero Vodochody (Ae 159 and Ae 270)
as well as other airframe producers (Raven of Wolfsberg Aircraft Co.) and continued through our involvement in the Czech
automotive industry (Skoda-Volkswagen) since 2000.
LETECKÝ
Z P R AV O D A J
3/2010
We can say that Evektor has more than 12-years experience
in creating FEM models for mechanical engineering and our
colleagues in the automotive crash simulation department were
the first pioneers to use the ANSA universal meshing preprocessor as the most powerful and effective tool for the FEM
models preparation. Having taken into consideration their best
experience with the FEM models preparation for the years and
previous involvements of the electric and avionic systems'
installation department in experimental EMC certifications on
various Czech aircraft development projects, a logic decision
has been made by the top management to extend engineering
activities of Evektor into the CEM field.
As the transfer of basic know-how in CEM from other aircraft manufacturers and research bodies was shown problematic or impossible for us, we have decided to create our own
know-how using a research project supported by the Czech
government. We created a consortium of three commercial
companies (Evektor, Hit and Mecas-ESI) and two universities
(VUT Brno and UTB Zlín) to apply for a grant with the project ARTEMIS (Analytic Research of Threats in ElectroMagnetically Integrated Systems) at the Ministry of Industry and
Trade of the Czech Republic. A financial support for this project was approved within the TANDEM Research and Development Program and made realistic our idea of aircraft structure evaluation using CEM at the beginning of 2007 [1].
To obtain the first experience with CEM tools, we selected
the commercially available PAM-CEM FDTD solver (for EM
field simulations) in suite with the CRIPTE MTLN tool (for
cable response calculations) delivered by ESI Group of France [2] and distributed by one of our partners in the project
ARTEMIS (Mecas-ESI s.r.o. Plzeň).
FDTD Solver
This solver is used for calculations of individual electric (E)
and magnetic (H) field intensity components inside and in close
proximity of aircraft electromagnetic model as well as appropriate surface currents that are excited by radiation on the aircraft skin.
In general, the fundamental mathematic model of electromagnetic field is represented by Maxwell equations. Computational electromagnetics includes many various ways of their
numerical solutions that can be classified for example according to the facts, if they are solved in their differential or integral forms or if the solutions are in the time or frequency
domains. Finite differences time domain method (FDTD) is
only one of the groups of possible numerical solutions, but it
has the widest spectrum of applications and commercial software implementations in spite of that it was firstly published by
Yee already in 1966. The following brief explanation can be
found in [4].
When Maxwell's differential equations are examined in this
way, it can be seen that the change in the E-field in time (the
time derivative) is dependent on the change in the H-field
across space (the curl). This results in the basic FDTD timestepping relation that, at any point in space, the updated value
of the E-field in time is dependent on the stored value of the
E-field and the numerical curl of the local distribution of the
H-field in space. The H-field is time-stepped in a similar manner. At any point in space, the updated value of the H-field in
C Z E C H A E R O S PA C E P R O C E E D I N G S
time is dependent on the stored value of the H-field and the
numerical curl of the local distribution of the E-field in space.
Iterating the E-field and H-field updates results in a marchingin-time process wherein sampled-data analogs of the continuous electromagnetic waves under consideration propagate in
a numerical grid stored in the computer memory.
Illustration of a standard Cartesian Yee cell used for FDTD,
about which electric and magnetic field vector components are
distributed, is in the left part of the Fig. 1 (Yee 1966). Visualized as a cubic voxel (volumetric pixel), the electric field components form the edges of the cube, and the magnetic field
components form the normals to the faces of the cube. A threedimensional space lattice is comprised of a multiplicity of such
Yee cells. An electromagnetic wave interaction structure is
mapped into the space lattice by assigning appropriate values
of permittivity to each electric field component, and permeability to each magnetic field component.
8
nents, and conversely. This scheme, now known as a Yee lattice, has proven to be very robust, and remains at the core of
many current FDTD software constructs.
Furthermore, Yee proposed a leapfrog scheme for marching
in time wherein the E-field and H-field updates are staggered
so that E-field updates are conducted midway during each
time-step between successive H-field updates, and conversely
(Yee 1966). On the plus side, this explicit time-stepping scheme avoids the need to solve simultaneous equations, and furthermore yields dissipation-free numerical wave propagation.
On the minus side, this scheme mandates an upper bound on
the time-step to ensure numerical stability (Taflove & Brodwin
1975). As a result, certain classes of simulations can require
many thousands of time-steps for completion.
MTLN Solver
This solver is used for calculations of the responses on wiring
structures embedded in the above mentioned aircraft EM
model. It uses the model based on the multi-conductor transmission line networks (MTLN) stressed by an electromagnetic
field. The wiring structures (models of cable bundles) can be
very large complex, they can be shielded or not and also dielectric insulators can be taken into the account. The following
brief explanation can be found in [5].
The MTLN solver is based on electromagnetic topology
(EMT) that relies on computations through multi-conductor
transmission line network (MTLN). The networks defined
through the code consist of cable lines (tubes) connected each
to other through junction. The junctions here represent avionic
units or any other components of the system under examination. The measured S-parameters or their equivalent computed
values are fed through the tubes to the junctions as an input.
The code thus analyzes the network, which has been ”volume
decomposed“ into junctions and tubes. The effect of network
radiation is simulated by introduction of external RF sources on
the cable network.
For large electrical systems, the EMT solution method
offers a strategy to handle the complexities associated with
electrical wiring in systems under examination. The first and
most important aspect of EMT is the assumption, that volumes
can be decomposed into subvolumes that can interact each to
other through a ”preferred path“ such as apertures and cables.
The assumption is based on the premise, that there is enough
shielding between the volumes and there is no energy transfer
through the shielding.
Fig. 1 — Illustration of a standard Cartesian Yee cell used for FDTD
(above). Volume discretization of an aircraft fuselage EM model and its
adjacent space into the Yee lattice (below)
This description holds true for 1-D, 2-D, and 3-D FDTD techniques. When multiple dimensions are considered, calculating
the numerical curl can become complicated. Kane Yee's seminal 1966 paper in IEEE Transactions on Antennas and Propagation proposed spatially staggering the vector components of
the E-field and H-field about rectangular unit cells of a Cartesian computational grid so that each E-field vector component
is located midway between a pair of H-field vector compo-
Fig. 2 — Left: Volume decomposition diagram. The entire system is
topologically broken down into different volumes. Right: Topological
network consisting of the external and internal levels, included within
each other
As mentioned above, the topological networks that result in an
EMT circuit are essentially composed of junctions, tubes and
sources. Junctions are elements that terminate or connect
9
LETECKÝ
transmission lines. Common types of junctions are open-circuited, short-circuited, 50-ohm and matched loads. Another
example of junctions is an n-port device that can be defined by
scattering parameters (S-parameters) or by any similar n-port
parameters. Tubes represent the connections between junctions
and can be defined from transmission line per length unit values (resistance per meter, capacitance per meter, inductance per
meter and conductance per meter). Parameters of junctions
and tubes can be derived from formulas, tables or measured
values. Sources can be defined by the results of the FDTD analysis and they can be either voltage sources (derived from the
local components of E-intensity) or current sources (derived
from the local components of H-intensity).
Z P R AV O D A J
currents at each port at each port of the junction and Zc is the
characteristic impedance. The scattering parameter (S) can be
written in terms of the admittance matrix (Y) as in equation (2)
with the characteristic impedance matrix at the junction of E2
with N1 conductors along tubes is represented through equation (3):
(2)
(S) = [(1) - (Zc)(Y)] [(1) + (Zc)(Y)]-1
Zc1,1
(Zc) =
...
Zc1,N1
:
:
:
:
ZcN1,1
...
Consider a large electrical system such as an aircraft shown in
Fig. 2 (left). Assume that the entire system can be decomposed
into sub-volumes. The aircraft in Fig. 2, for example, can be
decomposed into such volumes as cockpit, lower forward,
wing, tail and aft volumes which can interact with one another
through an interaction path such as apertures and cables. Inside the cockpit, there are situated critical and electromagnetically vulnerable electronic instruments (e.g. computers and
LCD displays) and wiring cables as main constituents of the
avionic system of the aircraft. The topological networks that
represent the system interaction are essentially composed of
junctions and tubes. Junctions that represent volumes are elements that terminate or connect transmission lines. Tubes
represent the connections between junctions and can be defined by transmission line per unit parameters. Inputs for junctions and tubes can be derived from formulas, tables, measured
values or simulated results. Fig. 2 (right) shows the diagram of
the topological network of the aircraft shown in Fig. 2 (left).
There are 5 main junctions and 4 tubes that form the upperlevel network consisting of the cockpit, the wing, the fuselage
and the tail. There exists inside the cockpit junction the lowerlevel network of the cockpit avionics and cable wirings.
For topological network equations, consider a simplified
sub-component of the large electrical system. Fig. 3 above
shows two avionic units in a subsystem that is represented by
two circular junctions connected by a tube-like structure. At
the junction of E2, vectors for both the incoming (W1(L)) and
outgoing (W2(O)) waves are related by the junction scattering
matrix (S12) through the scattering equation, expressed as
(W2(O)) = (S12) (W1(L)).
(1)
For multi-conductor lines, the incoming and outgoing waves
at the junction of E2 can be represented, respectively, by
(V)+(Zc(I)) and (V)-(Zc(I)) where (V) and (I) are voltages and
(3)
ZcN1,N1
Propagation of the incoming and outgoing waves along the
tube in the opposite directions with an equivalent distributed
source (Ws) can be characterized by a propagation matrix (Γ).
The resulting propagation equation becomes
(W1(L)) = (Γ)(W1(O)) + (Ws).
Fig. 3 — Two avionic units E1 an E2 connected together by a cable
(affected by an outer field) and their EMT representation
3/2010
(4)
Signals on an entire transmission-line network is expressed
through the BLT (Baum, Liu and Tesche) equation, which is in
essence, the multi-conductor transmission line (MTL) network
composed of the incoming wave vector, the outgoing wave
vector, and the source wave supervector (vector of vectors),
respectively shown above as [W(O)], [W(L)], and [Ws]. The
network scattering and propagation supermatrices (matrix of
matrices) are represented, respectively, by [S] and [Γ]. Using
the supervector and supermatrix forms shown by (1) and (4),
the BLT equation can be expressed as:
{[1] - [S] [Γ]} [W(O)] = [S][Ws],
(5)
where [1] is the identity supermatrix.
Solution of the BLT equations (5) have been incorporated
into the electromagnetic topological code CRIPTE developed
by French national aerospace research centre ONERA since
1990. It allows calculations of the response of complex cables’
networks in a frequency range up to several GHz. It was completed by the LAPLACE tool that allows the calculation of L
and C matrices of cable cross-sections.
3 Use of ANSA for electromagnetic model creation
When trying the first CEM simulations using the above mentioned FDTD and MTLN solvers, we created simple geometry
shapes and wiring structures using a special devoted pre-processor obtained as a part of the PAM-CEM suite. The EM
models, created directly using this pre-processor by linking of
simple entities at a level of school examples, were seemed perfect for meshing and consequential EM simulations on them.
A ball-shaped body of metallic material radiated by a planar
wave showed a predictable arrangement of surface currents on
its radiated hemisphere as well as EM field intensities in its
close environment. Simulation results of simple antenna dipole were predictably shown by the same way and we had a naive
idea simply to substitute these trivial geometric shapes by an
aircraft model and then we are prepared to solve any interaction of this aircraft with a severe electromagnetic environment.
Our initial enthusiasm was quickly turned to an absolute
gloom, when importing the aircraft CAD model transferred
using the neutral IGES format and the special pre-processor
announced more than ten thousand errors that prevented its
C Z E C H A E R O S PA C E P R O C E E D I N G S
subsequential processing. If the special CEM pre-processor
seemed to be suitable for healing the above mentioned simple
structures, its use for healing of the real aircraft might be considered as a real bottle-neck in the whole process of any CEM
simulation.
Fortunately, we had an opportunity to see our colleagues,
who prepare FE models for static calculations, crash simulations, etc., as they create their models using the universal preprocessor ANSA. We asked them for help to clear the bottleneck of the EM model pre-processing. After the above mentioned experience with the import directly to the special CEM
pre-processor, it seemed as a miracle, when our transferred
large-scale aircraft model was imported into this universal preprocessor with no error and using not only the neutral IGES
format, but also using the native CAD format (CATIA). When
the powerful tools of ANSA were consequentially used for
simplification and healing, we have obtained a quite superior
EM model that was meshed and exported using a mesh format
(NASTRAN) for an import into the special CEM pre-processor.
If the imported model is simplified, healed and meshed,
there was a more reliable way open for us — to import the
meshed model into the special CEM pre-processor with a minimum of possible errors the consequential required processing
(surface & volume meshing and EM simulation) might be prevented. When using an imported model prepared in this way,
we meet only a few problems with imported geometry and
these residual problems are well definable on the imported
meshed structure, so we can return to ANSA for a repair of
non-importable items only, which is very quick and comfortable for us.
Pre-processing of geometric models
The universal pre-processor ANSA [3] became the most
powerful tool for us during the aircraft EM model preparations. Preparing a geometric model, we proceed in the following
three steps:
- pre-processing using an original CAD,
- pre-processing using the universal pre-processor ANSA,
- pre-processing using the special pre-processor CEM.
The first step of the EM model pre-processing consists in
a removal of non-essential details (screws, rivets, small holes,
faceted edges, sliver edges,...) of substructures. The nonessentiality of any detail of the pre-processed substructure
depends on the maximum frequency of the model applicability (for the full-wave model, i.e. FDTD solver, it can be 1/12 up
to 1/10 of the minimum wavelength).
The simplified substructures are imported into ANSA and
the second step of pre-processing is initiated. The model substructures are associated with the information about materials
and material domains are created on them. The material domains are homogenous ”watertight“ closed volume bodies of the
same material (as interpreted by the FDTD solver).
The following processing of the material domains
depends on the kind of an associated material now. In case
of thin metallic materials (sheets), the domains are substituted by middle faces and in case of thick metallic or dielectric material, they are interpreted as volume bodies using
10
outer surfaces (upper and lower ones if necessary).
A special way of pre-processing is required for multi-layer
composite materials that may be provided by an additional protective metallic layer (e.g. expanded copper foil). For the purpose of integration of these materials into the FDTD model,
a special research was carried out for us by University of Thomas Bata (UTB) at Zlín. They decided to substitute any composite material by a fictive one-layer material with a constant
thickness and with the relative permittivity and loss angle corresponding to the reflection and transmission coefficients measured on samples of the original composite material. The material model derivation using genetic algorithms is shown in [6].
If the constant thickness of the material domain in the direction of aircraft radiation is necessary, only the outer surfaces of
these domains can be matched to a smooth aircraft surface and
the internal surfaces of these material domains are leaved
unmatched.
Then the various material domains are joined each to other
and their adjacent contact surfaces are matched by such a way,
when there is no unwanted contact-less part among them. The
model created by this way is meshed and exported to the special CEM pre-processor for the third step of pre-processing.
Having imported the geometric structure of the aircraft into
the special CEM pre-processor, an automatic checking of the
EM model correctness is necessary. If the special pre-processor discover any errors preventing the subsequent processing,
it is advisable to return with the model to ANSA and to make
the necessary corrections there. A healing of a model geometry using the tools of special CEM pre-processor is very laborious and time consumable.
If the EM model is errorless, the following operations have
to be carried out with it in the third stage of the pre-processing:
- addition of a model of examined wiring structures (see hereinafter) into the geometric aircraft model,
- setting of a volume discretization of the model and its close
environment (application of the Yee lattice as shown in Fig.
1),
- surrounding of the model and its close environment by unechoic layers. This significant improvement was brought into
the FDTD method by Berenger in 1994 and it makes a simulation of an infinite space around the examined model to prevent disturbing reflections of radiated field at the marginal
cells of the Yee lattice.
- setting of a time discretization for the FDTD calculations,
- setting of a radiation direction,
- setting of boundary conditions.
Having finished the third step of pre-processing, the EM
model is prepared for a surface and volume meshing, by means
of which the volume discretization is really executed as
a necessary condition for the subsequential running of the
FDTD solver.
Pre-processing of wiring structures
When cable bundles are embedded inside the airframe and we
examine the responses of a field radiation on them using the
MTLN solver, the model of an examined cable bundle is created in the similar three steps as the pre-processing of the EM
model geometry:
11
- pre-processing using the original CAD,
- pre-processing using the universal pre-processor ANSA,
- pre-processing using the special pre-processor CEM.
The first step of a wiring structure pre-processing in the original CAD consists in cancellation of unwanted branches of
the examined cable (the parts of the bundle, where no critical
signals are transmitted between aircraft equipment) and export
the cable bundle model into the universal pre-processor
ANSA, where the geometric model of the aircraft is already
prepared.
In the second step we have an opportunity to check, if some
cable paths are not in a collision with the geometry of aircraft.
This check using ANSA is very valuable for us, even if the
aircraft and cable geometries enter into the special CEM processor separately. Having repaired any collisions of the aircraft
and cable geometries, the cable geometry is imported into the
special CEM pre-processor as mentioned above.
For the specification and calculations of the BLT equation
(5) by the MTLN solver, the imported cable structure is defined using the tubes and junctions and boundary conditions are
subsequently defined for them. Also the electrical properties
of the cable (resistance, capacitance, inductance and conductivity) per unit of the cable length are to be entered in this third
step of the wiring structure pre-processing. If these cable parameters are not available by measurement for their direct entrance, they can be calculated from the geometric arrangement
of the cable crossectional structure using the LAPLACE tool.
LETECKÝ
Z P R AV O D A J
3/2010
solution, the MTLN solver can calculate required responses of
the wiring structure for the radiation. These responses could be
also displayed by a suitable post-processor. An example of
a critical cable arrangement in the small aircraft VUT 100
Cobra as well its cable response display is shown by the Fig. 5.
This cable response calculation was compared to the cable
response measurement in two specified cable sections and the
comparisons are also parts of the Fig. 5. If an uncertainty specification of such EM simulations and chained cable response
calculation is very difficult to be established by analytic methods, the experimental method of measurement and simulation of simplified aircraft fuselage model with a single shielded
cable was used for validation of the comparisons mentioned
above.
Processing of the EM model in its entirety
If the examined aircraft model as well as its internal wiring
structure are prepared for the next processing by the above
mentioned pre-processing steps, the surface and volume meshing is executed on this entire EM model.
The FDTD solver calculates all the E and H field intensities'
time histories at the edges and surfaces of all the Yee cells (Fig.
1) situated in the room (of the above mentioned unechoic layers) with the examined model inside. The calculated E and H
field intensities and surface currents can be displayed using
a suitable post-processor. An example of E-field intensity layout at a small aircraft is shown by the Fig. 4.
Fig. 5 — Frequency characteristic of open-circuit voltage amplitude
calculated on the cable embedded in airframe when radiated by a high
intensity radiated field (HIRF) in the band of 100 MHz up to 10 GHz
4 Results of the ARTEMIS research project and
possibilities of future development
Fig. 4 — E-field intensity lay-out at the small aircraft VUT 100 Cobra
when radiated by a surface wave with the frequency of 100 MHz
(upper picture) and 1 GHz (lower picture)
When the job of the FDTD solver is finished, the system
knows the field intensities in the neighbourhood of the examined wiring structure. When using them for the BLT equation
The previously mentioned ARTEMIS research project [1] was
established as the first breakthrough step of Evektor into the
practical CEM. However, the following applicable results were
obtained with essential help of partners in the consortium in
continuance of the project:
1) Procedures for effective creating of applicable geometrical
EM models using the ANSA universal pre-processor and
their meshing according to the FDTD solver requirements.
2) Applicable material models of protected or unprotected
composite materials or advanced plastic materials (ULTEM)
C Z E C H A E R O S PA C E P R O C E E D I N G S
to be used for a processing with the FDTD solver.
3) Validation of FDTD simulations’ results using experimental
test methods as well as comparisons with other Maxwell
full-wave simulations.
As mentioned above, the universal pre-processor
ANSA became the most valuable tool used for preparation of aircraft EM models in Evektor company. If we
take into the consideration a future use of other existing
and currently developed EM solvers and models (as an
applicability of the FDTD method is limited due to its
frequency range up to a few GHz), ANSA and the skilled engineers operating with this tool are the fundamental assumptions of the CEM division’s activities in our
company.
A participation of Evektor in the consecutive HIRFSE collaborative project (7-th Framework Programme
of EU) where a common platform of the EM model data
(AMELET) is established to match the data format for
various EM solvers (both the existing and future ones),
makes possible a continuance of this research. Evektor
is interested also in continuance to use ANSA as the
most powerful tool for creating of EM models and to
integrate it into the HIRF-SE framework as an organic
part making possible an effective pre-processing for any
EM solver as well as computational model. A foreseen
strategy is to create a data transmission wrapper (software submodule), by means of which the ANSA output
files are to be arranged in a structured HDF5 format
with a structure Amelet-HDF to be specified by our
partners within the HIRF-SE project. Output script of
ANSA is to be developed using the above mentioned
given data structure and in a close cooperation and with
a necessary help of BETA CAE Systems S.A.
Another question of the future development is a postprocessing of solvers’ data outputs. If we use the FDTD
and MTLN solvers purchased in the suite of PAM-CEM,
there is only a possibility of result post-processing using
the post-processor PAM-VIEW included in this suite. If we
consider the above mentioned use of existing and developed EM solvers in the common platform, there seems to be
no problem to read the output data of solvers plugged in
the HIRF-SE framework by the META post-processor and
to use its unique properties for EM observables' post-processing. There might be used a similar strategy of HDF5
data transfer from the common platform AMELET-HDF
into the META post-processor as in the case of ANSAAMELET data transmission mentioned above. If the
META postprocessor is plugged in the HIRF-SE framework, we can use the well-proven experience of our engineers to upgrade the EM result display and presentation
(see Figs 4 and 5) at the same level as in case of other
simulations of physical reality (virtual strength tests, crash
simulations, etc.), where a direct interface from solvers to
META exists.
5 Conclusions
The first and competitively used advantage of the application of universal pre-processor ANSA for CEM is the
12
possibility to employ up to 25 skilled engineers simultaneously when creating a large-scale EM model at the
Evektor company.
The second great competitive advantage that we wish
achieve with the help of international cooperation within
the HIRF-SE project is a possible consequent processing of any created EM model using various (existing as
well as newly developed) EM solvers. The solvers will
be based on a common platform and plugged in the
above mentioned HIRF-SE framework. It is the most
important facility, when there is no general-purpose EM
solver in CEM up to now, by means of which all the
requirements necessary for evaluation of electromagnetic threats of flight might be satisfied. We expect a precipitous research and development of various methods
and solvers in the field of CEM in the next future and
the effective creating of EM models using ANSA makes
our reliance to be successful in this field.
As the universal pre-processor ANSA has become
standard for effective creation of simulating models at
Evektor, we are pleased to take this opportunity and
express appreciation with using this powerful and reliable tool.
6 Acknowledgement
The research presented in this paper was financially
supported by the grant of the Czech Ministry of Industry and Trade FT-TA4/043 ARTEMIS (Analytic Research of Threats in ElectroMagnetically Integrated Systems) and it will be continued using Evektor’s participation on the large collaboration FP7 project HIRF-SE
(High Intensity Radiated Fields — Synthetic Environment), Grant Agreement No. 205294.
References
[1]
Final Report about Continuance and Results of the
Project FT-TA4/043 ARTEMIS under the TANDEM
Research and Development Program (Ministry of
Industry and Trade of the Czech Republic) in the
Period 2007 to 2010, Evektor, spol. s r. o., July
2010 (in Czech)
[2]
PAM-CEM /FD 4.1 User's Manual, ESI-Group 2004
[3]
ANSA version 12.1.5 User's Guide, BETA CAE
Systems S.A., July 2008
[4]
Finite-difference time-domain method. From
Wikipedia Free Encyclopedia
(http://en.wikipedia.org/wiki/Finite-difference_timedomain_method), June 2009
[5]
Islam, N. E.: Application of Advanced Concepts and
Techniques in Electromagnetic Topology Based
Simulations: CRIPTE and Related Codes; Research
Report No. TR-08-0522, University Missouri of
Columbia, December 2008
[6]
Goňa, S., Křesálek, V.: Derivation of Equivalent
Material Models for Composite Laminated Materials;
Radioengineering, 2009, Vol. 18, No. 3, September
2009, pp. 272-284
13
LETECKÝ
Z P R AV O D A J
3/2010
CFD Calculation of the Wing
with High-lift Devices
CFD výpočet křídla se vztlakovou mechanizací
Ing. Armand Drábek, Ing. Petr Vrchota / VZLÚ, Plc., Prague
This article deals with the CFD calculation of the initial configuration of an airliner wing equipped with double slotted flap and modified leading edge (Fixed Droop Nose + Double Slotted Flap). No slat on the leading edge was
used. Two deflections of the double slotted flap were calculated by EDGE program. This work is referring to previous article Grid Generation in ICEM CFD Software for CFD Calculation of Wing with High-lift Devices. The limitation of lift coefficient by Mach number was observed.
Tento článek se zabývá CFD výpočtem výchozí konfigurace křídla dopravního letounu s dvouštěrbinovou klapkou
a modifikovanou náběžnou hranou (sklopná náběžná hrana + dvouštěrbinová klapka). Konfigurace křídla byla bez
slotu na náběžné hraně. Tato práce navazuje na předchozí článek Grid Generation in ICEM CFD Software for CFD
Calculation of Wing with High-lift Devices (Generování sítě v programu v ICEM CFD pro CFD výpočet křídla se
vztlakovou mechanizací). Bylo vypozorováno omezení součinitele vztlaku Machovým číslem.
Keywords: Double Slotted Flap, fixed Nose Droop, Angle of Attack, Lift Coefficient, Drag coefficient, Single Slotted Flap, CFD.
Introduction
A big effort has been made recently to decrease the complexity of high-lift devices and the noise of aircraft. Simpler high lift
devices can be observed on modern aircraft. A highly efficient
simple slotted flap is used more often than complex triple slotted flaps, for example. The engine is the main source of the
aircraft noise during take-off, while high-lift devices and landing gear are the main sources of noise during approach. Sources of the noise can be roughly separated into two main parts.
The first is mechanical noise and the second one is aerodynamic noise. There are a lot of sources of aerodynamic noise.
One of them is flow passing at relatively high velocity through
slot. This work is the first part of a project, assessment of aerodynamic characteristic and verification of fulfilment of requirements of regulations. The other steps can be optimization of
position and geometry of the flaps if the requirements are not
fulfilled and the last step is aero-acoustic calculation. Comparison shows that this configuration produces less noise than the
previous one.
The main aim was to find out the possibility to satisfy the
requirements of the authorities on maximum lift and lift and
angle of attack during approaching by replacement of the slat
with a more efficient slotted flap.
The results were compared with the classical configuration
of high lift devices, slat on the leading edge in combination
with slotted flap on the trailing edge. Every calculation and
comparison of the results were done from the aerodynamic
aspect, only.
Condition of computations
A flow solver for unstructured grids, the EDGE code, is used
for all CFD calculations. The EDGE code was developed at
FOI. It is based on a finite volume approach with median dual
grids. The solver adopts an edge-based formulation for arbitrary elements and uses a node-based finite volume technique to
solve governing equations. The governing equations are integrated explicitly toward steady state with Runge-Kutta time
integration. The convergence is accelerated with agglomerati-
on multigrid and implicit residual smoothing. A central spatial
discretization is used for the convection of the mean flow and
a second-order upwind scheme is used for the turbulence, in
which the second order is enforced by a total-variation-diminishing limiter. A compact discretization of the normal derivatives of the viscous terms is used.
All calculations were performed as fully turbulent in RANS
mode with k-ω SST two-equation model of turbulence. The
grid was created as multiblock, structured from hexahedral
elements. Grid creation is described in [4].
The boundary conditions were placed in the sufficient
distance from the wing to have negligible effect on the flow
field around the wing. Boundary conditions were commonly
used in this type of external aerodynamic calculation. The free
stream velocity was M = 0.2 at the input boundary. All computations were made at 0 meters of International Standard
Atmosphere. The calculations were done for angle of attack
from 0 up to 10 degrees. Results of calculations over 10 degrees of angle of attack were not analyzed here because values of
lift coefficients still decreased in connection with the strong
separation of the flow on the wing. Flap deflections 35 and 40
deg. was analyzed.
Analysis of FND+DFS configuration
Maximum lift coefficient was 2,14 reached at 4 degree angle of
attack for higher deflection of the flaps. The slightly lower
value of the lift coefficient was obtained for smaller deflection
of the flaps for the same angle of attack (4 deg.) CL = 2,1. Lift
coefficient slowly decreases when angle of attack is increased
behind this value. It corresponds that the flow separation is
spreads from the trailing edge without sudden separation on
the leading edge. The values of the lift coefficient are affected
by separation of the flow for angle of attack higher than 5
degree. Unsteady calculation should be used for higher angle
of attacks. The lift and polar curves for both calculated variants are depicted in Figures 1 and 2.
The flow was separated on the suction side of the first
flap for higher deflection angle. The other separated regi-
C Z E C H A E R O S PA C E P R O C E E D I N G S
14
2,2
2,2
2,1
2,1
2
2
1,9
1,9
CL [-]
1,8
CL [-]
1,8
1,7
1,7
1,6
1,6
1,5
1,5
0
0
2
4
6
AoA [°]
8
40 deg
10
0,1
12
0,2
CD [-]
0,3
0,4
40 deg
0,5
35 deg
35 deg
Fig. 1 — Lift curves
ons were on the aileron for all range of angle of attack and
in the middle part of the wing for angle of attack higher
than 4 deg.
Visualization of pressure coefficient in Figure 3 shows
flow separation on aileron and the first flap even for small
angle of attack. Massive flow separation occurs at inner
part of wing (approximately 12 meters of wing span) for
angle of attack higher than 4 degree. This flow separation
extends to inner region of the wing at higher angles of
attack. Leading edge of a central part of the wing has a significant suction peak for all range of angle of attack in
comparison with the other part of the wing leading edge.
Flow separation was also observed in some regions of the
main flap. The flow around the second flap is without separation for whole range of angle of attack.
Fig. 2 — Polar curves of wing
a)
Analysis of single slotted flap wing
Results of configuration with double slotted flap were compared with previous computations of configuration with slat and
single slotted flap. Results of these previous calculations showed that significant separation on flap from 10 deg angle of
attack and from 14 deg angle of attack outboard wing separation was occurred. The streamline on the wing and flap surfaces are depicted in Fig. 4. The left side of this figure responds
the angle of attack 10 deg. while the angle of attack 14 deg. is
depicted on the right side. The lift curve is depicted in Fig. 5.
These results were adopted from our former colleague
Dr. Hrnčíř.
Comparison of FND+DFS and S+SSF configuration
Lift curves of both configurations, which were compared
with the previous calculation (slat + single slotted flap) are
depicted in Figure 5. FND+DSF configuration reached
much lower maximum lift coefficients than slat with single
slotted flap configuration. Maximum lift coefficient was
2.14 at 4 degrees of angle of attack for 40 degrees of the
flap deflection of FND+DSF configuration while the maximum lift of the configuration with the slat and single slotted flap was 2.78
Higher lift coefficients of FND+DSF were obtained only
in cases of small angles of attack in comparison with the
single slotted flap configuration (from 0 deg up to 6 deg).
The stall angle of the FND+DSF was much lower than
previous configuration. It can be caused by the absence of
the slat in combination with efficient flap system.
b)
Fig. 3 — Pressure coefficent: a) visible separation on the inner part of
the wing; b) suction peak at the central part of the wing
Limits of M∝2CL
High values of CL cannot be maintained indefinitely as
speed is increased, for soon surface pressure less than
absolute zero would be indicated. Theoretical limit of negative value of Cp can be found in [2]. A gas cannot be in tension. Hence the limiting suction pressure is a perfect vacuum over entire upper surface. The limiting pressure on the
lower surface is stagnation pressure. The limiting value of
Cp to reach sonic condition is for Mach number 0.2 of the
incoming flow -16.3.
Very high suction peak was observed during 2D calculation of the representative airfoil of the wing with
FND+DSF. The minimum Cp was in close proximity of leading edge and his value was -22. The Cp was below theoretical value of reaching the sonic condition and therefore
the small region on the leading edge with supersonic velocity was occurred. Minimum negative value of Cp was still
far from the theoretical limit, which is -35.8 for the input
15
LETECKÝ
Z P R AV O D A J
3/2010
Fig. 4 — Significant flap separation from 10 deg angle of attack and outboard wing separation from 14 deg angle of attack
Fig. 5 — Lift curves of FND+DFS and S+SSF configurations
Fig. 6 — Cp distribution for 2D verification
velocity M = 0.2 and represents pressure of absolute zero.
The distribution of pressure coefficient on the airfoil and
flaps are depicted in Figure 6.
This phenomenon was not observed during 3D calculation, because of the influenced flow by the tip of the wing.
The minimum value of Cp was in the middle of the wing
and it was roughly half value of the airfoils value.
Conclusions
The following conclusions have been made from results of
this study. Maximum lift of FND+DSF configuration was
reached at very small angle of attack (4 deg). It can be caused by absence of slat or small deflection of fixed nose
droop (high curvature of the leading edge) in combination
with relatively high deflection of efficient double slotted
flap. The separation of the flow for higher angle of attack
can be caused by high suction peak on the leading edge of
the wing (no suppression by slat effect). Slat is usually
used for reduction of the suction peak on the main wing by
his circulation (wing can operates in the range of higher
angle of attacks, as you can see in Fig. 6). It is possible that
supersonic region will be occurred even if the low-speed
calculation is done. This is highly probably especially for
optimization of aerodynamic characteristics of the wing
high-lift devices by 2D optimization of the specific cross
sections.
The variant of the wing with higher deflection of the flaps
gave higher lift coefficient, but the flow was separated on
the almost whole span of the suction side of the main flap,
while for the smaller deflection angle only the first flap had
separated flow and the second flap had flow without separation.
Separated flow on the aileron could be caused by lift
distribution along the span of the wing (higher lift coefficient on the flap than the value of the aileron lift coefficient aileron operated on the stall condition).
Higher lift coefficients were obtained in the range of
small angles of attack in comparison with previous case
(from 0 deg up to 6 deg). The maximum lift coefficient was
significantly lower than slat and single slotted flap configuration. It is caused by the absence of the slat (low stall angle
of attack) and separated flow on the flap and on the aileron.
It is possible to optimize the position, deflection and the
shape of the both flaps, now. For position optimization can
be used Nelder-Mead optimization method and for shape
optimization can be used genetic algorithm, for example.
A further step can be aero-acoustic calculation of both variants and previous one.
References
[1]
Anderson J. D. Jr.: Computational Fluid Dynamics The basics with
applications; McGraw-Hill International edition, 1995
[2]
Smith A., M., O.: High-Lift Aerodynamics; Journal of Aircraft, Vol.
12, No. 6, 1975
[3]
EDGE manual:
http://www.foi.se/upload/projects/edge/documentation-latest/edgequickstart.pdf
[4]
Drábek A., Vrchota P.: Grid Generation in ICEM CFD Software
for CFD Calculation of Wing with High-lift Devices; Czech
Aerospace Proceedings, No. 1, 2010
C Z E C H A E R O S PA C E P R O C E E D I N G S
16
Modern Design of Avionic Equipment for
the L159 Aircraft
Moderní konstrukce leteckých palubních přístrojů pro letoun L159
doc. Ing. Rudolf Jalovecký, CSc., Department of Aerospace Electrical Systems,
Defence University, Brno
This article introduces the aircraft cockpit set of devices, based on new principles, which has been developed for engine control systems of the aircraft series L159 and L159B. The design of all devices is based on the requirement to keep
current sensors of measured quantities. Therefore, characteristics of the existing emitted signals and analog emitters
are shown. After digitalization the signals are led to newly designed operating indicators working with digitalized signals from analog sensors and processed in a microprocessor system. The developed set of three devices has single
mechanical structure, in which only the input module and the appropriate software provide the desired function of the
device.
Článek informuje o kolekci leteckých palubních přístrojů pro kontrolu motorových systémů letounů řady L159
a L159B, postavenou na nových principech. Konstrukce vychází z požadavku zachování současných snímačů měřených veličin. Proto je uvedena charakteristika stávajících vysílačů signálů, na které jsou napojeny konstrukčně nové
ukazovatelé pracující na principu digitalizace analogového signálů ze snímačů a jeho následného zpracování v mikroprocesorovém systému. Vyvinutá řada tří přístrojů vychází z jednotné mechanické konstrukce, kde pouze vstupní
modul a následné programové vybavení zajistí požadovanou funkci přístroje.
Keywords: the digitization of the signal cabin air apparatus, air motor apparatus.
1 Introduction
The massive development of digital computer technology, the
miniaturization and production of microchips with rich peripheral equipment, their affordability and the extent of its use gives
a presumption in aviation technology, which penetrates into the
innovated cockpit equipment to indicate the flight and apparatus parameters needed for airframe or engine control. The
company MESIT přístroje spol. s r. o. plays a dominant role in
the field of development and production of on-board equipment in the Czech Republic. Already in 1996 the company
began to develop and subsequently construct an indicator of
the longitudinal balance within the framework of the project L159. This indicator is based on single chip microprocessor and its introduction into service was successful. Also other
two multiprocessor devices that followed — fuel quantity indicator and exhaust gases temperature indicator — for the
Aero L-159B aircraft have been successfully built and employed and implemented into use. Nowadays the company
MESIT přístroje spol. s r. o. designes and introduces a new set
Fig. 1
of three on-board devices — fuel quantity indicator, revolution
per minute (RPM) indicator and exhaust gases temperature
(EGT) indicator — which should replace the existing instruments supplied by the firm Ametek.
2 Arrangement of indicators and dashboard
mounting conditions
The current set of instruments in the aircraft L159B cockpit is
shown in Figure 1. The situation is shown where fuel quantity
indicator is a product of company MESIT přístroje spol.
s r. o. but the other three instruments are produced by Ametek.
The newly developed devices, their basic functions and its
labelling are:
Fuel Flow Indicator — LUN 1662 — measures fuel mass
consumption, this value is evaluated and displayed. Simultaneously the rest fuel total mass is calculated.
Revolution Per Minute indicator — LUN 8300 — measures
rotation speed of engine shaft, evaluates the speed and displays
it inside the interval of percentage range from 0% up to 110%.
Exhaust Gases Temperature Indicator — LUN 8390 —
measures, evaluates and displays the temperature of exhaust
gases.
In the technical specs award for the development of
three on-board equipment on aircraft L159 and L159B
occurred a number of components and operating conditions, such as mentioned are entirely understandable requirement consistent appearance of the existing apparatus
pointer, the same connector plugs, the same measuring
range, etc. As a very hard condition can be conservation the
existing transmitters. However, some severe requirements
have become advantageous because the manufacturer does
not have its own on-board instruments to address the problems of transferring non-electric quantities in the appropriate electrical parameters, and accuracy of the transfer
and distribution of measured quantity variables in the aircraft.
17
LETECKÝ
Z P R AV O D A J
3/2010
Fig. 2
Fig. 3
3 Characterization of current transmitters
Transmitter of immediate fuel consumption of company Eldec
Corporation (P/N 9-127-79 Fuel Flow Transmitter) has an
interesting structure and, therefore, if you zoom in more. In
Figure 2 is shown the principle of measuring the fuel flow. Fuel
by own flow spin turbine, which also rotates cylinder two fixed
magnets and is output through spring revolving segment with
two other magnets. A pair of fixed magnets in the first reel
gives the "start" pulse. The fact that fuel is conducted through
the cache management, which it directs on the rotating segment, which is driven through spring, there is slow-down the
rotation segment about angle, which is equal to the quantity of
flown fuel. Rotary magnets attached to the rotating segment
generated in the second coil signal ”stop“. The phase shift between the start and stop signal is directly proportional to fuel
flow. The manufacturer guarantees the transfer of linearity in
the range from 200 to 6200 ppm (Pounds Per Hour).
Transmitter speed (Figure 3) Company Sekura eng. what.
(transducer, motional P / UP / N K379-C42100), is a simple
pulse generator with a sinusoidal output signal at a frequency of
0 to 18,414 kHz. Indicated value of 100% RPM (Revolutions
Per Minute - the number of revolutions per minute) is defined
signal frequency 16,740 kHz, with a sensitivity threshold at
around 10% RPM speed signal with a frequency of 1679 Hz.
Transmitter output gas is temperature thermocouple (Fig. 4)
type K (chromel-alumel) Honeywell Company (P/N 30947052) with a range of measurements to 1150° C
4 Construction of new concept characteristics
The design characteristics of motor systems can be divided into
electrical and mechanical part. In both cases, was to unify the
developers as much as possible the design characteristics and
thus more efficient than their production.
4.1 Electrical design characteristics
Electrical design of all three indicators based on research studies [1, 2] and previously developed combinations Fuel for
aircraft L159, RAVEN and revolution for a number of Zlin 141
aircraft, 142 in Figure 5 indicated block diagram of electrical
design characteristics. Despite the fact that the orientations of
the three devices aimed at enterprise use of the popular and
often used processor ADuC812 occurred in the course of development work to ”change“ to newer and more powerful type
of ADuC841. Company Analog Devices has introduced the
processor as a direct substitute for the previous type and they
equip the processor. Given the general trend of companies
Fig. 4
engaged in the development of single microprocessor has this
much more memory to the program (64kB), built external
RAM (2kB), more internal SRAM (4kB) and several other
embedded peripherals such as 2 channels PWM (Pulse-Width
Modulation) and ”fourth“ internal counter (applicable only to
the management of communication over the serial interface
with speeds to 115 kBd zero error when generating communication speed).
In addition to the integrated peripherals are connected to the
microprocessor further external modules.
Input modules
The only module that is different in each index, the adjustment
module input signal. In fuel quantity indicator regulates both the
input signals from the transmitters of that so that the entry counters in the microcontroller was brought rectangular signal (TTL)
of the variable width of the 16-bit counter in the microprocessor
to evaluate the accuracies of 0.5 s. Due to the measurement of
pulse width from 2.0 to 60 ms is sufficient accuracy.
Also, the tachometer input module provides an adjustment
signal generated by the rectangular signal of variable pulse
Fig. 5
C Z E C H A E R O S PA C E P R O C E E D I N G S
18
Fig. 6
width proportional to speed. Further processing is the same as
for quantity indicator.
Input module for the input gas thermometer ensures the processing of very low voltage from the thermocouples, the voltage gain to 2,5 V (corresponding to a temperature of 1200° C)
and, of course, compensate for the cold.
Modules to display the measured data
To view the measurement data is in all three used a combination of indicator analogue and digital display measured values.
Analog information is displayed above the scale of hand
movements; the movement of the hand is used for stepper
motor with its control. The kernel driver stepper engine is
made up of integrated circuit IMT901, which allows up to 8division multiple electronic step, thus movement of the hand
scale even more soft (stepper motor has a value of one step of
0.9° at microsteping to be 0.1125°). From the microprocessor
is controlled by the driver only two pulses, indicating the direction of rotation and the number of performed steps, which
allows implementing very simple control algorithm in the microprocessor hand position, but also brings the need for initial
synchronization ”zero“ position of hand over the power scale.
Therefore, all three of the index (and in fact the fourth —
Fuel), carried out when you turn on the so-called ”zero“ hand
over the movement of an optical sensor retro movement to 0
over the scale pointer. While the revolution of quantity meter
and the entire scale is linear, the thermometer is a non-linear
scale, respectively break-down in the value of 300° C. This
software solves the non-linearity of a microprocessor.
To view the digital information is used quadruple seven-segments bulb-display with a high life and a defined wavelength
emitted of yellow-green light. This display is controlled by
a set of four 8-sliding housing registers, which are controlled
by the microprocessor SPI protocol (four wires).
The device on an aircraft used in both day and night, is
equipped with a pointer under shine. The intensity of this backlight and brightness of the display microprocessor controls two
channels using PWM modulation based on the digitization of
the input voltage management podsvitu.
The need for information on the measured parameters in the
FDR (Flight Data Recorder - flight recorder) is out for all devices in the range of analog signal from 0.25 to 4.75 V, which is
spreading to the extent measured. The adjustment scale and
the control microprocessor controls, using one of the two 12apartment D / A converter. Impedance adjustment, tension and
gain protection against short circuit at the output provides the
basic operational amplifier circuits.
Since the ADuC841 microprocessor is equipped with the
possibility of programming the soldered state (In-Circuit Programmable) and also for the possibility of setting up all parameters of each device is equipped with a microprocessor
module standard RS232 serial interface, which communicates
with PC at a speed of 19200 Bd. All setup and of course, control function is implemented a special program on the PC.
4.2 Mechanical design characteristics
Mechanical design is based on the characteristics of previous
devices tested structures designed on a similar principle (number of Fuel and revolution). In Fig. 6 is foreseen mechanical
construction equipment. As stated in the electrical modules.
Fig. 5 and are structurally identical modules of printed circuit
boards. Equipment includes:
- Board of microprocessor where the microprocessor
ADC841, crystal, and other districts RS232 auxiliary circuits (gates, filters in the circuits of power, etc.).
- Board resources, which are made up DC/DC converter, providing office equipment in the range of supply voltage 1536V, the necessary protection against the transmission and
protection against electromagnetic pulses.
- Board of display, where the quadruple seven-segment display, four shifts registers, lamps control transistors providing
PWM brightness control and optocoupler to capture the
starting position stepper engine.
- Board of engine drivers, including their own integrated
motor drivers and resistors for the implementation of diagnostics without malfunction stepper engine.
- Board of input module, which is soldered directly to the
input connector and contains all the circuits to adjust the
input signal according to the type of pointer.
5 Software and equipment setup program on the PC
All software of all three characteristics is written in assembler
for the CPU core 8051 and is implemented by functional
modules.
The function of the individual program modules are described in the Plan of software aspects [3,4,5] and can be divided
into several areas:
- Handling integrated processor peripherals (counters, A/D
and D/A peripherals, internal EEPROM memory, PWM outputs, etc).
- Handling external peripherals indicator (stepper engine displays, light bulbs podsvitu, RS 232, etc.).
- Handling interrupt vector (RS 232, pulse counters, RTC circuit, etc.).
- Implementation of all the mathematical operations required
in processing measured quantity variable (digital filter, computing speed, fuel, gas outlet temperature) is always in units
suitable for controlling stepper engine, or display. D/A converter.
- Diagnostics indicator (stepper engine connection of transmitters, etc.).
In Figure 7 an example of diagram thermometer output gas is
shown. After setting the necessary initial conditions in an infinite loop solves the following tasks:
- Measure the input voltage from the thermocouples and digi-
19
LETECKÝ
Z P R AV O D A J
3/2010
Fig. 7
Fig. 8
talize to digital values.
- There will be diagnostics of digitized voltage.
- There will calculate the digital filter at a moving average
with a triangular dependency.
- Individually hand position is calculated stepper engine to
display the temperature (at the same time solves the nonlinearity over the scale).
- Calculate the temperature of output gas for display devices
(digit).
- Calculated temperature for the analog output voltage for
MFD (12-bit data for D/A converter).
- Measure the voltage for the management of managing the
display brightness and implements the management of the
brightness of bulbs and light display.
- Implements the diagnosis of the entire device (activity stepper engine connection of transmitters, etc.).
A similar manner is realized software revolution and quantity
meter. Of course, all the algorithms are modified for processing measured. For quantity meter is included in the cycle calculation module to calculate the total fuel consumption, which
is realized by integrating the time instant fuel consumption and
maintenance of two buttons located on the device housing. The
first one is used for short-term view of total fuel consumption
of the display and the second button then, while pressing the
first, reset the total fuel consumption.
Software for PC, which is used to setup and control functions of all three characteristics, is realized in the development
environment, Delphi 7 Professional. Basic program window is
shown in Figure 8. The connection is realized with a serial
interface RS232. The program includes a common part again
to serve the same peripherals, and then a separate three modules, which are activated by selecting the type of pointer. The
Fig. 9
program can be controlled and set the content of internal and
external EEPROM ADuC841 processor, processor to record
a program. The contents of memory can record how the parts
(for each type of peripheral devices and independently) or at
the end of development, once the entire memory at once.
Conclusion
Development of those on-board equipment for the control of
engine has been accompanied by a wide range of technical and
operational problems. Not negligible are also operational tests
and tests for resistance to high negative temperatures (-55° C)
because the aviation directives are strict. Figure 9 shows the front
panels of the developed devices. And from their look it is apparent that the mechanical construction is exactly the same.
Acknowledgement
Work presented in this paper has been supported by the Ministry of Defence of the Czech Republic (Research Plan No.
MO0FVT 0000403).
[1]
[2]
[3]
[4]
[5]
References:
Jalovecký, R. a kol.: Indicators with Stepper Motors for Air
Application; Preliminary Research Report, Brno, 1996 (in
Czech)
Jalovecký, R. a kol.: Indicators with Stepper Motors for Aviation Application. Technical Solutions to the Electrical Part of
the Indicator with a Stepper Motor for Aviation; Brno, 1996
(in Czech)
MESIT Přístroje spol. s r.o.: Plan for Software Aspects of
LUN1662, (in Czech)
MESIT Přístroje spol. s r.o.: Plan for Software Aspects of
LUN8300, (in Czech)
MESIT Přístroje spol. s r.o., Plan for Software Aspects of
LUN8390, (in Czech)
C Z E C H A E R O S PA C E P R O C E E D I N G S
20
Hardware in the Loop Simulation
of FBW Components
HIL simulátor FBW komponent
Zdeněk Hanzálek 1 , Pavel Hospodář 2 , Martin Hromčík 3 and Libor
Waszniowski 4 / Department of Control Engineering, Czech Technical
University in Prague, Faculty of Electrical Engineering
Jiří Doubrava 5 / AERO Vodochody
The paper presents hardware in the loop simulator developed for validation of control algorithms and electro-hydraulic servo actuators designed for flight by wire control system of a small jet plane. The architecture and functionality of
the flight by wire system is briefly described and requirements on validation tests to carry out on the hardware in the
loop simulator are mentioned. The flight by wire control system under development consist of dual channel flight control computer and dual channel electronic control unite controlling dual channel electro hydraulic servo actuator with
mechanical redundancy. It supports several control modes as autopilot and recovery mode. An architecture and software and hardware components of the developed simulator and hydraulic stand are described. The simulator instrumentation is based on distributed peripherals communicating via CAN open industrial field bus. The real time simulation of the plane model is performed via an embedded computer with Power PC processor and Linux operating system. Matlab Simulink environment is used for the development of plane models and control algorithms and data visualization and evaluation. FlightGear, an open source flight simulator, is used for flight visualization. An example of
hardware in the loop simulation test is presented.
Článek popisuje HIL simulátor vyvinutý pro ověřování řídicího algoritmu s elektrohydraulickým servomechanismem
pro navržený FBW řídicí systém proudového letounu. Stručně je popsána architektura a funkčnost FBW řídicího
systému včetně požadavků na ověřovací zkoušky prováděné HIL simulací. Vyvíjený FBW řídicí systém se skládá
z dvoukanálového řídicího počítače navazujícího na dvoukanálovou řídicí elektronickou jednotku elektrohydraulického výkonového servomechanismu s mechanickou redundantní vazbou. Navrhovaný řídicí letový počítač podporuje
letové módy od autopilotních až po vybrání letu z nezvyklé polohy-recovery. Dále je popisována architektura SW
a HW komponent ve vyvíjeném simulátoru a navazujícím hydraulickém zkušebním stendu. Vybavení simulátoru je
založeno na rozšiřitelném periferním zařízení komunikační sběrnice CAN. Real Time simulace letových modelů je
prováděna vestavěným Power PC s operačním systémem Linux. Prostředí Matlab Simulink je použito pro vývoj letových modelů, řídicích algoritmů, zobrazování dat a vyhodnocení. Dále je pro letové zobrazení letu použit letový simulátor Flight Gear. V závěru je uveden příklad provedené HIL simulace s hodnocením.
I. Introduction
This paper presents proof of concept validation of components
of the Flight-by-Wire (FBW) control system via hardware in
the loop (HIL) simulation. The validated components are control laws of the flight control computer (FCC), dynamics of the
new electro-hydraulic servo actuator (EHSA) with mechanical
redundancy and control logic of dual channel, fail-safe electronic control unit (ECU).
The basic principle of the hardware in the loop (HIL) simulation is based on connection of real hardware devices (EHSA
with ECU, hydraulic and mechanical subsystems and control
surface in our case) with a computer based simulator providing
a real-time numerical simulation of the rest of the system (control law of FCC and plane in our case) in the closed loop. The
1
Associate Professor, Department of Control Engineering, Czech Technical
University in Prague.
2
Research Assistant, Aeronautical Research and Test Institute in Prague, and
Department of Control Engineering, Czech Technical University in Prague.
3
Research Assistant, Department of Control Engineering, Czech Technical
University in Prague.
4
Research Assistant, Department of Control Engineering, Czech Technical
University in Prague.
5
Development Engineer, Department of Aircraft System Development,
AERO Vodochody, Ltd.
idea of connection of the real part of the system with the modeled rest of the system is based on the following arguments.
Due to the presence of the real devices in the closed loop, the
simulation is as near to the reality as possible in this validation
phase. Due to the utilization of the model in the closed loop,
the simulation is cheaper, easier and less time consuming than
experiments on the real device, risks of the damage of the real
device are eliminated and simulation of the system behavior in
extreme modes of operation which are hard, dangerous or even
impossible to induce on the real device is possible. The absence of the real device in the simulation particularly simplifies the
development of an airplane control system, since it allows systematically and effectively validates the system, even if the
airplane or a certified component is currently not available.
A measurement and data logging implementation is much easier and flexible on a stand than on a plane. All input signals are
deterministic in the computer based HIL simulation. The
reproducibility of an experiment is therefore easier to achieve
as compared to fly tests.
Contrary to poor numerical simulation, HIL simulation gives
more comprehensive evidence of the system correctness,
which is important especially in conservative industries as aviation is. Our experience is that management responsible for our
project has education and know-how in mechanical enginee-
21
ring. HIL simulation is therefore more acceptable for them
than numerical simulation, even though it would be precise
enough. Concerning the price of HIL simulation compared to
the poor numerical simulation, it is worth mentioning that the
construction of the stand used for HIL simulation does not
necessary increase the cost of development because devices
used in HIL simulation can usually be used also in another test
as integration tests or aeroelasticity tests that must be done.
The objective of the project in which we have adopted the
HIL simulation is to design a concept of the FBW control system of the small jet plane, demonstrate the new features it
brings and validate it by proof of concept experiments. The
intention is to upgrade current control system by implementing
new features of active control improving pilot comfort and
control quality. These functions include fast oscillations dampers, care-free control, autopilot and recovery mode. Because
it is not convenient to change completely the concept of the
control system (from mechanical with hydraulic actuators to
FBW) at once, it was decided to plan the changes of the control system in several successive steps, where each next step is
based on the validated components and concepts of the previous one.
There are three main concepts of the new FBW control system considered in our work. Since the current control system
use mechanically signaled and hydraulically powered actuators (the airplane is naturaly stable) it was decided to maintain
this mechanical control system due to its proved reliability and
redesign the hydraulic actuators to mechanically and electrically signaled hydraulic actuators (Electro-Hydraulic Servo
Actuators - EHSA) by augmenting them by electric servovalves, in the first phase of the control system upgrade. Electric
signaling via FBW system provides primary control, while
mechanical system provides redundancy. This system therefore joins the reliability of the mechanical system with the possibility to implement some modes of active control via FBW
system. The maintained mechanical linkage of the control stick
with the controlled surface (via mechanical signaling of the
actuator and its mechanical feedback), however, prevents
implementation of dampers of the aircraft fast oscillations,
since fast deflection of the controlled surface would be undesirably transmitted to the stick.
Different concept of the control system must be used to
decouple the stick and controlled surface, which would allow
unrestricted active control. Before employing fully reliable
FBW, which would allow removing mechanical redundancy,
corresponding know-how and experiences are required. Therefore, the next step toward implementation of the active control
system seems to be installation of second actuator connected in
series between the main actuator and the surface. Since the
output of this serial actuator is not coupled with the mechanical control system, it can provide a new degree of freedom for
active control implementation. Since the serial actuator become rigid in the case of fault in FBW system, it does not degradate the reliability of the mechanical redundancy.
The final concept of active control system is fully reliable
FBW system without mechanical redundancy, which, however
represents the biggest change of the current control system
concept.
LETECKÝ
Z P R AV O D A J
3/2010
Our attention currently focuses especially on control laws of
the FCC and EHSA from the following reasons:
❑ FCC control laws are central to FBW functionality and
new features brought by FBW installation.
❑ FCC control laws depend on the airplane dynamic and
control quality requirements. It is therefore better to
design control laws in-house and not to buy third party
solution.
❑ EHSA is central to safety of the control system.
❑ EHSA is central also to integration with current system.
Therefore EHSA is based on the existing hydraulic devices already used on the plane.
❑ Actuators dynamics and nonlinearities affect the close
loop dynamic. They must be therefore examined and considered in control laws design and FBW evaluation.
❑ EHSA is supposed to be developed and manufactured inhouse due to existing know-how and technology in this
field.
On the other hand, we do not focus on flight sensors, inceptors and hardware of computers since their development is supposed to be subcontracted.
A modular HIL platform is used to evaluate possible contribution of the FBW to the airplane controllability. Control law
designed in Simulink is executed by embedded computer
representing FCC. The real EHSA with the most important
mechanical parts and hydraulic subsystem is mounted on hydraulic stand and controlled by dual channel ECU. The angle of
the control surface is measured via sensor and the aerodynamic
force loading the rudder is imposed via a hydraulic cylinder.
The system contains an embedded computer executing the
numerical model of the airplane in real time. Peripherals are
interconnected via CANopen field bus, which makes this architecture very flexible and extensible. Simulink provides user
interface, which allows flexible test design and evaluation in
the same environment where the model and control law was
developed. The flight simulator FlightGear (FG) is used to evaluate the FBW system from ergonomic point of view by pilots.
The rest of the paper is organized as follows. The FBW
rough architecture and basic functionality is described in Section II. This chapter also defines requirements on validation
tests. Section III deals with the developed HIL simulator. It
describes she simulator architecture, the stand with mechanical
and hydraulic equipment and its electronic instrumentation. An
example of a validation test passed on the HIL simulator is
described in Section IV. Concluding remarks are mentioned in
Section V.
II. Control System
This chapter briefly describes architecture and functionality of
the current control system of the plane and the considered
upgrade to FBW system with mechanical redundancy.
A — Modes of control
We consider three hierarchical levels of aircraft control:
1 Aircraft dynamics accommodation — (e.g. increasing
damping ratio of short period components of aircraft
motion)
2 Autopilot — controls and stabilizes aircraft position angles and speed
C Z E C H A E R O S PA C E P R O C E E D I N G S
3 Flight Control Systems (FCS) — aircraft control accordingly to specified flight track using navigation variables
for both vertical and horizontal aircraft guidance, typically: control altitude, control vertical speed, final approach
for vertical (glide slope-GS) and horizontal (localizer LOC) level (with instrument landing system), horizontal
leading with VOR system and recovery.
The mentioned hierarchical levels can be accomplished manually by pilot or automatically by a FBW system. It is also meaningful to accomplish lower levels (e.g. short period dampers)
by FBW and the upper levels by pilot. Realize, however, that
this is not possible in FBW system with mechanical redundancy (see Section II.C). The mechanical coupling of control stick
with control surface would transfer the short period actuation
of the surface to the stick which must be hold by pilot. We suppose, and we will validate it by HIL simulation, that this will
not bring any problem when FBW system accomplish also
higher hierarchical level of the aircraft control (at least autopilot), since pilot does not need to hold the stick in this case.
To allow using automatic dampers without autopilot an
additional serial actuator must be used or mechanical redundancy must be removed [1].
B — Current Control System
Currently, the airplane is controlled by typical irreversible control system based on mechanically signalled hydraulic actuators. The schema of the control system is depicted on Figure 1.
Fig. 1 — Scheme of the current control system
The control signal is transmitted from control stick to
actuator via system of mechanical rods and levers. Hydraulic actuator is equipped by mechanical feedback system which moves the hydraulic spool to the position proportional to the difference of the control sick deflection
and the control surface deflection. The hydraulic spool
controls flow of hydraulic medium to the hydraulic
cylinder where it moves piston driving the control surface. Reliability of the actuator is provided by redundancy
of the hydraulic part. There are two hydraulic cylinders
with pistons on the common rod and two spools also on
common control rod. Both these hydraulic systems are powered from two separated hydraulic circuits.
Since the control system is irreversible, an artificial feel of
force nonlinearly proportional to stick deflection is provided by
Artificial Feel Unit (AFU — system of mechanical levers,
springs and an actuator). The neutral position of the force provided by AFU can be moved by electric trim actuator.
The control system is equipped by autopilot (AP) stabilizing
aircraft position angles. It acquires fly data from avionic system
and controls electric actuator deflecting the control stick.
Actuator is disengaged or engaged by clutch. AP disengages
22
itself when a force measured by Force Sensor (FS) indicates
that pilot is trying to overpower AP. AP always trims to zero
force to prevent sudden stick deflection by AFU in the case of
AP disengaging due to a fault.
Notice that this concept of AP implementation has some
disadvantages. Since the AP actuator is installed at the beginning of the mechanical system, the control process suffers
from nonlinearities (such as slashes and friction) in the closed
loop. Moreover the AP actuator must drive the mass of the
mechanical system and overpower the force of the AFU. On
the other hand, its torque is limited to ensure fault tolerance of
the control system (pilot must be able to overpower it in the
case of a fault).
C — FBW with mechanical redundancy
To overcome disadvantages of autopilot implementation to the
mechanical control system, as described in previous section, an
upgrade of the control system to FBW system with mechanical redundancy has been proposed (see Figure 2).
The hydraulic actuator has been equipped by electric servovalves (one for each one channel) and set of switch and bypass
valves providing disengaging of the servovalves in the case of
a fault. This composes dual channel Electro-Hydraulic Servo
Actuator (EHSA) with mechanical redundancy.
Position of the EHSA is controlled by dual channel Electronic Control Unit (ECU - one for each servovalve). Each servovalve has two coils driving its spool. Each coil of each servovalve is driven by different ECU channel. The
final spool position is therefore average of actuation of both ECU channels.
Similarly, disengaging valves can be controlled from both ECU channels. Primarily, both
electrical servovalves are controlled synchronously by both ECUs. When a fault of one ECU
occurs, the second ECU still can control both
servovalves (but with half gain since only one
Fig. 2 — Scheme of FBW system with mechanical redundancy
coil is used). When a servovalve fails, it is bypassed by
a bypass valve and the corresponding cylinder does not lock
the second one. When either both servovalves or both ECU
channels fail, cylinders are switched to mechanically controlled spool. All valves are designed to be fail-save, therefore the
EHSA is switched to the save mode (mechanically controlled)
when control voltage or hydraulic circuit pressure is lost.
Reference of the EHSA position is obtained from dual channel Flight Control Computer (FCC) providing plain control
according to specified control mode (see Section II.A). Communication of FCC and ECU is provided via dual CAN bus.
23
The described system of redundancy is depicted on Figure
3. It shows that all components of the FBW system that are not
considered to be sufficiently reliable (FCC, CAN, ECU) are
dual channel and are connected in the way that each component is connected to the both channel of proceeding and following component. Therefore when faults of two different components, even in different channels, occur (e.g. fault of CAN1
and ECU2) the system as a whole does fail.
Fig. 3 — Concept of FBW system redundancy
Since our work focuses on control laws and actuators, our
FBW system dedicated to proof of concept experiments consist of dual channel FCC, dual CAN bus, dual channel fail-safe
ECU and dual channel EHSA with mechanical redundancy.
The remaining parts of the system are simulated by HIL simulator.
Fig. 4 — Electro-Hydraulic Servo Actuator
D — FBW Test requirements
To validate the functionality of the FBW system described in
Section II.C, several HIL simulations have been specified. The
goals of these tests are:
1 to identify the real dynamic and static characteristics of
the control system
2 to evaluate the control quality
3 to evaluate robustness of the control system
4 to analyze the effect of faults in the system
5 to evaluate acceptability of automatic control system
maneuvers for pilots
These tests are carried out:
a) in several points of the flight envelope
b) for different configurations of the airplain (especially for
different rotational inertias)
c) off-load and for various aerodynamic load of the control
surface
Identification the real dynamic and static characteristics
of the control system
The dynamic model of the airplane and many mechanical and
hydraulic components of the control system are nonlinear.
Moreover, dynamics of some components (hydraulic actuators
in our case) is unknown at the design time, since these components are still under development. The control law design is
therefore based on a simplified linearized model omitting or
estimating unknown dynamics and nonlinearities. Even though
LETECKÝ
Z P R AV O D A J
3/2010
the robustness analysis is used to verify the control law at the
design time, an experimental validation of the system behavior
based on real system dynamics must be done at the evaluation
time.
Identification of the real control system dynamics can be
done in the open loop. Various amplitudes and periods of harmonic input signals are used to find the nonlinear behavior.
The most important issues were the dynamic and static characteristics of the new actuators.
Evaluation of the control quality
A metric of control quality (e.g. the rate of damping)
is measured in the system with the real dynamics
(containing real actuators and mechanical component
with reduced inertia) and real nonlinearities (friction
and backslashes of mechanical components, hysteresis of force sensor and actuators). It is measured in the defined
point of the flight envelope and the effect of the aerodynamic
load is evaluated via simulation of this load in the HIL simulator.
Evaluation of robustness of the control system
Backlashes, friction and other nonlinearities in the closed loops can cause undesirable oscillations of the
system. Effect of these nonlinearities can progress due
to wear of mechanical components. These oscillations
can be excited by appropriate signals during the HIL
simulation. The intention is to identify the source of
these oscillations and tune the controller to be robust
against them.
Analyzes of the effect of faults in the system
It is necessary to validate the operation of the system
under various fault conditions. Various faults should be injected to the electrical system to verify that the system is fail-safe.
Very important is to validate the effect of switching from
electrical to mechanical signaling since it can cause very fast
movement of control surface.
It should be validated that ECU identifies and isolates faults
and change control mode accordingly. Control quality and
robustness should be evaluated also in a degradate mode of
control (e.g. fault of one hydraulic circuit).
Evaluation of acceptability of automatic control system
maneuvers for pilots
Acceptability of maneuvers performed by automatic control
system by pilots must be evaluated on a flight simulator. It is
important especially for modes of the highest hierarchical
levels, e.g. recovery.
III. Hardware in the Loop Simulator
In this chapter we describe the HIL simulator developed for
FBW tests. Architecture of a general HIL simulator is depicted
in Figure 5. Sensors and Actuators provide a physical interface
with the Control Unit Under Test. The model of the controlled
plant is simulated by the Control and Simulation Calculation
block. This block also controls the data acquisition and actuation and controls the overall simulation process. The simulation is executed in real-time. The most commonly used mathematical model of the controlled plant is discrete and supposes
C Z E C H A E R O S PA C E P R O C E E D I N G S
24
Fig. 6 — Data Flow of HIL simulation
Fig. 5 — Architecture of a Typical HIL System [1]
equidistant synchronous sampling. Sensors, Actuators and
Control and Simulation Calculation block must be therefore
precisely timed and synchronized.
As important as closed loop simulation itself are supporting
tools providing Human Machine Interface (HMI) of the simulator, development and parameterization of the model and Post
Simulation Analysis. While blocks composing the closed loop
with the Control Unit Under Test (Sensors, Actuators and Control and Simulation Calculation) affect the accuracy of the
simulation, the supporting tools (blocks HMI and Development and Post Simulation Analysis) affect the efficiency of the
development and validation process.
There are several factors to consider when designing a HIL
system [1]:
❑ The system should accept a variety of control unit configurations.
❑ A small change in the control unit must not warrant
a design of a completely new system.
❑ The system should perform both open and closed-loop
testing.
❑ The system should be scalable and open.
❑ The system should be of reasonable cost in terms of components and development time.
A — Architecture
Based on the requirements on HIL simulations specified in
Section II.D, HIL simulator architecture has been proposed.
Data flow is depicted in Figure 6. Simulation is conducted in
the closed loop of Plane Model, Flight Control Law, ECU and
EHSA with control surface and related mechanical system.
This data flow is equivalent for all axes. Visualization is provided by flight simulator Flight Gear.
The described data flow is deployed to the architecture
depicted on Figure 7. All components of the FBW system are
installed on the hydraulic stand representing the central part of
the system. Since different experiments require only some
parts of the hydraulic stand, we distinguish two stands — aircraft hydraulic stand and actuator hydraulic stand (see Section
III.B). Both stands can be equipped by load actuators simula-
Fig. 7 — HIL simulator architecture
ting aerodynamic force. Simulation of the plane dynamic is
provided by digital Simulator (see Section III.C). The model
for simulation is developed by Matlab Simulink which is used
also for data acquisition and visualization. Flight visualization
is provided by flight simulator FlightGear. Real time communication is provided via CAN bus with CANopen protocol.
Communication with development environment and FlightGear is provided via Ethernet with TCP protocol.
B — Hydraulic Stand
The Hydraulic stand is the central part of the HIL simulator. It
consists of the real hydraulic actuator and related hydraulic
equipment, mechanical transmission and the hydraulic actuator
LA, which is controlled to simulate the aerodynamic load of
the control surface. The hydraulic stand therefore contains all
real components providing any undesirable features as a friction, backlash and other nonlinearities or unknown dynamics. It
is therefore the crucial part of the HIL simulator need for the
system validation.
We distinguish aircraft hydraulic stand (Figure 9) and actuator hydraulic stand (Figure 8). Aircraft hydraulic stand contains complete hydraulic and mechanic control system for three
Fig. 8 — Actuator hydraulic stand with one EHSA and load actuator
25
axes. It contains also cockpit with stick and FlightGear flight
projection. Aerodynamic load of control surfaces is usually
simulated by spring, but hydraulic load actuators can be installed if required. Completeness of the control system allows
validation of the flight control laws and interaction of the FBW
system with the current control system.
Actuator hydraulic stand is subset of the hydraulic and
mechanic system consisting of only one EHSA and load actuator. This smaller system allows conducting experiments
related to EHSA easier. It can be configured for different
experiments such as measurement of static and frequency
characteristics of actuators, analysis of resistance to autooscillations, measurement of actuators stiffness, analysis of
sensitivity to reduced mass change and test of reaction on
faults.
C — Simulator
Simulator is an embedded computer providing data acquisition
and real time computation of simulation steps. Its HW is based
on the MPC5200B embedded processor with 760 MIPS Power
PC core. It is equipped with 128MB SDRAM, 32MB Flash,
10/100 Ethernet, two CAN ports and many other peripherals.
Its SW is based on open source components. As an operating
system is used Linux 2.6, as a driver of CAN is used SocketCAN and as an implementation of the CANopen protocol is
used CANFestival (http://www.canfestival.org/).
D — Flight visualization
To validate acceptability of automatic control system maneuvers by pilots, the flight visualization is provided by the flight
simulator FlightGear. The aircraft dynamics is computed by
simulator and FlightGear only visualizes the flight variables on
monitor or projection.
Experiments can be conducted in cockpit of the aircraft
stand (see Figure 9) where the linkage of the stick with EHAS
is maintained. This cockpit also maintains main ergonomics
aspect of the real aircraft (stick deflection range and corresponding force).
Also second experimental cockpit can be used (see Figure
12). This cockpit is intended for FBW system without mechanical redundancy. It is equipped with side stick without mechanical linkage of the stick and the surface. Therefore, it can be
used for validation of control modes that can not be used on
FBW system with mechanical redundancy (e.g. dampers).
E — Actuators and Sensors of the HIL simulator
Control Surface Angle Sensor
The control surface angle is measured via absolute single-turn
optical shaft encoder with 13 bit resolution and high measurement linearity. It is an industrial device compliant with CANopen standard. It is designed for really synchronous position
acquisition of several axes. Therefore it is suitable for synchronous measurement in HIL simulator.
Load Actuator and Force Control
The aerodynamic force loading the rudder is simulated via
Load Actuator (LA) consisting of hydraulic cylinder and servovalve. The servovalve is controlled by a digital controller
ForceCtrl. The force is measured by four foil strain gauges forming the full Wheatstone bridge attached to the connecting rod
LETECKÝ
Z P R AV O D A J
3/2010
of the cylinder. Even though an AC power supply of the bridge can be used to eliminate disturbances, the DC power supply of the bridge has been used, since the measured force can
change dynamically (the frequency band of the controller has
been estimated to 1kHz at the design time) and the required
frequency of the AC power supply would be therefore too
high.
The value of the current aerodynamic force, which is used as
a reference for the loading force, is computed by ForceCtrl
according to the following equation:
(1)
FAeroLoad = kα ⋅ α + kβ β
where:
α
is deflection angle of the rudder
β
is drift angle
kα and kβ are constants
Alternatively, the reference for loading force can be obtained
from the Simulator computer as a constant or time varying signal Fref.
ForceCtrl is connected to two CANopen networks. It is
a slave in the main network used for simulation control. This
network is synchronized with the frequency of simulation
(more than 12ms). ForceCtrl receives β and Fref from Simulator computer and transmit α and FMeasured to Simulator computer via this network. ForceCtrl is also a master of the second
network used to acquire α from the Rudder Angle sensor. This
network is synchronous with the ForceCtrl controller period
which is 1ms.
IV. Example of Validation Test
In this section we present an example of HIL test. A short period damper is used for demonstration, since it best demonstrates the influence of the effect of EHAS dynamics.
A — Short period damper
The longitudinal motion consists of two oscillating parts. First
part is short motion (short period) and it is damped relatively
high. But for more convenient and more precise flight control
is necessary to damp this part of motion. Regulator for linear
model is designed by the help of a root locus method. Damping of short period satisfies standard MIL-F-8785 C for category A and C 0.35 < a < 1.3. In some cases is used wash out
filter for filtering of steady-state value. Final damper damps
only short period motion (component of pitch rate with higher
frequency). Result of the Model in the loop simulation (see
model in Figure 13) in time and frequency domain are depicted
in Figure 14.
Fig. 13 — Model for Model in the loop simulation
C Z E C H A E R O S PA C E P R O C E E D I N G S
26
Fig. 14 — Step response and Bode Diagram of the damper
B — Identification
Real system includes dynamic of actuator and distance-velocity lag in communication. Distance-velocity lag could destabilize system. Therefore we must indentify this dynamic and include it to the control system. For this we use ARX identification method (least-square method to estimate parameters).
Comparison of measured and estimated elevation is on the
Figure 15.
C — Smith predictor
Next we must remove a problem with distance-velocity lag.
For this we use the Smith predictor. Feedback with model system in regulator deducts outer feedback (real feedback), but
only inner feedback without distance-velocity lag is working.
This controller has been simulated in the loop with real
EHSA (see Figure 16). Result of this simulation is on Figure
17.
Fig. 15 — Response of the model and real system
V. Conclusion
We have described the HIL simulator designed for validation
of flight control algorithm and hydraulic actuators behavior.
The most important components of the simulator have been
described and their crucial parameters have been mentioned.
The design of the simulator is tailored to the architecture and
parameters of the FBW system under development and requirements on validation tests.
The most important features of the simulator are its open
architecture allowing extensibility and integration to the Matlab environment allowing efficient test design and post simulation analysis.
A practical experience with the developed simulator has
been demonstrated.
Fig. 16 — Model for HIL simulation
Acknowledgement
This work was supported by Ministry of Industry and Trade of
the Czech Republic under Project FT-TA3/044.
References
[1] Waszniowski, L., Z. Hanzálek, P. Hospodář, M. Hromčík
and J. Doubrava: ”Hardware in the Loop Simulation of FBW
components“, In Proceedings of the AIAA Modeling and
Simulation Technologies Conference, 2009
[2] LabVIEW FPGA in Hardware-in-the-Loop Simulation
Applications, National Instruments Corporation, 2003
Fig. 17 — Comparison of damped response (HIL simulation) and non
damped response (model)
Editorial Note: Some formerly mentioned pictures are
printed on the outer cover page of this issue in colour for
better clarity.
27
LETECKÝ
Z P R AV O D A J
3/2010
Fatigue Prediction Based on Finite Element
Analysis of Riveted Joints
Únavová predikce na základě MKP analýzy nýtovaného spoje
Ing. Jan Papuga, Ph.D. / Evektor, spol. s r.o., Uherské Hradiště - Kunovice
The paper reports on an application of the PragTic freeware fatigue solver to fatigue predictions of riveted joints. The
input data for the analysis concern very detailed finite element models and thus the fatigue analysis follows the same
trend evaluating local stresses and strains. Although the input data were not complete as regards the material
parameters, the evaluation process brought along interesting experience, which is commented here.
Článek popisuje použití freewarového únavového řešiče PragTic na únavovou predikci nýtovaných spojů. Vstupními
daty pro takovou analýzu jsou velmi podrobné MKP modely a tak i únavová analýza sleduje podobný trend ve
vyhodnocování místních napětí a deformací. Ačkoli vstupní data týkající se materiálových parametrů nebyla
kompletní, proces analýzy přinesl některé zajímavé zkušenosti, které jsou zde popsány.
Keywords: post-processing of a finite element analysis, riveted joints, fatigue analysis.
Nomenclature
b
[-] . . . . . . . . . . . . . . . . . .fatigue strength exponent
c
[-] . . . . . . . . . . . . . . . . . .fatigue ductility exponent
E
[MPa] . . . . . . . . . . . . . . . . . . . . . . .tensile modulus
εa
[-] . . . . . . . . . . . . . . . . . . . . . . . . . .strain amplitude
ε'f
[-] . . . . .fatigue ductility coefficient in axial loading
FEM / FEA . . . .finite element method / finite element analysis
γ'f
[-] . . . .fatigue ductility coefficient in torsion loading
LR
[-] . . . . . . . . . . . . . . . . . . . . . . . . . . . .lifetime ratio
N
[-] . . . . . . . . . . . . . . . . . . . . . . . . .number of cycles
υ
[-] . . . . . . . . . . . . . . . . . . . . . . . . . . .Poisson's ratio
σ'f
[MPa] . .fatigue strength coefficient in axial loading
σl
[MPa] . . . . . . . . . .lower nominal stress at the cycle
σm
[MPa] . . . . . . . . . . . . . . . . . . . . . . . . . .mean stress
σu
[MPa] . . . . . . . . . .upper nominal stress at the cycle
τ'f
[MPa] .fatigue strength coefficient in torsion loading
Introduction
In order to optimize the structural design phase of its current
and future aircraft projects, Evektor aims also at using the
modern trends in computational structural analysis, evaluating
their possible assets, and, if possible, on including new methods into its portfolio of analyses. One part of such attempts
was recently described in [1], in which detailed finite element
analyses of riveted joints were reported. Both [1] and the work
described here are outputs of IMPERJA project (No. OE08006
of EUREKA programme) ”The Fatigue Performance of Riveted Joints in Airframes“. Thanks to this support, the analyses
were possible in such detail that would not be otherwise feasible during the common engineering design.
The riveted structures are hard nuts to crack, because of the
high deformation induced during the riveting and the complex
stress and strain field around rivets. The paper [1] mentioned
among others also various analytical methods used for estimating the rivet flexibility and opposed to their simplicity, which
allows their quick use but with obvious neglecting of many factors that influence the real joint. The situation in fatigue prediction is the same, because the used analytical methods can
hardly follow all the details of external load types or joint con-
figuration. The necessity to use the fatigue S-N curves of an
identical joint type can be very negative constraint limiting the
applicability of such a solution or even invalidating the analytical outcome.
On the other hand, the detailed computational analysis faces
other problems — be it the time-consuming preparation and
analysis phases or e.g. questionable covering of the reality including too many interacting effects, which were in some way
mastered in the analytical computation used for years long, or
where the uncertainty of the results was either known or prescribed in standards. Its applicability in the aerospace industry is
thus nowadays limited mainly to the use of the finite element
technique, but the fatigue prediction is not the same case and
the analytical empirical solution is prevalent. Nevertheless, the
knowledge on various fatigue prediction models is continuously improving as well as the calculation speed, enabling so use
of more complex prediction methods. The fatigue life computation based on detailed models is therefore already feasible,
but the outcome has to be further studied so that its transferability to real engineering practice can be evaluated.
Test specimens
The specimens used in the analyses were of two types. They
are marked as EVE1 (see Fig. 1) and EVE2 (see Fig. 2). They
are typical representatives of riveted joints used in Evektor in
new airplane designs. EVE1 specimen relates to a real riveted
joint as used in the airframe of the VUT100 Cobra four-seat
airplane. The EVE2 specimen corresponds to a typical real
riveted joint of the bottom side of integral wing tanks of the
EV-55 Outback small utility airplane. The specimens were tested at Brno University of Technology (EVE1) and at VZLU
(EVE2) at four load levels each. Thanks to their chain-like
composition, such testing provided sufficient data sets for
determining of two S-N curves, because the experimental campaign thus generated 16 results for EVE1 or 32 results for
EVE2. Unfortunately for the fatigue predictions reported further, the fatigue experiments reported only on the total lifetime
covering the whole duration of dynamic loading until the final
rupture of individual joints. The crack initiation phase has not
been separated in any way. Final geometric mean lifetimes
valid for individual load levels are reported in Tab. 1.
C Z E C H A E R O S PA C E P R O C E E D I N G S
Fig. 1 — Geometry of the EVE1 specimen and its further specification
28
Fig. 2 — Geometry of the EVE2 specimen and its further specification
Material data
Tab. 1 — Final average experimental lifetimes achieved for the two
specimen types at various load levels. The numbers relate to the total
lifetime until the complete break of the specimen.
FEM models
The finite element models that are post-processed during the
fatigue prediction are the main topic of another paper [1].
A brief sketch follows here. The models were built in
MSC.Patran and solved with MSC.Marc. Only one half of
each joint was modelled thanks to the symmetry conditions,
which means that 3.5 rivets were modelled in each of the two
analyses. The mesh quality can be evaluated in Fig. 3. Interfaces between individual parts in the model were set as touched
contacts, allowing the separation of both bodies and then buildup of another contact. The friction coefficient on the interfaces
was set to 0.12 in conformity with Müller [2], because no own
measurement of this value was taken.
The sealant layer in both specimens could not be modelled
at all due to the already too extreme demand on the calculation time and because of convergence problems found. The quality of the models in [1] was checked with quasi-static experiments concerning a simpler two-rivets configuration without
any presence of the sealant. Some discrepancy can be therefore anticipated due to its neglect, but we had no other option in
the end. The quasi-static experiments of simpler joints including the sealant layer have not been made, nor their analysis in
MSC.Marc, because this would require extensive tuning of the
visco-elastic material model of the sealant.
Fig. 3 — Example of
the meshed locality
around the central
rivet in EVE1 specimen
In addition to uncertainties related to omission of the sealant
layer and unknown ratio of the crack growth phase in the total
life of the riveted joint, the quality of material parameters input
into the fatigue prediction model was another problem that we
had to face. Within the IMPERJA project, only the tests of
complete joints were proposed and no basic material experiments were conducted. A search for available open data resulted in the comparison depicted in Fig. 4. While we can see that
the curves from various papers are quite close to each other for
high-cycle fatigue, the trend in low cycle fatigue below 1000
reversals differ. Here in these cases, the expected lifetimes in
Tab. 1 are above this threshold. In the end, the curve referred
to by Everett [6] was chosen for next analyses. A comparison
of the trends in Fig. 4 can show that the highest ratio of lifetimes between the chosen and any other curves is approximately two. The final retrieved data are reported in Tab. 2.
Although the curves are quite similar in the area of interest,
we cannot be satisfied. None of available data sets corresponds
to the clad 2024T3Clad (EVE2) and no reference to D16C
(EVE1) material was found at all. In the end, we had to use the
only data set we had in hands for the two specimen types.
Usual assumption for clad duralumin is that the lifetimes of
such materials are lower, i.e. the curve should lie below the
curve valid for non-clad duralumin [7] and thus by using the
Everett data we should end with over-optimistic results with
lifetimes higher than real.
Fig. 4 — Strain-life curves for 2024T3 and 2024T351 materials
as found in various papers
29
LETECKÝ
Tab. 2 — Material properties of 2024T3 extracted from Everett [6]
Fatigue prediction methods
The analysis is based on detailed FEA-models, which means
that the local solution utilising either stress-life or strain-life
curves has to be applied. The most important difference between the two methods is that the stress-life concept processes
elastic stresses, while the strain-life solution is using real elastic-plastic data. In many other applications, the stress-life solution is handier, because of the possibility of skipping the timeconsuming non-linear material solution. Nevertheless, the task
solved here concerns the case where both the material and
contact non-linearities had to be involved in order to be able to
simulate the riveting process. Theoretically, the stress-life
solution still could be used for fatigue life prediction, but the
local loads have to be extrapolated to the real level from two
load levels at which Hooke's linear condition is valid. Under
such conditions, some doubt would still pertain as regards the
validity of such an assumption because of the contacts that are
changing during the shear loading. During further analyses, it
was found here for EVE1 specimen that the material below the
countersunk head is remaining plastic even after the riveting
tool removing. This means that no elastic region, from which
the extrapolation to the final load level could be realized, is
available. Even the theoretical hope on examination of stresslife solution in our evaluations is gone.
Another peculiarity of strain-life methods is that the unknown fatigue life demands implicit solution. The basic Basquin and Manson-Coffin formula states:
σ'f
b
c
ε a = ε a,el + ε a , pl =
⋅ (2 N ) + ε ' f ⋅(2 N ) .
E
The total strain amplitude εa is divided into its elastic (subscript
el) and plastic (pl) parts, which are further related to the final
numbers of reversals 2N or full cycles N. Because of the addition of the two parts together, the solution is not explicit. The
use of Newton-Raphson iterative formula for the retrieval of
the result prolongs the calculation time substantially in comparison with the explicit calculation available by stress-based
methods.
The equation above is the basic formula used in the methods
described further. It refers to strain amplitude only and does
not incorporate the mean stress effect at all. A lot of various
formulations on including it in the solution were proposed (see
uniaxial methods section in Help for PragTic [10]). Two most
well-known methods are the Smith-Watson-Topper (SWT)
method:
(σ a + σ m )⋅ E ⋅ ε a
= σ 'f ⋅(2N ) + E ⋅ ε 'f ⋅(2N ) ,
b
c
and the Landgraf formula, in which the mean stress effect
affects only the elastic part of the total strain-life formula:
σ ' −σ
b
c
ε a = f m ⋅ (2N ) + ε 'f ⋅(2N ) .
E
3/2010
Z P R AV O D A J
Although there are many other proposals, these two were chosen for our purposes because of their general acceptation, simplicity and the fact that no other material parameter is involved
in their use. There is no sense in applying any more advanced
formula available in the PragTic solver if the basic material
parameters are only estimates.
The note above on classification of the formulas to the group
of uniaxial methods concerns the kind of processing the local
load history. The complex strain (or stress) tensor is transformed to one scalar equivalent strain (stress) value, which is then
processed in the implicit solution mentioned above. Also here
more potential kinds of solution exist. The use of von Mises
stress and effective strain, both signed by the sign of first invariants of the stress and strain tensors respectively can be the
right choice:
(
1
(ε x − ε y )2 + (ε y − ε z )2 + (ε z − ε x )2 + 3 γ xy2 + γ yz2 + γ zx2
2 
2
ε = sign (ε x + ε y + ε z )⋅
1 + ν eff
)

,
ε pl
ε el
+ ν pl
.
ε
ε
1
2
2
2
σ = sign (σ x + σ y + σ z )⋅ (σ x − σ y ) + (σ y − σ z ) + (σ z − σ x ) + 6 τ xy2 + τ yz2 + τ zx2 .
2
where
ν eff = ν el
[
(
)]
Because the tensor reduction using the square root leads to
positive values only, the signing can be necessary, but only in
some specific cases, e.g. for fully reversed loading of an unnotched specimen, where the transformation to pure Mises stress
would lead to repeated loading in equivalent parameters. On
the other hand, the local loading is very complex in the case
presented here (high preload caused by riveting, multiple contact surfaces, etc.), and the external loading is only in push, i.e.
no periodical push-pull loading is induced. The experience of
the author is that the choice of the right model is not simple, if
the local loading is so complicated as here. Therefore, both the
signed and pure (not signed) variants of stress and strain tensors reductions were checked.
In addition to these two variants, our early attempts included
also the variant with the sign set from the sign of principal
stress with the maximum absolute value. Such a solution is
used e.g. in the fatigue wizard in Ansys. Nevertheless, the output of this method was not likely to be realistic because of
a plenty of isolated points with localized damage and no obvious relation to other highly damaged points (see Fig. 5).
Because of the local load complexity, our early attempts
included also checking the multiaxial kind of solution —
Socie’s combined method ([8], [10]) more precisely. The
multiaxial methods were designed in order to cope with the
problem of signing noted above by adding other parameters to
the equivalent local load value. The Socie method necessitates
use of the strain-life curve obtained during the torsion loading.
Because such data are available seldom, Socie himself [8]
proposes to use the same exponents b and c also for the torsion
loading, whereas the following engineering formulas should be
applied to, when retrieving the torsion strain-life curve material
parameters from axial load data:
τ 'f = σ 'f
3
γ 'f = ε 'f ⋅ 3
.
The scarce complete information reporting both axial and torsion strain-life curves does not legitimate such assumptions.
C Z E C H A E R O S PA C E P R O C E E D I N G S
30
All four material parameters can differ quite substantially from
these estimates, though they seem to be acceptable for checked
structural steels1. Under the condition of so limited credibility
of material parameters, the application of a solution with such
an increased complexity is not legitimate.
conservative (safe) side, with LR parameter higher than 1. The
parameters describing the critical node locations in the local
cylindrical systems established in each hole (see Fig. 6) are
relative distance from the hole center:
Use of the PragTic fatigue solver
α angle measured from the direction of the external acting
force and relative depth:
PragTic is a freeware fatigue solver programmed by the author
of the paper. It can be downloaded for free from
www.pragtic.com website. It is a tool developed because of the
research focus of the author on complex problems in multiaxial
fatigue analysis, but its domain of usability is substantially
enlarged in an attempt to make it a universal fatigue solver. At
this moment, it offers either strain-life or stress-life solution
including multiaxial loading that can be done either at isolated
points (strain gauges data) or on a FEA-model.
The details on real practical use of PragTic during fatigue
analyses of the riveted joints are provided in [9]. In short, the
FEA data was transferred from MSC.Marc to PragTic via
MSC.Patran, which can generate the ASCII output formatted
acceptably for the universal ASCII import function in PragTic.
Because the import function allows definition of a file structure map, the input of FEA result files could be optimized as
regards its total duration. This fact is noted here because of the
necessity to get into PragTic quite a lot of individual result files
both in stress and strain tensors at many load increments.
The necessity to work solely with strain-based fatigue calculation methods enforces direct preparation of local load histories of stress and strain tensors at all evaluated notes. Their
composition is realized within PragTic by defining the load
regime as a sequence of various result files (see [9]).
The output of the calculations by the two methods described
previously is exported from PragTic to a FEMAP neutral file.
Different variables based on the damage value are exported
within it, so that the damage map on the FEA model was easily readable. The optimum variable chosen for graphical evaluation of results is the sixth root of damage2.
Results
Except for running the analyses by SWT and Landgraf methods of calculation, two kinds of equivalent strain and stress
calculation were adopted — one without any signing, while
second with signing by the sign of the first strain or stress invariant (I1). The most critical localities for all four variants can
be examined in Tab. 3 together with the lifetime ratio calculated from Npred predicted number of cycles and Nexp number of
cycles retrieved in experiment:
LR =
N exp
.
N pred
Majority of the values achieved in the calculations is on the
1
Such a check can be done on the basis of material parameters
available in the material database on the page
www.pragtic.com/vmat.php. Currently, the materials SNCM630,
S52C, S45C, SCM435, SCM440, SF60, SFNCM855, SNCM439,
9254 AL FG, StE460, 1045, AISI 304, AISI 1141 and AISI 1141 MA
(i.e. no duralumin at all) are there with the two strain-life curves data
provided.
2
The same variable is a choice for graphical interpretation of damage
results e.g. in FemFat (www.femfat.com).
r∗ =
z∗ =
x node
,
r hole
z node
t sheet
where the znode distance in the z direction is measured from the
top surface of each particular sheet. The holes are numbered
according to Fig. 7. The meaning of the top side corresponds
to the side nearer to the countersunk head, while the bottom
side is the one close to the driven head. In addition to the values provided in Tab. 3 an example of graphical outputs for the
SWT criterion without signing and EVE1 specimen is depicted
in Fig. 8.
Fig. 6 — Definition of local cylindrical coordinate systems
Discussion
The cells highlighted by grey colour in Tab. 3 show the results
for the calculation method nearest to the experimental reality.
Apparently, the variant of the tensor reduction without any
signing leads to better results here. The signing by I1 results in
extremely conservative prediction. No similarity between critical places for the two kinds of tensor reduction can be found.
There is still quite a distinct disproportion between the prediction and experiment at lower number of cycles (load regimes
LR1 and LR2 above all). It is necessary to remind here that the
strain-based fatigue prediction methods are intended to be used
only for the calculations till the crack initiation. In the low
cycle fatigue regime (high loads, here LR1 load case above
all), the crack initiation covers only several percents of the total
life of the structure, while it corresponds to an almost entire
lifetime in the high cycle fatigue (here LR4) — see e.g. [11] if
interested in such an analyses. In consequence, a part of the
deviations found can be attributed to this fact.
Another comparison that can be done is between the results
for the SWT and Landgraf criteria. Results provided for the
SWT method are usually better when compared with the retrieved from the Landgraf solution.
Unfortunately, the minimum knowledge of real material
parameters related to the specimens tested in EVE1 and EVE2
riveted joints does not allow any more substantial evaluation of
the final results. The analyses presented here thus show above
all the potential hidden in the industrial use of local fatigue prediction methods in fatigue post-processing of FE-analyses of
riveted joints. Nevertheless, even with material parameters
derived from own experiments on the same lot of sheets, the
31
LETECKÝ
Z P R AV O D A J
3/2010
Tab. 3 — Results describing the final lifetime ratio LR and localizing the critical nodes. Top sheet is the one with the countersinking
more elaborate analysis should cover either analysis of the
onset of the crack growth phase (i.e. end of the crack initiation
phase) or involvement of the crack growth prediction. If the
second option is introducing another unknown error to the final
prediction, the first one is realizable only by spending a lot of
time and money for the early crack detection.
In correspondence to these facts, the potential of the local
detailed solution presented here seems to be depleted. The problems related to obtaining the converged and correct solution of
the FE-analyses were already discussed in [1]. The model of
three and half rivets was touching the limits of the FEA-solver
available in Evektor. There is some hope related to the continuous development of hardware and software, but the increase in
the task size to e.g. 100 rivets does not seem to be routinely solvable in 10 years from now. A quick applicability of the local
detailed solution is further restrained by the complexity of the
problem when preparing the fatigue analyses setups. Thus,
because the interest of Evektor in riveted structures is not fading
off, our further attempts in fatigue predictions on riveted joints
will be directed to completely other solution and will focus on
analyses of internal forces carried through individual nodes and
the spring elements as described also in [1].
on a riveted structure. The analyses of two different joint types
follows the local trend, with complete modelling of the rivet
and sheets in detail, including also the riveting phase in the FEA
(see [1]). Fatigue post-processing presented here is based on
accessibility of the PragTic fatigue freeware, in which the setup
of the transient analysis based on FEA-result sequence is enabled. Although the results for the SWT method with tensors
reduction by Mises rule without any signing could be evaluated
as acceptable under given conditions and limitations, the results
themselves are not sufficiently supported by experimental analysis of used duralumin sheets. Only the complete joint was tested, and so no more substantial analysis of retrieved results
could follow. Because of the complexity of the problem highlighted in the discussion section and the extreme computation
time (used for the FEA solution preparation above all), the further attempts at our company are likely to follow the way of
simplified models composed from 1D and 2D elements.
Acknowledgements
Conclusion
The work described in this paper was realized within the
IMPERJA project ”The Fatigue Performance of Riveted Joints
in Airframes“ (No. OE08006 EUREKA programme) with
a financial support provided by Ministry of Education, Youth
and Sports of the Czech Republic.
The presented paper describes the application of a modern fatigue prediction tool based on FE-analysis results to a problem
highly specific to aerospace industry — fatigue life prediction
Editorial Note: Some formerly mentioned pictures are
printed on the inner cover page of this issue in colour for
better clarity.
References
[1]
[2]
[3]
[4]
[5]
[6]
Fárek, J.: FE-Modelling Methodology of Riveted Joints; Czech
Aerospace Proceedings, Vol. 2010, No. 2
Müller, R. P. G.: An Experimental and Analytical Investigation on
the Fatigue Behavior of Fuselage Riveted Lap Joints; Delft
University of Technology, Delft 1995
Duprat, D.; Davy, A.; Boetsch, R.; Boudet, R.: Fatigue damage
calculation in stress concentration fields under variable uniaxial
stress; International Journal of Fatigue, Vol. 18, No. 4, 1996, pp.
245-253
E. U. Lee; A. K. Vasudevan; G. Glinka: Environmental effects on
low cycle fatigue of 2024-T351 and 7075-T651 aluminum alloys;
International Journal of Fatigue, Vol. 31, 2009, pp. 1938-1942
Fatemi, A.; Plaseied, A.; Khosrovaneh, A. K.; Tanner, D.:
Application of bi-linear log-log S-N model to strain-controlled
fatigue data of aluminum alloys and its effect on life predictions;
International Journal of Fatigue, Vol. 27, 2005. pp. 1040-1050
Everett, R. A.: The effect of load sequencing on the fatigue life of
2024-T3 aluminum alloy; International Journal of Fatigue, Vol. 19,
1997. pp. 289-93
[7]
Schijve, J.; Jacobs, F. A.; Tromp, P. J.: The significance of cladding
for fatigue of aluminium alloys in aircraft structures; [NLR TR
76065 U]. NLR, Amsterdam, 1976
[8]
Bannantine, J. A.; Socie, D. F.: A multiaxial fatigue life estimation
technique. In: Advances in Fatigue Lifetime Predictive Techniques,
ASTM STP 1122. Eds.: M. R. Mitchell and R. W. Landgraf.
Philadelphia, ASTM 1992. pp. 249-275
[9]
Papuga, J.: PragTic in Application - Fatigue in Riveted Joints;
CD-ROM with presentations from WCFA'09 & PUM3 meeting;
CSM, CTU in Prague and ALV, Prague, 2009*
[10]
Papuga, J.: Help for PragTic. Prague, 2008**
[11]
Socie, D. F.: Critical plane approaches for multiaxial fatigue damage
asessment; In: Advances in Multiaxial Fatigue, ASTM STP 1191.
Eds: D. L. Dowell and R. Ellis. Philadelphia, ASTM. 1993. pp. 7-36
*
**
The presentation is available online on
http://www.pragtic.com/UM3ttbl.php.
Available online on http://www.pragtic.com/program.php#help.
C Z E C H A E R O S PA C E P R O C E E D I N G S
32
Static Strength Verification Procedure
of the Real Composite Structures
Metodika ověření statické pevnosti reálných kompozitových
konstrukcí
Ing. Vladimír Snop / VZLÚ, Plc., Prague
The static strength tests of real composite structural parts and large parts of primary aircraft structure
have been performed in VZLU strength test lab recently. For strength verification of these structural parts
there was necessary to find and verify philosophy of strength tests solving, design implementation of experimental loads, tests performance and results evaluation. The philosophy was verified during static tests of
landing gear beams and fuselage bulkhead structural parts according to standard conditions and after
conditioning and during static test of small transport aircraft model fuselage section. Only philosophy of
static strength test of model fuselage section is described in this article.
V pevnostní zkušebně VZLU byly uskutečněny v nedávné době statické pevnostní zkoušky reálných kompozitových
uzlů a velkých částí primární konstrukce letounu. Pro ověření pevnosti těchto konstrukčních částí bylo třeba nalézt
a ověřit metodiku řešení pevnostních zkoušek, návrh zavádění experimentálních zatížení, vlastní provedení zkoušek
a vyhodnocení výsledků. Metodika byla ověřena při statických zkouškách konstrukčních uzlů podvozkových nosníků
a trupové přepážky za standardních podmínek a po kondiciování a při statické zkoušce modelové sekce trupu malého
dopravního letounu. V tomto článku je popsána pouze metodika statické pevnostní zkoušky modelové sekce trupu.
Keywords: static strength test, primary composite structural part, model fuselage section.
1. Introduction
The static strength tests of large composite structural parts model fuselage section of a small transport aircraft, were performed in VZLU strength test lab in 2009. Although the development of the composite model fuselage section did not consider new full-scale aircraft and it wasn't under the CAA and
EASA supervision either, CS-23 airworthiness certification
requirements were met during the tests. According to mentioned certification requirements and generally respected Building Block Approach, the test program for composite model
fuselage section strength verification was proposed. It has consisted of test specimens, details - panels, subcomponents and
full-scale components [2]. The mentioned group of tests have
respected Damage Tolerance certification requirements including environment influence. Tests with VID and BVID damages and with environment influence were performed on panels and landing gear beam. Only philosophy of static strength
test of model fuselage section is described in this article.
2. Test specimen
The fuselage section shell (Fig. 1) had been assembled from
right and left halves which were connected by bonded join
with inner and outer straps. The opening holes for passenger’s
windows were in both shells. The shells were reinforced by
two integrated bulkheads made of sandwich structure with
foam core. The front and rear wing hinges were fixed on the
top side of bulkheads by bonded and bolted joints. In the bottom part of both bulkheads were places for front and rear landing gear beams installation. The beams were fastened on
fuselage structure by two pins. The forces from landing gear
beams were distributed to fuselage structure by bolted and
bonded flanges. The floor panels were bolted to floor spars
which were pasted into shell bottom part. Test specimen
weight was 220 kg (without the mass of all mock-ups). Dimensions of the fuselage section were 3760x1750x1640mm.
3. Load cases selection
The load cases analysis was performed with reference to significant loaded components on fuselage, landing gear beams
and wing hinges. According to this analysis, load cases, for
which static test were performed, were specified. Load cases
were chosen so that their minimum number sufficiently
covered the loads of fuselage section. The forces and their
points of action were defined so that histories of the component’s required loading were applied on the section tested.
During the load cases selection the preference was given to
the cases with maximum values of vertical forces and bending moment in vertical plane, because these components
are the most important. The following load cases were performed [1]:
1. Load case No: 505, gust acting on the vertical tail,
2. Load case No: 82, one main gear wheel landing,
3. Load case No: 42, landing with maximum angle of attack,
4. Load case No: 74, main gear horizontal landing.
4. Load distribution
Fig. 1 — Fuselage section set-up
The test specimen was fastened to test stand by fitting jigs (see
Fig. 2).
33
LETECKÝ
Fig. 2 — Test
stand set-up
Z P R AV O D A J
3/2010
damper crumpling setup. The specimen deformations were
made possible by rod-end bearing in load applying points.
Landing gear mock-up mass was 175 kg. The mass of both one
were balanced. The scheme of the loading forces on fuselage
section is shown in Fig. 4.
Fig. 4 - Load forces scheme - Load case 505
5. Test realization
The loads were applied to the fuselage section by sleeves, front
and rear load frames and left / right landing gear mock-ups.
The sleeves were divided into horizontal level, consisting of
two similar structural halves connected by bolted joints. Their
structure was composed from steel profiles bolted to wooden
boards with thickness 60 (80) mm, in which the fuselage contour was caught with overlap of 5 mm for rubber band sealing.
The gaps between sleeves and fuselage contour were tightened
by two component sealant during the sleeves assembly to fuselage. After the first test series there were failures of fuselage
skin below connections of both sleeves halves. Therefore the
connections were repaired by bolted patches. Weights of sleeves were 445 / 245 kg. The actual fuselage and sleeves mass
were balanced by mutual balancing system. The front and rear
frames, for load application of further sections of rear whole
fuselage, consisted of sheet-metal flange with a thickness of
8mm, to which the tubes and profiles were welded. This parts
supported the load actuators fastening points.
The frames were bolted to the flanges of fuselage section by
40 fitted bolts. The flange section dimensions were 4 x 40 mm
approx. Drilling of frames (11H7 holes), that were fixed to the
floor, were performed in vertical fuselage position. Joint fitments were performed by epoxy sealant. The frames mass
were 440 and 460 kg. These were balanced by front and rear
balancing system. The concept of applying a load to the front
fuselage part is in Fig. 3.
The test was performed in VZLU Strength Test Lab according
to specification. Fuselage section (the test specimen) was checked after delivery to test facility. NDI by ultrasound method to
find the defects and failures was also performed before and
after the test. The results were recorded into specimen documentation. The strain gauges were installed on the fuselage
section in specified places. The places for displacements measuring were defined too. The calibrations of actuator displacement sensors were performed using IST Hydropuls control
load system. Load cells, including cables, were calibrated in
TIRAtest loading machine. The tuning of loading channel’s
loops regulation were performed after loading system assembly. Load and position’s limits on individual channels were preset with ”Group Unload“ action for limits exceeding. The ”Dither“ values 0.5% and 400MHz were set up for all channels.
The reason of limitation start load reaction in actuators during
the switch-on of pressure in hydraulic distribution. All actuators were activated during the function test and connected to
test specimen. The actuator limits were adjusted and activated
so as to make possible safe test shutdown during the random
channel overloading. The actual regulation constants were set
up for position and load control of all channels in program
”Loop Tuning Tools“ during the preloading activity on test
specimen. These constants are different for each test and
actuator’s assignment. They depend on specimen rigidity and
loading system behaviour. Test set-up is shown in Fig. 5.
Fig. 5 — Test set up - Load case 505
Fig. 3 — The load distribution to front part of fuselage section
The landing gear forces were applied to the specimen by use of
main landing gear mock-ups which have possibility of various
Partial static strength tests were performed by following procedure:
C Z E C H A E R O S PA C E P R O C E E D I N G S
34
● Preloading, check of loading and measuring systems, dis-
MDL Fuselage section, Load case 74, Moment of failure-2,
100709
60
50
40
30
20
Load (kN)
placements and strain gauges measuring,
● Preloading and static test No. 1 up to limit load (LL), displacements and strain gauges measuring, permanent
deformations evaluation,
● Preloading and static test No. 2 up to limit load (LL), displacements and strain gauges measuring,
● Preloading and static test up to ultimate load (UL), displacements and strain gauges measuring, failures evaluation,
● Static test up to failure, strain gauges measuring, failures
evaluation.
10
0
-10894
896
898
900
902
904
906
908
910
912
-20
-30
-40
-50
-60
-70
-80
Time (s)
6. Test results
During the test of load case No. 82 at the 50% UL, skin fuselage failure occurred. The structure was reinforced in damaged
location (outer and inner fuselage skin) by bonded patches.
During the test load case No. 74 at the 59% UL, fuselage
structure was damaged in presence of intense crackling noise.
Process of static test is in Fig. 6. Registration of the failure
moment is in Fig. 7.
Card 1 - Load
Card 4 - Load
Card 10 - Load
Card 7 - Load
Card 8 - Load
Card 9 - Load
Fig. 7 — Registration of failure moment
Loading forces - load case 74
MDL Fuselage section, Load case 74, Up to 100% UL-2,
100709
60
50
40
30
Fig. 8 — Fuselage section failure - left rear part
20
Load (kN)
10
0
200
300
400
500
600
700
800
900
1000
-10
-20
-30
-40
-50
-60
-70
depends on the kind of test specimen and loading process. In
following Fig. 9 is shown a change of displacements on higher
loading level for load case No. 74, test to failure, loading level
55%UL. The most extensive viscoelastic creep effect was
recorded at the place of maximum loading — rear fuselage
section part (actuator 5B). The displacement is approximately
0.6mm in place 5B.
The viscoelastic creep had no affect on actual setting of
actuator regulation constants. The viscoelastic creep was
irrelevant at standard loading rate of fuselage section
(5%UL/10sec).
-80
Load case 74, actuator 5B, before failure
Time (s)
Card 4 - Load
Card 10 - Load
Card 7 - Load
Card 8 - Load
Card 9 - Load
Fig. 6 — Process of static test
Loading forces - load case 74
In subsequent visual inspection the numbers of failures were
located (bulkhead cracks, skin crimps and delaminations).
7. The viscoelastic creep
During the composite fuselage section loading a viscoelastic
creep was found at higher loading levels with time delay (approx. 30sec) for strain gauge and displacements measurement.
The viscoelastic creep means farther deformations of composite specimen during the constant load. Deformation speed
86,0
position (mm)
Card 1 - Load
85,5
85,0
84,5
84,0
83,5
83,0
840
850
860
870
880
890
Time (s)
T2 - Card 10
Fig. 9 — Viscoelastic creep of fuselage section
in place 5B
900
35
LETECKÝ
8. Conclusion
Z P R AV O D A J
3/2010
References
The static strength of large composite structure (composite
model fuselage section) was verified according to mentioned
philosophy. In this way the corresponding part of CS/FAR 23
including Damage Tolerance requirements were met. The fatigue strength test with fail-safe pass and residual static strength
test will be the next step of the model fuselage section certification program.
[1]
[2]
[3]
[4]
Snop, V.: Zkušební program statické pevnostní zkoušky
kompozitní sekce trupu MDL; VZLU Report R-4340, 2008
Cabrnoch, B. et al: Projektová studie celokompozitního trupu
MDL; LA composite Report LA033/MPO/06, 2006
Jironč, J.: Návrh zkoušek a zkušebních zařízení kompozitních
konstrukčních uzlů; VZLU Report R-3990, 2006
Fárek, J.: Výběr případů zatížení pro zkoušku celokompozitového
trupu; Evektor EV-55539-01-ST, 2008
Composite Material Machining and
Manufacturing of Molds Using
the CNC Machine Tool Unit
Obrábění kompozitů a výroba forem na CNC obráběcím centru
Ing. Radek Mazal / HPH s.r.o., Kutná Hora
Společnost HPH s.r.o., český výrobce kluzáků, uvedla v roce 2009 do provozu výkonnou obráběcí číslicovou jednotku
vhodnou pro letecký průmysl, speciálně pro výrobu rozměrných dílů kluzáků. Kromě obrábění kompozitů umožňuje
i frézování dřeva, překližek, plastů, pryskyřic, hliníkových slitin a dalších materiálů. Uvedeny příklady obrábění
trupu, křídla a dalších rozměrných polotovarů.
Last year, the company of HPH s.r.o., the Czech manufacturer of gliders, launched its own powerful CNC Machine
Tool Unit. The POWER FC 9000 CNC Machine Tool Unit was developed as a result of the three-year cooperation of
HPH s.r.o. and the company of SAHOS s.r.o., under the financial support of the Ministry of Industry and Trade
of the Czech Republic.
That way SAHOS s.r.o. extended its line of products by the
new series of CNC Machine Tool Unit and HPH s.r.o. acquired a powerful tool to be used in the aviation industry. The
design of the unit basic dimensions was based on the sizes
of the individual components of the 304S Glider and of the
other types of gliders under development. The unit features
external dimensions of 11.1 x 5.5 x 6.0 meters and the
workspace of 9.0 x 3.0 x 2.0 meters. The entire unit is large
enough to accommodate even the biggest components like
a wing or fuselage of the glider of 18 meters in wingspan
and the machine is capable of machining in five axes, therefore from the bottom as well.
In addition to the machining of composite materials, the
CNC Machine Tool Unit is capable of milling the solid
materials like artificial wood, massive wood, plywood,
wood-base fiberboards, wood-base particleboards, as well
as the much more consistent materials like aluminum alloys. Moreover, using the special accessories the unit is capable of machining the foamed polystyrene, plastics, epoxy
resins and polyurethane resins.
POWER FC 9000 CNC
The frame of the CNC Machine Tool Unit is welded from
the massive thick-walled steel profiles and the sheetmetals, thermally processed, and filled with the special
high-consistency concrete, all with a view to achieve the
necessary shape resistance and solidity of the machine.
The performing parts of the machine were designed
with respect to its application in machining the semifinished materials made of composites, and already during
the development of the machine, the resistance of the
selected components was tested for the carbon and aramid dust, which extremely wears down the machine
during the processing of these materials.
The shifting parts are embedded with the help of ball
linear guides equipped with the wiping units. The ball
screws or pinion gears driven by the digital alternating
servomechanisms transform all movements to the linear
shifting.
The machine is controlled by the Heidenhein iTNC 530
fully digital control system (optionally by Siemens Sinumerik 840 or Fidia NC15).
The machine features a tool changer for the fully automated operation.
Left: POWER FC 9000 CNC Machining Unit
C Z E C H A E R O S PA C E P R O C E E D I N G S
POWER FC 9000 CNC Machine Tool Unit Technical Specifications:
36
Utilization of the following was determined and tested for
the purposes of aviation industry:
Cutting of the composite moldings after their removal from
the mold — in particular, this is the actual shaping performed
by the milling, cutting and drilling tasks. Because of a spatial
character of the peripheral contours, it is necessary to master
the technology of machining in 5 axes.
Component Chucking Methods
Correct clamping of the product provides for the accuracy of
machining process, as well as the surface quality of the
machined part. Therefore, the suitable machine tool bench
had to be selected, as well as the basic connection surface for
all types of the machined parts, which would correspond to
the machined material, also with respect to the subsequent
processing. We also contemplated using the chucking method
via the grid of the threaded cases embedded into the board
made of synthetic resin bonded paper, but because of the high
variety of the semi-finished products, we chose the simple
and effective method of direct clamping using bolts screwed
into the base plate made of MDF. The MDF material is sufficiently rigid and simultaneously does not damage the tool in
the case of potential failure.
Utilization of the CNC Machining Unit in Aviation
Manufacturing
HPH s.r.o. is very particular about the quality of processing
the individual parts its customers are used to, and it has to be
achieved by a thorough manual finishing of all components,
which is time-consuming and requires the highly qualified
workforce. In addition, even the most careful manual work
will never fully assure the equality and exchangeability of all
parts. This fact negatively surfaces whenever a necessity arises to mutually fit the individual components during the
assembly or servicing.
With respect to the ever-increasing pressure on making the
production more efficient, we had to develop the technology of
machining the composite parts. Utilization of CNC technology
is particularly advantageous when deployed in the following
two major manufacturing operations:
1. design and manufacturing of the models and molds
2. cutting and finishing of the laminated composite semifinished products
Each of these operations has its own special characteristics.
During the research for the suitable technological conditions,
we used our experiences with CNC technology in finishing the
wood and the wood-fiber materials. Characteristics of these
materials are basically similar to the composites (fiber structure + bonding agent, combustibility).
Wingtip after demoulding
Prototype Models and Molds
In manufacturing composite preproduction models and
molds, emphasis is put on the price and speed of the production. The materials meeting such requirements are the
polyurethane foams of density from 120 to 450 kg/m3,
which allow molding up to 5 pieces of the pilot series. In
addition, the 80 kg/m3 polyurethane foam can be used for
simple shapes and for final shaping of up to 1 to 2 pieces.
However, all mentioned materials require the surface stabilization after their machining in order to utilize the standard
follow-up technologies for the preparation of the model or
the form for the composite part.
Serial Models and Molds
In manufacturing composite serial production models and
molds, emphasis is put on the quality of the surface and
on the mechanical characteristics under the repeated molding of the products. The materials meeting such requirements are the polyurethane foams of density from 600 up
to 800 kg/m 3, which allow molding up to 70 pieces, and
when using the block materials with special filling of 700
to 1,700 kg/m 3 in density, it is possible to mold the thousands of pieces. The surfaces of the mentioned materials
are easy to finish, and the special fillers allow their easy
polishing.
37
LETECKÝ
Z P R AV O D A J
3/2010
Examples of Suitable Machining Operations
1. Fuselage Machining
The size of the main structural components of gliders is usually a matter of meters. Such large machined pieces cannot be
clamped onto an integral device; they must be attached to the
bench using the chucking system. Mounting of the machined
piece must be done with a help of the measuring points determined in advance, which can be ”detected“ in the machining
space using the probe attached to the spindle.
The holes in the fuselage represent a special issue for the
machining process, especially the openings for the driving
units. These holes are usually marked with a laminated contour, so they can be easily detected. Nevertheless, they are present at the place of a connecting hem for both halves of the
fuselage. At this place the laminate is up to 10 millimeters
thick, so the machining task represents more of a milling into
a solid mass than the cutting of the thin-walled plate.
Carbon fiber fuselage
2. Wing Machining
In machining the glider wing and tail surfaces, the contours at
the dividing plane are especially suitable, where the burrs of
the overflowed material remain after the molding process.
Other surfaces suitable for machining are the contours of the
wing bracing struts and the elevator bracing struts. In addition,
grooves for the sealing tape are made in the coat alongside the
bracing struts.
The holes in the wings represent a special issue for the
machining process, especially the two openings for the installation of the water valves situated at the bottom of the wing,
the two openings for the filling caps situated at the top coat of
the wing, and the shaft openings for the braking flaps, also
situated at the top coat of the wing.
Glass fiber fuselage,
technological sample
Trailing edge machining
The following operations were selected for comparison
purposes:
- cut of the contour at the dividing plane
- machining of the contour of the wing bracing struts
- machining of the contour of the elevator bracing struts
- milling of the holes into the wing coat
Comparison of Labor Consumption: Manual Machining
versus CNC Machining (complete machining of a single
wing):
C Z E C H A E R O S PA C E P R O C E E D I N G S
3. Cantilever Machining
Composite plates of 220 x 150 x 25 millimeters in size were
used as semi-finished products. This product is suitable for
machining, especially because of the simple task preparation.
The clamping of the machined piece was made with a help of
technological openings, it was adequate, and it provided the
necessary rigidity of the ”machine-tool-piece“ set. The savings
of labor in the case of these plates was tremendous because of
the technological demand on their manual production.
Machining of wing cantilever beam semiproduct
Cantilever after machining
38
ved the comparison of the times necessary for the preparation works and for the actual production of the component.
The time saved in the process of actual machining varies
around 70%. Nevertheless, the timesaving is much lesser in
the process of work preparation, and in the case of fuselage machining, it even showed the negative figures. The
CNC machining preparation works particularly involve the
purchase of the computer three-dimensional model of the
component, CAM programming of the machining process,
selection of the optimal chucking method, and also the
”rough“ machining of the testing sample. Most of these
operations are based on data preparation, whereat the work
with the serial-production components is more beneficial
because such data have to be prepared only once and they
can be reused. The perceived disadvantage of the prolonged data preparation for the piece production (e.g. the prototype design & development) is compensated more than
enough by the significantly shorter production time and
also by the quality of the final product. In addition, such
applies twofold for the components that are complex in
shape, where not only their shape must be maintained, but
also their symmetry.
HPH s.r.o. developed the CNC Machine Tool Unit and
the follow-up technology of composite material machining
thanks to the program named IMPULS and thanks to the
financial support of the Ministry of Industry and Trade of
the Czech Republic. A similar machine has never been available to the Czech aviation industry until now. Most of the
work in the past had to be done manually or using the conventional methods, or CNC equipment had to be purchased
for very high prices. The deployment of CNC technology
in HPH s.r.o. proved to be practical and only the amount of
own work and the external job orders will show the return
on investment in the future. HPH s.r.o. would like to offer
its free machine production capacity to the other members
of the Asociace výrobců sportovních letadel (Association of
Sport Aircraft Manufacturers), with a view to benefit the
entire Czech aviation industry.
References
[1]
[2]
[3]
[4]
Utilization of the CNC Machine Tool Unit
The above-mentioned examples show the overall decrease
in labor demand for the production of testing samples. The
timesaving rates within the individual phases of the production are however very different. Our assessment invol-
Szabo, František, Ing.: Obrábění plastů na CNC obráběcím
centru; MM Průmyslové spektrum, MM 2010/1, Article
code: 100104, p. 38, 3. února 2010
Pekař, Martin, Ing.: Vývojová zkouška obrábění - křídla
a VOP, T9-2009-E9, Zpráva k projektu FI-IM4/168, HPH
s.r.o., 24.05.2009
Pekař, Martin, Ing.: Vývojová zkouška obrábění - ořez
trupu, T7-2009-E9, Zpráva k projektu FI-IM4/168, HPH
s.r.o., 24.05.2009
Pekař, Martin, Ing.: Vývojová zkouška obrábění
kompozitové desky tloušťky 25mm, T2-2009-E9, Zpráva
k projektu FI-IM4/168, HPH s.r.o., 28.01.2009
39
LETECKÝ
Z P R AV O D A J
2/2010
Aerodynamic Optimization Procedure
of Flapped Sailplane Airfoil
Postup aerodynamické optimalizace profilu větroně s klapkou
RNDr. Jaroslav Hájek, PhD., Doc. Ing. Zdeněk Pátek, CSc., M.S., Mgr. András
Szöllös, Ing. Marian Zabloudil / VZLÚ, Plc., Prague
Aerodynamic optimization of the sailplane airfoil is a task of high complexity caused by the extreme variety of sailplane flight regimes. As a consequence, an optimization of the sailplane airfoil with a trailing edge flap is a highly multidisciplinary task, and microgenetic algorithm technique proved as advantageous. Development of resulting optimized
airfoils in dependence on the development of optimization criteria and constraints was tracked. The final airfoils featured better drag characteristics than an initial well-performing airfoil already especially designed for high-performance
sailplanes.
Aerodynamická optimalizace profilu větroně představuje kvůli rozmanitosti letových režimů obtížnou úlohu. Proto
optimalizace profilu větroně s klapkou na odtokové hraně patří k rozsáhlým multidisciplinárním úlohám a byl na ni
úspěšně použit mikrogenetický algoritmus. Sledoval se vývoj optimalizovaných profilů v závislosti na vývoji optimalizačních kritérií a omezení. Výsledný profil vykázal lepší odporové charakteristiky než původní velmi výkonný profil
rovněž navržený speciálně pro výkonné větroně.
Nomenclature
cD
cL
cm
cm
Re
α
δf
drag coefficient
lift coefficient
pitching moment coefficient
absolute value of the pitching moment coefficient
Reynolds number
angle of attack
flap deflection
1 Introduction
The sailplane is a very specific aircraft from the point of view
of aerodynamic optimization. Complexity of optimization originates from the fact that there are significantly different regimes during one flight [1], [2]. A typical cross country flight
consists of circling at thermals at low speed and of straight
interthermal glide at higher speed as well. The difficulties are
amplified by the uniqueness of each flight caused by the variability of weather dependent on location and time, especially
variability in physical characteristics and distribution of thermals. The weather conditions which bring largely non-constant
ratios between circling and straight glide parts during different
flights but they also bring even different flight speeds in different circling segments and different speeds in different straight
glides during one flight. In addition, it is necessary to envisage
the possibility of outlanding which imposes subsequent constraint on low landing speed. Solution (for sailplane classes
where it is allowed) was found in trailing edge flap settable in
several deflections up and down and thus adapting the airfoil
to different conditions by modifying airfoil curvature [3].
2 Optimization technique
The optimization demand arises from the large principal differences between the flight regimes which necessarily lead to the
optimization with controlled and evaluated compromises. Optimization of flapped sailplane airfoil has to consider several
typical flight regimes with different flap deflections and Reynolds numbers but also to consider other requirements imposed by the possibility of outlanding and by the possibilities of
manufacturing technologies.
As the problem is evidently multicriterial, a genetic algorithm approach was evaluated as promising optimization technique. µARMOGA microgenetic algorithm developed by
VZLU was used. The functioning of µARMOGA can be described as follows: After initialization of the population by
Latin hypercube sampling (LHS) and evaluation depicted as
archive update, the evolution goes through selection, mating
and mutation to evaluation of the new population. Every
n-generation the population statistics is updated, range-adaptation takes place, followed by knowledge based (elitist-random)
reinitialization. Thorough description of the algorithm is to be
in [4][5].
3 Aerodynamic Characteristics Computation
Method
The XFLR5 v5.00 software was used for the computations.
The en criterion with n = 9 was used for the transition from
laminar to turbulent boundary layer.
4 Optimization
The aerodynamic optimization of high-performance-sailplane
airfoil is generally driven by the aim to minimize drag. This
goal was expressed as the minimization of the area in polar
curve diagram determined
- from the left side, by the vertical axis at cD = 0,
- from the right side, by the envelope (from the left side) of
polar curves for individual flap settings at Reynolds numbers corresponding to the relevant flight regimes (see Tab.
1),
- from the bottom side, by the value of cL = 0.10,
- from the upper side, by the value of cL = 1.30.
Simultaneously it was required to achieve not inferior maximum lift performance (to meet outlanding requirements,
cL MAX ≥ 1.60 at δf = 14 deg at Re = 1.217·106) and not inferior
moment characteristics (to meet control requirements — it was
supposed initially that the moment did not decline from the
moment of initial airfoil so there was no specific constraint
imposed).
Due to supposed manufacturing technology limitations, the
C Z E C H A E R O S PA C E P R O C E E D I N G S
Fig. 1a — Initial airfoil's lift curves
40
Fig. 1b — Initial airfoil's polar curves and envelope polar curve
airfoil maximum thickness was constrained to 13.5 percent as
minimum reasonable value, the flap chord was fixed as 15 percent of the airfoil chord, the thickness in the place of the flap
axis of rotation at 3.75 percent at least and the thickness at the
trailing edge at 0.4 percent.
An earlier developed high-performance sailplane airfoil
designated AF135-I (see Fig. 1) was used as an initial airfoil
for the genetic manipulations.
Fig. 2 — AF135-1A shape
Tab. 1 — Airfoil's flight regimes
5 Results
5.1 Optimized Airfoils without Moment Constraint
The initial optimization runs generated new airfoils, more
effective from the point of view of drag. But the analyses showed that this improvement is at the expense of the higher absolute value of the moment and the overall aerodynamic performance was compromised by this fact. Example see the AF1351A airfoil (Figs. 2 and 3).
5.2 Airfoils with Moment Constraint
Thus it was necessary to complete the optimization by addition
of another constraint limiting the moment. The new constraint
was set as follows:
Airfoil at flap deflection of 0 deg, at Re = 1.217·106,
cm≤ cm initial airfoil at cL = 1.10.
Two airfoils designated AF135-2A and AF135-2B respectively were evaluated as the most promising new designs
(Fig. 4).
In comparison with the initial AF135-I airfoil section, the
AF135-2A is featured by improved drag at lower and medium lift coefficients, and extremely high maximum lift coefficient of cL MAX = 1.90 (at flap deflection of 14 deg), higher
by 0.11 compared to the cL MAX of AF135-I. The absolute
value of the moment is decreased (Fig. 5).
The AF135-2B provided lower drag coefficient for nearly whole range of lift coefficients, with the exception of
41
LETECKÝ
Z P R AV O D A J
Fig. 3a — AF135-1A lift curve
Fig. 3b — AF135-1A envelope polar curve
Fig. 3c — AF135-1A moment curve
Fig. 3d — AF135-1A moment vs. lift curve
Fig. 4a — AF135-2A shape
Fig. 4b — AF135-2B shape
2/2010
C Z E C H A E R O S PA C E P R O C E E D I N G S
42
Fig. 5a — AF135-2A and AF135-2B lift curves
Fig. 5b — AF135-2A and AF135-2B envelope polar curves
Fig. 5c — AF135-2A and AF135-2B moment curves
Fig. 5d — AF135-2A and AF135-2B moment vs. lift curves
cL ≥ 1.48, and the maximum lift coefficient virtually the
same compared to AF135-I (in fact, higher by 0.02 which
was not of any significance regarding the c L MAX reliability
in XFOIL computations). The moment is also better than at
AF135-I but the improvement is less pronounced than at
AF135-2A (Fig. 5).
Unfortunately, unfavourable deficiency was revealed on
both of airfoils: the ”hump“ at the lift curve, i.e. the slight
decrease of the cL followed by its re-increase. This phenomenon occurred in the zone of the flap deflections δf ≥ 5
deg at the angle of attack of 4 degrees, so at the configuration and the angle of attack currently used in the flight.
The analysis showed that the beginning of the ”hump“
Fig. 6 — AF135-3A shape
43
LETECKÝ
Z P R AV O D A J
Fig. 7a — AF135-3A lift curve
Fig. 7b — AF135-3A envelope polar curve
Fig. 7c — AF135-3A moment curve
Fig. 7d — AF135-3A moment vs. lift curve
indicated the formation of laminar bubble on the upper surface. The magnitude of the ”hump“ itself was done mainly
by the drop of the suction at rear part of the upper surface
of the airfoil, significantly pronounced with growing angle
of attack. Once the reason of the phenomenon was discovered, the remedy was found in the further completion of
optimization constraints.
5.3 Final Optimization of 13.5 percent Airfoils
Following the analysis of the aerodynamic performance of
the airfoils developed so far, the optimization procedure
was finally completed in two ways.
a) The optimization was completed by constraint limiting the minimum value of lift curve slope to exclu-
2/2010
de the ”hump“ on the lift curve even for high flap
deflections.
Airfoil at flap deflection of 14 deg, at Re = 1.217·106, lift
curve slope dcL/dα (in deg) ≥ 0,02 at 0 ≤ α ≤ 10 deg.
b) At low lift coefficients (cL ≤ 0.3), at the flap deflections -10 deg and -14 deg, there was a very narrow
(in terms of cL) laminar drag bucket which caused
very narrow flight speed range in which the flap
should be set from -10 to -14 deg to obtain optimum
drag (i.e. to move on the envelope drag curve
during high-speed flight by optimum manner). This
feature would be inconvenient for the pilot a thus
could compromise sailplane performance.
C Z E C H A E R O S PA C E P R O C E E D I N G S
44
To assure wider laminar buckets at all flap positions, the
optimization criteria were once more redefined as the minimization of the sum of three areas defined as follows:
1. For the airfoil with flap deflection of -10 deg at
Re = 2.434·106, the area in the polar curve diagram
bounded by cD = 0 axis and polar curve from the left
and the right side respectively, and cL values between
0.15 and 0.50 vertically.
2. For the airfoil with flap deflection of 0 at
Re = 1.521·106, the area in the polar curve diagram
bounded by cD = 0 axis and polar curve from the left
and the right side respectively, and cL values between
0.40 and 1.00 vertically.
3. For the airfoil with flap deflection of +10 deg at
Re = 1.217·106, the area in the polar curve diagram
bounded by cD = 0 axis and polar curve from the left
and the right side respectively, and cL values between
0.80 and 1.30 vertically.
The resulting airfoil was designated AF135-3A (Figs. 6 and 7).
References
[1]
Thomas, F.: Fundamentals of Sailplane Design; College Park
Press, College Park 1999, ISBN 0-9669553-0-7
[2]
Pajno, V.: Sailplane Design; Macchione Editore, Varese 2006,
ISBN 88-8340-274-X
[3]
FAI Sporting Code Section 3 — Gliding; Fédération
Aéronautique Internationale, Lausanne 2009
[4]
Szőllős A, Šmíd M, Hájek, J.: Aerodynamic Optimization via
Multi-objective Microgenetic Algorithm with Range
Adaptation, Knowledge-based Reinitialization, Crowding and
Epsilon-dominance; Advanced Engineering Software 200, Vol.
40, pp. 419-430
[5]
Hájek, J, Szőllős A, Šístek, J.: A New Mechanism for
Maintaining Diversity of Pareto Archive in Multi-objective
Optimization; Advanced Engineering Software 2010, accepted
for publication
6 Conclusions
The optimization procedure using the microgenetic algorithm
proved to be an efficient method for complex multicriterial
aerodynamic optimization. The criteria and constraints
should be thoroughly defined to keep the genetic reproduction in desired way.
The newly developed airfoil features overall better aerodynamic performance than the initial airfoil. The initial airfoil was already of high performance, so the improvements are
in the order of a few percent but such improvements are valuable in high-performance sailplane design.
Room for Your Notes
Fatigue Prediction Based on Finite Element Analysis
of Riveted Joints
Colour illustrations to the article published on pages 27 to 31.
Fig. 5a
Fig. 5b
Fig. 5 — Examples of outputs of fatigue calculations with evaluating the SWT criterion with the sign derived from the sign
of the maximum absolute principal stress. Read the next section for an explanation of the scaled variable - the sixth root
of damage.
Fig. 8a
Fig. 7 — Numbering of holes. The picture shows the top (T)
sheet above all
Fig. 8 — Calculation results presented in sixth root of damage
for EVE1 specimen, SWT method without any signing and four
levels of extern loading - a) LR1, b) LR2, c) LR3, d) and e) LR4
Fig. 8b
Fig. 8c
Fig. 8d
Fig. 8e
Hardware in the Loop Simulation of FBW Components
Colour illustrations to the article published on pages 20 to 26.
Fig. 9 — Aircraft hydraulic stand
Fig. 10 (Right) — Installation
of Elevator actuator
on the aircraft stand
Fig. 11 (Below) — Installation of the aileron EHSA
on the aircraft stand
Fig. 12 (Right) — Experimental
cockpit of FBW system without
mechanical redundancy
© ASSOCIATION OF THE
AV I A T I O N M A N U FA C T U R E R S