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