sprint abilizer
Transkript
sprint abilizer
April 2011 ISSN 1211—877X CZECH AEROSPACE Proceedings LETECKÝ zpravodaj In this issue: J OURN A L FOR C Z ECH A EROS P A CE RESE A RCH Feasibility tests of friction welding of high melting temperature materials in VZLÚ Simulation of level of quasistatic microaccelerations on-board of a spacecraft Small aircraft ECS optimization using 1D tool Microaccelerometer Error Analysis Analysis of dynamic characteristics of gas turbine Kinetic analysis of oxidizing decomposition of carbon fiber reinforced epoxy composite Analysis of options for calibration of micro accelerometer MAC No. 1 / 2011 CZECH CZECH AEROSPACE Proceedings J OURN A L Editorial address: FOR C Z ECH A EROS P A CE RESE A RCH VZLÚ / Aeronautical Research and Test Institute Beranových 130, 199 05 Prague - Letňany Czech Republic Phone.: +420-225 115 224, Fax: +420-286 920 518 Editor-in-Chief & DTP: Martina Monteforte Hrabětová ([email protected]) Production department: Zdeněk Kočí ([email protected]) Distribution: Zuzana Tyrová ([email protected]) 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 - Letňany Czech Republic ID. No. 65991303 in cooperation with the Czech Technological Platform for the Aviation and Space. Printing: Aeronautical Research and Test Institute 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 1/2011 © 2011 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 L e t e c k ý z p r avo da j 1/2011 Contents / Obsah 2 Feasibility tests of friction welding of high melting temperature materials in VZLÚ, Plc. Zkoušky proveditelnosti třecího svařování materiálů s vysokou teplotou tání realizované ve VZLÚ, a.s. Ing. Petr Bělský / VZLÚ, Plc., Prague 7 Experimental modal and FEM analysis of space instrument DUMMY Experimentální modální a MKP analýza kosmického přístroje DUMMY Ing. Vladimír Dániel, Ph.D., Ing. Tomáš Jamróz, Ing. Jiří Had /VZLÚ, Plc., Prague 10 Simulation of level of quasistatic microaccelerations on-board of a spacecraft Modelování úrovně mikrozrychlení na palubě kosmického prostředku Ing Viktor Fedosov Ph.D. /VZLÚ, Plc., Prague 16 The influence of structural damping on flutter characteristics of a small sport aircraft Vliv strukturálního tlumení na flutterové charakteristiky malého sportovního letounu Ing. Martin Zejda / Aerospace Research Centre, FME CTU in Prague; Ing. Aleš Kratochvíl / Aerospace Research Centre, FME CTU in Prague; Doc. Ing. Svatomír Slavík, CSc. / Czech Technical University in Prague, Faculty of Mechanical Engineering, Department of Aerospace Engineering 20 Small aircraft ECS optimization using 1D tool Optimalizace ECS malých dopravních letadel pomocí 1D simulačního nástroje Ing. Jiří Hejčík, Ph.D. / Institute of Aerospace Engineering, Brno University of Technology Prof. Ing. Miroslav Jícha, CSc. / Department of Thermodynamics and Environmental Engineering, Brno University of Technology 23 Development of series production technology for stator and rotor blades Vývoj technologie výroby statoru a rotoru pro opakovanou výrobu. Ing. Karel Barák; Ing. Michal Řehák/ Department of Aerospace Engineering 25 Microaccelerometer Error Analysis Rozbor chyb mikroakcelerometru Ing. Milan Chvojka 31 Analysis of dynamic characteristics of gas turbine Analýza dynamických vlastností spalovací turbíny Tomáš Jamróz, Jiří Had, Vladimír Dániel / VZLÚ, Plc., Praha 34 Kinetic analysis of oxidizing decomposition of carbon fiber reinforced epoxy composite Kinetická analýza oxidačního rozkladu uhlíkového epoxidového kompozitu Ing. Zdeněk Mašek 40 Analysis of options for calibration of micro accelerometer MAC Analýza možností kalibrace mikroakcelerometru MAC RNDr. Vojtěch Zadražil / VZLÚ, Plc., Prague 47 Worst Case Analysis Methods Implementation in an Accelerometer Measuring Loop Analysis Aplikace postupů Analýzy nejhoršího případu (WCA) při analýze měřicí smyčky akcelerometru Jozef Zakucia, Milan Merkl / VZLÚ, Plc., Prague 54 Experimental Assessment of the Fragments Effect to the Fuselage of the Aircraft Experimentální určení účinku střepin na drak letounu Ing. Miroslav Lošťák, Doc. Ing. Miloslav Petrásek CSc. VUT Brno 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 Feasibility tests of friction welding of high melting temperature materials in VZLÚ, Plc. Zkoušky proveditelnosti třecího svařování materiálů s vysokou teplotou tání realizované ve VZLÚ, a.s. Ing. Petr Bělský / VZLÚ, Plc., Prague The paper provides information about first feasibility tests of FSW and FSSW of high melting temperature materials in the territory of Czech Republic by VZLÚ, Plc in November 2010. Experiments were focused on friction welding of microalloyed steel USIBOR 1500 P intended mainly for the use in the automobile industry. Tested subjects consisted of four types of materials for working parts of welding tools (pure tungsten, W25Re2HfC alloy, silicon nitride and PCBN). In addition, the influence of welding parameters on mechanical properties and macroscopic pattern of the weld joints had been studied. Příspěvek stručnou formou informuje o prvních zkouškách třecího svařování materiálů s vysokou teplotou tání realizovaných na území České republiky ve VZLÚ, a.s. v listopadu 2010. Uskutečněné experimenty byly zaměřeny na kontinuální a bodové třecí svařování mikrolegované oceli USIBOR 1500 P určené především pro použití v automobilovém průmyslu. Během těchto testů byly testovány 4 druhy materiálu pro pracovní části svařovacích nástrojů (čistý wolfram, slitina W25Re2HfC, nitridová keramika a kubický nitrid bóru). Zkoumán byl vliv svařovacích parametrů na mechanické vlastnosti a makrostrukturu svarových spojů. Keywords: Friction stir welding, USIBOR 1500, microalloyed steel, PCBN, silicon nitride, tungsten rhenium Introduction The solid state welding processes known as friction stir welding (FSW) and friction stir spot welding (FSSW) are relatively new when compared to fusion welding methods. Principle of FSW technology was invented and patented in 1991 by The Welding Institute (TWI) in Cambridge UK [1]. The process uses a non-consumable rotating tool to generate frictional heat in the work piece. The conventional friction stir tool consists of a pin and a shoulder. The pin is plunged into the material to be joined and stirs the material at the joint interface. The shoulder remains in contact with the top surface of the material to be joined and generates additional heat while constraining the plasticized material. The FSSW is a variant on the friction stir welding process that does not involve any lateral movement along the work piece. During FSSW process a rotating tool with probe is plunged into two overlapped blanks to be jointed. The rotating tool generates friction heat in the specimens. The heated and softened material close to the tool plastically deforms and a bond is made between the upper and lower sheet. FSW and FSSW offer many advantages over conventional fusion welding processes because of their low heat input and absence of melting and solidification process. It results in better material properties, fewer weld defects, low residual stresses and improved dimensional stability. In the past, commercial applications of FSW and FSSW have been limited to lower melting temperature materials, e.g. aluminium, zinc, lead etc. The joining methods have proven very beneficial in these applications, but these materials represent less than 10% of the welded products in the world. On the other hand, steels and other high melting temperature materials (HMTM) represent greater than 80% of the welded product worldwide. Thus, the economic impact of FSW and FSSW in HMTM could markedly dwarf that of lower melting temperature materials. For example automotive industry represents huge opportunity for the next dissemination of the promising methods. Recently, the increasing use of advanced high strength steels (AHSS) for occupant protection and potential weight savings in high productivity manufacture has encouraged the development and adoption of new joining technologies capable of meeting the quality, reliability and cost requirements in the sector. There is a possible question whether FSW and FSSW methods could be utilized for welding of the AHSS steels and fulfil automotive industry requirements. The automotive industry at the territory of the Czech Republic has a long tradition of engineering and production. At the present time czech automotive manufacturers produce over most 1 million vehicles per year. The motor industry covers roughly 20% of the total exports and about 21% of the Czech exports to EU. The success was not possible to achieve without continuous innovations and cooperation of the industry with technical universities and R&D institutes. One of the most extensive production cycles in automotive manufacturing is joining, particularly welding. That way any innovation in the area can have big impact on resulting price and competitiveness. FSW and FSSW are very promising joining methods that could to find utilization in czech automotive industry. VZLÚ, Plc. officially started FSW research activities in 2005 in frame of the project ARC-B2 (Aerospace Research Centre) supported by the Czech Ministry of Education, Youth and Sports. A lot of various studies focused on butt, 3 L e t e c k ý z p r avo da j 1/2011 cimens. The specimens for continuous FSW tests consisted of two plates 200x110 mm. Specimens for FSSW tests were realized by using two coupons 45x150 mm with a 45x50 mm overlap in case of lap-shear specimens and two coupon 50x150 mm with a 50x50 mm overlap in case of cross-tension specimens. All specimens were degreased prior to welding using a household cleaner and alcohol. FSW and FSSW Tools Friction stir welding of high melting temperature materials such as steels require a tool material that possesses a combination of properties including high strength, toughness, high-temperature thermal stability, hot hardness, high temperature oxidation resistance, and a decreased affinity to the weld materials. During the feasibility studies in VZLÚ, Plc. were tested 4 tool materials: Fig. 1 - Continuous FSW of USIBOR 1500 steel sheets lap, spot and T-welds of aluminium alloys and other low melting temperature materials were performed through the years 2005-2009. Research activities focused on friction stir welding of high melting temperature materials such as steel, titan and nickel base alloys are next very important milestone for understanding all possibilities of the modern joining technology. The paper gives short information about first feasibility tests of FSW and FSSW of advanced high strength steel realized at the territory of the Czech Republic by VZLÚ, Plc in November 2010. Table 1 – Chemical composition and basic mechanical properties of USIBOR 1500 steel Material The test material used in the feasibility tests were sheets with thickness 1,5 mm made of Al-Si pre-coated quenchenable micro-alloyed steel 22MnB5, well known with the commercial name of USIBOR 1500 P [2]. The steel is intended mainly for use in automobile structural and safety components. USIBOR 1500 P is designed to be heat treated and then quenched during the hot stamping process. The mechanical properties of the final part make significant weight savings possible (up to 50% compared to standard high yield strength steel). The very high yield strength of the steel after heat treatment and hot stamping make it suitable for anti-intrusion components (fender beams, door reinforcements, B-pillars, windscreen uprights, floor and roof reinforcements, etc.). The material strength is propagated by adding a small fraction of boron to the carbon, manganese and chromium composition. Therefore this steel grade is also called “boron steel”, in colloquial terms. The nominal chemical composition and basic mechanical properties are listed in Table 1. The sheets were delivered in soft-annealed condition from Benteler Automotive Rumburk s.r.o. and cut into spe- Fig.2 – FSW and FSSW tools Pure Tungsten Pure tungsten is a steel-gray to tin-white metal. Very pure tungsten can be cut with a hacksaw, forged, spun, drawn, and extruded. The impure metal is brittle and can be worked only with difficulty. Tungsten metal has excellent high temperature mechanical properties and the lowest expansion coefficient of all metals. Pure tungsten has the highest melting point (3422°C) of all metals, and at temperatures over 1650oC has the highest tensile strength. The metal oxidizes in air and must be protected at elevated temperatures. Pure tungsten semifinished products in form of rods are quite easily obtainable on the market. W24.5%Re2%HfC alloy Historically W25%Re alloy has been produced into wire for the thermocouple market. The use of pure tungsten in thermocouple applications, as the positive element, posed a problem when heating tungsten above its recrystallization temperature (about 1200°C). This causes the wire to become brittle when reintroduced to room temperature. The addition of rhenium increases the recrystallization temperature, the ductility, and the ultimate tensile strength of the material. Also the hafnium carbide addition is very C z e c h A e r o s pa c e P r o c e e d i n g s beneficial - it markedly increases strength of the alloy at high temperature levels above 1300°C. The excellent material produced by Rhenium Alloys, Inc has recrystallization temperature about 1900oC, elastic modulus 400 GPa and the ultimate tensile strength at room temperature greater than 1400 MPa. The major advantage of the alloy is the high strength at elevated temperatures (250 MPa at 1926°C) [3]. Silicon Nitride Silicon nitride is a ceramic-based material characterized by high hardness, high thermal conductivity and low thermal expansion coefficient that give better thermal shock resistance than other ceramic materials. Silicon nitride is mostly used in high-endurance and high-temperature applications, such as gas turbines, car engine parts, bearings and metalworking and cutting tools. A disadvantage may be the limited fracture toughness. The working parts of welding tools used for feasibility tests in VZLÚ, Plc were made of ceramic semifinished products from CERATIZIT Austria G.m.b.H (rods made of ceramic material SNC1). PCBN Polycrystalline Cubic Boron Nitride (PCBN) is a highly efficient cutting material from a polycrystalline substance of a cubic boron nitride grain. PCBN is produced by sintering at high temperature and high pressure. Since PCBN is a very hard, wear resistance material, it seems to be an optimal material for friction stir welding of steel. But premature cracking and high costs could be limiting factors for its application. PCBN welding tools used in the study were realized in frame of cooperation VZLÚ, Plc and Bonar, a.s. Two different engineering designs of the welding tools were used (see Fig.x). The first design utilized short bar working parts (parts made of tungsten and Si3N4) inserted into shank holder and secured with locking bolts. The second tool construction consisted of a working tip, a locking collar, and a shank. The collar coupled the torque from the shank to the working tip by means of hexagonal connection. Position of collar and working tip was secured with locking bolts. The second tool design was used for PCBN and WReHfC alloy tips. Bodies of the tool holders, locking collars and shanks were made of nickel alloy Hastelloy X. All working parts of the tools have standard concave shoulder with diameter of 13 mm and concavity angle 7°. Also pin geometry was the same for all the welding tools. It had shape of truncated cone. The cone pin had base diameter of 5 mm and top angle 40°. Pin length was 1.27 mm for FSW tools and 2.45 mm for FSSW tools. 4 3) combination of ceramic spray ToolProtect and separation pad made of Phlogopite Mica or paper Welding Procedures All the continuous FSW tests of USIBOR 1500 steel were carried out on the manual knee-type milling machine FGU40 with 11 kW spindle drive motor. This old milling machine is characterized by good stiffness and sufficient power (spindle speed range: 35-1800 rpm, max torque: 2000 Nm). At the beginning of each weld the rotation tool slowly plunged into the joint line between two plates 200x110 mm. After attainment of full plunge 0.4 mm tool dwelled at the place about 5 seconds. Then followed accelerating (40 mm/min to distance 10 mm) and transition to planned welding speed. The tool rotational speed and welding speed were varied, which were 710 and 900 rpm, and 80, 160 and 224 mm/min, respectively. The welding tools were tilted 2.5° from the plate normal direction and inert gas (Ar) shroud was utilized for shielding which prevented base metal surface oxidation. Next experiments were focused on friction spot welding of the same material. They were conducted with NC bed-type milling machine FSG80-A2 under displacement control. The tool rotational speed and tool plunge rate were varied, which were 1800, 2500 and 3200 rpm, and 0.7 and 2 mm/s, respectively. Plunge depth was held constant at 2.65 from the specimen surface. Note that this plunge depth corresponds to 87% of the combined thickness (3.04 mm) of the overlapping strips. Since the pin length was 2.45 mm, this means that the tool shoulder plunged approximately 0.2 mm into the specimen during welding. Quasi-dwell of 1.5 s was applied at the end of each weld program. The dwell was accomplished by creating a two-step welding program that involved first plunging to nearly the full desired depth followed by further plunging the final 0.3 mm of depth at a slower rate 0.2 mm/s. Total time of welding cycle for plunge rates 0.7 and 2 mm/s was c. 5 and 3 s respectively. Heat protection of welding fixture During friction stir welding of steels is achieved very high temperature (above 900°C). It is necessary to protect the welding fixture from the welding heat and sticking of the weld test sheets to the fixture. Three types of protection of welding fixture and sticking prevention were tested: 1) backing bar made of 15 mm thick high speed steel Böhler S790 Microclean hardened to 66 HRC 2) backing bar made of pure oxide ceramic DISAL 100 (99% Al2O3) Fig. 3 - Time-transient temperature evolution Temperature field measurement During welding transient temperatures were measured with 6 thermocouples placed in two cross-sections in the middle of the welded sheets (see Fig.x). The thermocouples were of the k-type using wires 0.19 mm with a 5 L e t e c k ý z p r avo da j 1/2011 welded ball junction of about 0.5 mm. The CC High Temperature Cement based on Zirconium Silicate was used for fixation of the thermocouples. Data were acquired and saved using the USB external measurement system meM-Adfo BPL-16 from BMC Messsysteme GmbH. Examples of temperature records for CMT welds and sample FSW 2 are shown in Fig.2. At the figure is shown that temperature on the advancing side is markedly higher in comparison with the retreating side. The temperature difference of maximums measured with thermocouple 1 and 2 is c. 165 °C. Maximum temperature measured in distance 9 mm from weld axis reached to 490 °C. Results and discussion Feasibility of continuous FSW of high melting temperature material was successfully demonstrated. Fully consolidated butt welds of USIBOR 1500 sheets were produced when lower welding speed was applied. No surface or internal defects were evident when welding speed was up to 80 mm/min at 710 rpm or up to 160 mm/min at 900 rpm. Higher welding speed led to problems and defects formation (see Fig.x b). Three types of protection of welding fixture and sticking prevention were tested during the feasibility tests. Single backing bar made of high speed steel Böhler S790 hardened to 66 HRC turned out to be insufficient due to the test plate sticking. Also backing bar made of 99% Al2O3 was tested, but it was damaged during the first welding test, obviously due to the brittleness and too high temperature. Combination of ceramic release agent applied on backing bar made of Böhler S790 and separation pad made of Phlogopite Mica or paper was tested with very good results. Drawback was the ceramic powder or flogopit remnants at the weld root face. All welded panels were cut and metallographic analyses and tensile tests were carried out. The main objective of the metallographic examination was assessment macroscopic and microscopic pattern of the weld joints. The weld zones were compared with unaffected base material using optical microscope. A typical as-welded defect free microstructure is illustrated in Fig.x a). Four distinct regions were identified in the USIBOR1500 friction stir weld: 1) a coarse grained predominantly martensitic stir zone (CGSZ) located in the centre of SZ, 2) shoulder affected SZ (SASZ) with fine martensite located closely below shoulder, 3) a heat affected zone (HAZ) as transition region between the SZ and base material, and 4) base material with ferritic-pearlitic microstructure with carbide precipitations. The aluminium coating was stirred during FSW, producing aluminium inclusions randomly distributed along or inside the stir zone. In some cases the aluminium inclusions partially blocked consolidation of welded material and decreased tensile strength of the welds. All tensile tests were performed on a 500-kN TIRAtest 28500 testing machine at a constant crosshead displacement rate of 2 mm/min. Continuous FSW specimens without defects welded at lower welding speed failed outside the weld joint in base material. Ultimate tensile strength of the specimens was over most 541 MPa (joint efficiency over most 91%). Fig. 4 - Examples of macrographs for FSW welds: a) Si3N4 tool, 80 mm/min, 710 rpm; b) Si3N4 tool, 224 mm/mm, 710 rpm Joint strength of FSSW welds were evaluated by tensile shear and cross-tensile testing in accordance with standards ČSN EN ISO 14272 and ČSN EN ISO 14273. In case of the tensile shear test shims of the same thickness as the test specimens were used when gripping the samples to induce pure shear and to avoid the initial realignment during testing. The relationships between tensile shear failure load, tool rotational speed and plunge rate, are represented in Fig.x. It can be seen that the tensile shear strength increases with increasing rotational speed. It can be also clearly observed that higher tensile shear strength was achieved at lower plunge rate. Maximum tensile shear failure load 7.3 kN was measured for specimens welded at rotational speed c.2500 rpm and plunge rate 0.7 mm/s. Macrographs of the weld joints with the highest and lowest tensile shear strength are illustrated in Fig.x. Figure x shows the cross tension results as they vary by tool rotation speed and plunge rate. Analogous to tensile shear strength also cross tension strength increases with increasing rotational speed. Effect of plunge rate is not so clear because cross tension strength is higher for lower plunge rate 0.7 mm/s up to c. 2300 rpm and above the value the strength is higher for higher plunge rate. Maximum cross-tensile strength 3.65 kN was achieved for high rotational speed 3200 rpm and plunge rate 2 mm/s. As mentioned above, four types of material for working parts of welding tools were tested. Although fully consolidated welds was possible to realize by the means of all the tools, they markedly varied in toughness at high temperature, wear rate and chemical interaction between tool material and USIBOR 1500 steel sheets. Tools made of pure tungsten showed a tendency to wear rapidly and to change pin geometry in consequence to high loading (especially during plunging) and weldment material sticking. Average pure tungsten tool life during friction spot welding was only about 8 spot welds because of high deformation of pin. 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 Fig. 5 - Tensile shear failure load as a function of rotational speed and plunge rate Fig. 7 - Cross tensile strength as a function of rotational speed and plunge rate W25Re2HfC alloy tools appeared to be markedly more resistant to wear than tools made of pure tungsten but problem with weldment material sticking occurred too. It is not clear if the problem was caused by chemical interaction between tool and weldment material or by increased surface roughness (lathe-turning was used for final manufacturing operation). W25Re2HfC alloy is possible to classify as very robust tool material that can tolerate incorrect set-up and it does not require high precision FSW machine. Silicon nitride (Si3N4) and PCBN tools were characterized by very high hardness, excellent high temperature strength, low wear rate and high geometry stability. No problems with weldment material sticking occurred. On the other side catastrophic failure threatened if set-up welding conditions was incorrect. Tool life of PCBN tools was surprisingly lower than for Si3N4. During the welding tests radial cracks occurred in the PCBN tool shoulder. during the feasibility tests. Combination of ceramic release agent applied on backing bar made of Böhler S790 and separation pad made of Phlogopite Mica or paper turned out as the best solution. 3) Fully consolidated continuous FSW butt welds of USIBOR 1500 sheets were produced when lower welding speed was applied. Maximum tensile shear strength was achieved when welding speed was up to 80 mm/min at 710 rpm or up to 160 mm/min at 900 rpm. Higher welding speed led to problems and defects formation. 4) Higher shear tensile and cross-tensile strength of friction spot welds (FSSW) was achieved for higher rotational speed (above 2500 rpm). Effect of plunge rate is not so clear because tensile shear strength was higher for lower plunge rate but dependence of cross tensile strength on plunge rate changed with rotational speed. Maximum cross-tensile strength 3.65 kN was achieved for high rotational speed 3200 rpm and plunge rate 2 mm/s. Conclusions Feasibility tests of FSW and FSSW of high melting temperature materials were successfully performed using USIBOR 1500 steel. During the tests were tested four types of material for working parts of welding tools (pure tungsten, W25Re2HfC alloy, Silicon Nitride and PCBN). The influence of welding parameters on mechanical properties and macroscopic pattern of the weld joints has been studied. As a result, the following conclusions were achieved. 1) Silicon nitride (Si3N4) appears to be the best material for working part of FSW and FSSW tools. It is characterized by balanced performance ratio, high hardness, excellent high temperature strength, low wear rate and chemical inertness. 2) Three types of protection of welding fixture and sticking prevention were tested Fig. 6 - Examples of macrographs for FSSW welds with the highest (a) and lowest (b) shear tensile strength Acknowledgements These R&D activities were carried out in the ARC supported by the Czech Ministry of Education, Youth and Sports. The author would like to thank Rhenium Alloys, Inc for supply of W25Re2HfC sample and information about the excellent material. The author also gratefully acknowledges cooperation of BONAR a.s., CERATIZIT Austria G.m.b.H. and Benteler Automotive Rumburk s.r.o. References [1] Thomas, WM; Nicholas, ED; Needham, JC; Murch, MG;Temple-Smith, P;Dawes, CJ.Friction-stir butt welding, GB Patent No. 9125978.8, International patent application No. PCT/GB92/02203, (1991) [2] Alberto Turetta: Investigation of thermal, mechanical and microstructural properties of quenchenable high strength steels in hot stamping operations, doctoral thesis, University of Padua, January 2008. [3] Leonhardt T. Processing and Properties of Tungsten 25%Rhenium with and without Hafnium Carbide, Journal of The Minerals, Metals and Materials Society, Volume 61, Number 7, page 68-71, TMS, July 2009. [4] Mishra R.S., Ma Z.Y.: Friction stir welding and processing (part 6. Material specific issues), Materials Science and Engineering, Volume 50, p.1–78, August 2005. 7 L e t e c k ý z p r avo da j 1/2011 Experimental modal and FEM analysis of space instrument DUMMY Experimentální modální a MKP analýza kosmického přístroje DUMMY Ing. Vladimír Dániel, Ph.D., Ing. Tomáš Jamróz, Ing. Jiří Had /VZLÚ, Plc., Prague The paper focuses on the dynamic characteristics of the space instrument DUMMY. This is a substitute of flight hardware. It must also met the dynamic requirements for payload goals. The goal of the work was to compare results of experiment modal analysis (OMA) and FEM model data. The modes almost correspond to experimental modes. Frequency analysis was also performed and the results are compared with experiment parameters. Disagreements are the modes with a nonlinear contact vibration and modes with a significant geometric similarity. Článek se zabývá studií kosmického přístroje DUMMY. Jedná se o náhradu letového hardwaru. Tato náhrada musí také splňovat dynamické požadavky kladené na užitečný náklad. Cílem práce je porovnat výsledky experimentální modální analýzy (OMA) a MKP modelu. Modální analýzou byly nalezeny vlastní tvary kmitání. Tyto tvary odpovídají většině vlastních tvarů zjištěných experimentem. Dále je provedena frekvenční analýza a výsledky jsou porovnány s experimentem. Neshody jsou u tvarů s nelinearitou tvořenou kontaktem a u tvarů kmitání s vysokým stupněm geometrické podobnosti. Keywords: space instrument, modal analysis, frequency analysis, OMA, FEM sten rhenium Introduction Vibration properties of space instrument are significant in respect to design of the instrument. During the launch the instruments are affected by rocket engine, dynamic shock of solid booster burn out, stage and fairing separation and resonances incurred in the rocket. For evaluating the effects of vibration to instrument the modal analysis is fundamental. Experimental modal analysis is method for description of vibration and modal properties of instrument [2]. For experimental evaluation of modal properties more methods exists. In this work the operation modal analysis (OMA) was used. This method based on transfer function [1]. For designing the instrument parts the FEM are used. The modal analysis determines the modes and frequency analysis determines the transfer function or the frequency response function. number of modes in frequency range to 2000Hz was evaluates with using the function Multivariant MIF and SUM. The modes are in table (Fig. 2). In TestLab software the AutoMAC criteria were calculated for all modes. Modes 6 and 7 and the group of modes 9 to 13 has significant geometric similarity. Lower geometric similarity has modes 2 and 3 and 17 and 18. Experimental modal analysis - OMA Operational Modal Analysis was used for determine the modal parameters of interest space instrument – Dummy. The structure was drive by broadband noncorrelating random signal. The response was measured in 63 points. The purpose of experiment was to evaluate modal parameters in frequency range from 20 to 20000Hz. More the experiment checks the functionality of interconnection of systems LMS TestLab [3] and PRODERA. The instrument hangs on elastic rope during the test to simulate the free support (Fig. 1). The system was excited by three dynamic forces in three perpendicular dimensions. The excitation was realised by electro-dynamic drivers PRODERA. The stochastic independent broadband stationary random signals were used. The value of force was not measured. The modes were found by method Time MDOF. First the Fig. 1 - Experimental modal analysis Fig. 2 - AutoMAC criteria describes geometric similarity C z e c h A e r o s pa c e P r o c e e d i n g s 8 FE model The finite element model of Dummy and vibration adapter was prepared. The Dummy is done as 3D geometry and vibration adapter is done as 2D shell. The quadratic tetrahedron and quadratic rectangular was used. The screw connections between walls of Dummy are done as rigid connection. The sides are connects to the base only in three point. The sidewall can vibrate in transverse direction. In vertical direction this connection crates the nonlinear behaviour. The model contains slot in this place to allow transverse vibration. Nonlinear behaviour is impossible in frequency analysis [4]. The results of several first modes can be fault. The connection between disc adapter and Dummy is done as rigid by RBE2 elements. Model has free support. The mesh is on Fig. 3. Fig. 4 - Modal analysis – comparison of experiment end model, mean deviation is 16.4% Fig. 3 - Mesh of Dummy a disc adapter All parts are made from Duralumin 7075 T651 (www. matweb.com). Many steel screws are on Dummy. They increase the weight of Dummy in respect to model. More some electronic parts are inside the Dummy. Therefore the density is higher than in common Duralumin. This simplification spread the weight of screws equivalent to all Dummy structure. The adapter is full of hole which does not change the stiffness, but change the weight. The simplification above was used here too. Modal results The modal analysis was done. The comparison of calculated modes end experiment is in Fig. 4. The graphic result is on Fig. 5. The model has higher frequencies of modes. Average rate of frequencies between model end experiments is 83.51%. Some modes founded by TestLab were not found by modal analysis. The missing modes correspond to AutoMAC criterion (Fig. 2.) calculated by TestLab. Graphic representation of first mode is in Fig. 6. First mode represents nonlinear behaviour when the sidewalls with base panel forms contact which open and close in time. In TestLab software the modal dumping coefficients (Fig. 7.) were calculated for each mode. The reverse FFT method was used. Fig. 5 - Modal analysis – comparison of model and experiment Fig. 6 - First mode getting from experiment OMA method (left) and FEM model (right) - ahead to back oscillation Fig. 7 - Calculated modal dumping coefficients by TestLab software for each mode 9 L e t e c k ý z p r avo da j 1/2011 Frequency analysis The frequency analysis on model of Dummy and adapter was done. The excitation places were comparable with experiment. The excitation was realized by harmonic sinus signal with amplitude 0.5mms-2. The excitation was realized in all three axes together. The material damping coefficient was used of value 0.01. The results are showed only in interested places. They correspond to position of accelerometers used on experiment. Calculated crosspower spectrum of acceleration unit is g2/Hz. For calculating the crosspower spectrum two point results were used. Because we don't know the value of excitation force, we cannot compare the model end experiment. We will discuss only the rate of measured end experimental data. The Fig. 8 shows crosspower spectrums obtained from experiment and frequency analysis of mesh point 39159, it correspond to point 15p from OMA analysis which is the top corner of Dummy. The Fig. 9 shows crosspower spectrums obtained from experiment and frequency analysis of mesh point 28425, it correspond to point 48p from OMA analysis which is the centre of base plate of Dummy. Fig. 9 - Comparison of crosspower spectrum of Dummy – centre of base plate The experimental modes from no. 9 to no. 13 are similar. The model is not so affected by adapter vibration - first adapter mode no. 14 has resonance frequency 1388Hz. Conclusion The frequency shift between experiment and model modes is 16.5%. Some experiment modes were not found out by FEM analysis at all. These are shapes with high similarity as evidenced AutoMAC criteria. High frequency shift appears at nonlinear modes. The nonlinear behaviour is not possible to compute by frequency analysis. Frequency results present crosspower spectrums. They agree with experimental data in meaning of frequency. The amplitudes are not comparable in absolute value, because the experiment force value was not measured. The relative amplitude values agree with experiment except the range around first mode of disc adapter (1100Hz). Acknowledgment This work was supported by the Czech Ministry of Education, Youth and Sports, by the Grant No. 1M0501. Fig. 8 - Comparison of crosspower spectrum of Dummy – top corner of unite The results on figures are frequency shift of value 16.5% as a result from modal analysis of deviation between experiment and model. We can draw up the calculated frequency shift correspond to measured data. The best agreement of model end experiment is between 600Hz and 1300Hz. Under frequency 600Hz the results are affected by nonlinear behaviour of Dummy side walls. The nonlinear vibration is not possible to compute by frequency analysis. The side walls are not in contact with the base plate, the penetration shift the resonance frequency below the measured resonance. The Fig. 9 shows the results of base plate vibration. Over frequency 1300Hz the resonance peaks are shifted to higher frequencies. This frequency shift is affected by rigid screws used for connecting Dummy and adapter. The influence of rigid connection rise with frequency. As we note above we can compare only relative changes of amplitude. The big difference is mainly between 1000Hz and 1300Hz. It describes the first mode of disc adapter. References [1] Černý, O.: Teorie a praxe modálních zkoušek konstrukcí letadel; VZLÚ zpráva R – 2689/92 [2] Miláček, S.: Modální analýza mechanických kmitů; Vydavatelství ČVUT, ISBN 80-01-02333-8 [3] LMS TestLab polyMax Modal Analysis, Software. [4] MSC.Software, MSC.Nastran: Reference Guide; 2005. C z e c h A e r o s pa c e P r o c e e d i n g s 10 Simulation of level of quasistatic microaccelerations on-board of a spacecraft Modelování úrovně mikrozrychlení na palubě kosmického prostředku Ing Viktor Fedosov Ph.D. /VZLÚ, Plc., Prague This paper introduces the simulation methods of residual quasistatic microaccelerations on board of a spacecraft. On base of equations describing spacecraft orbital motion and perturbations acting on a spacecraft the main components of microaccelerations field were analysed. Some simulation results relating to orbit flight of different spacecrafts with allowance for drag are also provided. Článek uvádí do problému modelování úrovně zbytkových kvazistatických mikrozrychlení na palubě kosmického prostředku. Hlavní komponenty mikrozrychlení jsou analyzovány na základě rovnic popisujících orbitální pohyb kosmického prostředku a popisu negravitačních rušivých sil působících na satelit. Článek prezentuje výsledky modelování mikrozrychlení pro některé případy orbitálního letu kosmického prostředku s ohledem na odpor svrchní atmosféry. Keywords: quasistatic microaccelerations, orbital motion, simulation, drag Introduction Purpose of using of a high sensitivity microaccelerometer developing in the VZLÚ is measurement of low frequency (< 1 Hz) quasistatic microaccelerations on-board of an orbiting spacecraft. Nature of these accelerations is defined on base following reasoning. Let a spacecraft be a rigid body which has center of mass (CoM) in point C and let a point C’ be fixed with its frame. The difference between the gravitational field strength at the point C’ and the absolute acceleration of this point is called a residual acceleration at the point C’ (Figure 2). Base sources of the residual accelerations are: firstly, non-gravitational (non-conservative) forces such as high atmosphere drag, (in-) direct solar pressure, Earth Albedo and Earth infrared radiation which should be described with respect to spacecraft angular motion (may be active/stabilized or passive) and secondly, there are internal factors as movement and working of the spacecraft internal parts or, mainly, vibrations of the spacecraft structure during flight. In general, frequency spectrum of residual acceleration lies into wide range from ultra low frequencies to hundreds of Hz. The first group of residual accelerations sources has character of low frequency quasistatic accelerations (< 1Hz), the second - lies in range above 1 Hz. Primarily, effect of quasistatic residual accelerations has important impact to (bio-)technological experiments in orbit (Tab 1) or during missions orientated on improvement of near-earth space models (thermosphere density, Earth gravitational and magnetic fields and other). In conclusion of the brief introduction we have to define used reference frames (Figure 1). OXYZ is Equatorial or Inertial (Absolute Celestial) Coordinate System. Its center is taken at the Earth’s center. The X-axis of this system is directed towards the vernal equinox. The Z-axis is directed parallel to the Earth’s polar axis towards the North Pole of the Earth and the Y-axis is directed so as to form a right/handed coordinate system. Fig. 1 - Relative orientation of IRF (OXYZ), orbital RF (CXoYoZo) and SC RF (Cxyz) Datum point of Orbital Coordinate System CXoYoZo is taken at the Centre of Mass (CoM) of a spacecraft. X-axis (CXo) is directed parallel to the vector of angular momentum (A) of an orbiting spacecraft. Z-axis (CZo) is directed from the Earth’s center to the Zenith through spacecraft’s CoM the Y-axis is directed so as to form a right/handed coordinate system. Generally, the next coordinate system 11 L e t e c k ý z p r avo da j - Spacecraft reference frame Cxyz - is defined by satellite manufacturer. In the paper, unless otherwise indicated, we use spacecraft structural coordinate system such that its axes are oriented parallel to the Xo, -Yo and –Zo axes. Back and forth transformations between mentioned coordinate systems can be expressed by rotational angles (Direction Cosine Matrix) or by quaternion functions. These transformations are described in extensive literature (for example, [2], [3]). a (r ) = α r 1/2011 of a local vertical Denote by vector a (r ) acceleration of any point into V. This acceleration is equal to difference between absolute acceleration α r of the point and a (r ) at the point of the V: (2) a (r ) = α r − g (r ), (∀∆r ∈ V ) In addition, suppose that triaxial accelerometer is mounted at the point (for example, C’) into V. In this case, namely acceleration describing by (2) will be measured by the instrument. With respect to (1) and (2) the acceleration at the spacecraft CoM ( ∆r ∈ =V 0)) is equal to sum of external non-gravitational − g (r ), (∀ accelerations acting to the spacecraft: Seeing that absolute acceleration of any points of the V is described as • α (r ) = aC + ω × ∆r + (ω × ∆r ) × ω we can express quasistatic microacceleration at the point of the V as following: (3) • a (r ) = ΓC + ω × ∆r + (ω × ∆r ) × ω + aG , (∀∆r ∈ V ) aGG = g C − g (rC + ∆r ) is an acceleration due to • gravitational gradient at the point C’, ω , ω are vectors of where Tab. 1 - Base perturbations and related accelerations acting on-board of Russian Space technological systems [1],* - specialized small space platforms angular velocity and acceleration of the spacecraft. Whereas that ([5]): Analytical expression of the residual quasistatic accelerations Suppose that a spacecraft orbital motion is described by following differential equation: d 2 rC = g + Γ , where: dt C 2 (1) C We suppose that spacecraft’s CoM lies in internal space of a spacecraft (payload area V, see Fig 2) so we can approximate U (r ) as following: µ ,where is a gravitational parameter of a centre of attraction (for example for Earth = GEM, where GE is the Earth’s Gravitation Constant and M is Earth’s mass). Hence for any point C’ (figure 1) we can define its position vector as r = rC + ∆r (∆r ∈ V , ∆r / rC < 1) and g (r ) = − µr / rC 3 e.g., for the point C’ g (r ) = − rC 3 [3(e ⋅ ∆r )e − ∆r ] We can express quasistatic acceleration at the point C’ as • g C = − gradU (r ) r = rC rC µe a (r ) = ΓC + ω × ∆r + (ω × ∆r ) × ω + rC is a position vector of spacecraft CoM, ΓC is a perturbative acceleration due to non-gravitational perturbations acting to a spacecraft, g C is a field force with potential U (r ) and U (r ) = − aGG = µ r 2 er where er = r is an ort r µe rC 3 (4) [3(e ⋅ ∆r )e − ∆r ] This formula is principal expression allowing us to simulate quasistatic acceleration at any point on-board of an orbiting spacecraft. • In case, if parameters of spacecraft angular motion ( ω , ω ) are known we are able to define quasistatic acceleration at the point C’ situated at distance ∆r from spacecraft CoM taking into account non-gravitational perturbations ( Γ ) acting to CoM. How it was noted above the microaccelerometer mounted at the point C’ (any point of the V) will measure accelerations described by (4). We can use experimental data obtained by the accelerometer and data predicted by (4) in following way. We can reconstruct the spacecraft motion by spacecraft telemetry (for example, using data from on-board magnetometers, star-camera or by angular velocity sensors) and calculate the acceleration along the reconstruction by formula (4) in 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 point of accelerometer location. Comparison the calculation results with the results of processed acceleration measurements in the same time interval defines method how to verify the accelerometer during flight. 12 The main perturbation at this orbit is drag therefore we can consider that: Γ = aD Perturbation due to atmosphere drag ( aD ) is possible to calculate as 1 C D ρAV 2 , where: 2m aD = C‘ ∆r C V CD is aerodynamic drag coefficient, for the simulation case it is equal to 2.2 A is an effective cross-sectional area = ¼ of the spacecraft surface area is thermosphere (high atmosphere) density, for the simulation case is calculated by using of the thermosphere model TD88 in relation to reference environmental condition (solar flux and geomagnetic parameter) and spacecraft orbital parameters. TD88 detailed description is given in [6]. V is the spacecraft velocity relative to atmosphere rotation, for the simulation case it is possible to define high atmosphere as statically and V is equal to the spacecraft absolute velocity. Simulation time is equal to one orbit. Values of the atmosphere density and spacecraft acceleration (in spacecraft coordinate system) due to drag are shown in Figure 3. Fig. 2 - Payload space V Simulation The formula (4) was derived for a general situation without any frequency restrictions. But if the spacecraft attitude motion is like as motion of a rigid body (this motion is usually very slow), then formula (4) gives only a quasi-steady acceleration component, i.e., a part of a real acceleration that has frequencies within the range from 0 to 0.1 Hz. On this account, thereinafter, we will study spacecraft free motion as motion of rigid-body. Two cases of the motion are presented: spacecraft torque-free motion and motion under external moments of perturbations (gravitation moment is considered). 1. Suppose, that a spacecraft which is characterized by size (0.5 x 0.5 x 0.7 (m)) and mass (80 kg) moves at orbit with following parameters: - orbit period – 100 minutes; - circular orbit; - altitude – 450 km; - inclination – 98,8 o ; - perigee argument – 66,5 o - eccentricity – 0,00112 Fig. 3 - Accelerations due to drag Rotation motion of the spacecraft (as rigid –body) relative to its CoM is simulated by dynamic Euler equation in matrix form: J dω + (ω ×)( J ⋅ ω ) = M dt (5) where: J is an inertia tensor of the spacecraft (in the spacecraft refe• rence frame), ω ,isωa vector of spacecraft angular velocity, M is • a vector of external moments acting to the spacecraft and (ω ,X) ω is a skew-symmetric matrix of spacecraft angular velocity: 13 L e t e c k ý z p r avo da j 0 (ω×) = ω z − ω y − ωz 0 ωx 1/2011 ωy − ωx 0 Let the spacecraft inertia tensor is 0 0 4,93 J = 0 4,93 0 0 0 3,33 And initial values of x, z, z -components of satellite angular velocity vector are (rad/s): Fig 5 Components of angular acceleration vector M (external moments) = 0 ω x 1e − 3 ω = ω y = 1e − 3 ω z 1e − 2 Figures 4 and 5 show diagrams of vector components of spacecraft angular velocity and accelerations obtained by solution (5) at M = 0. Fig 6 Quasistatic accelerations along to X axis of the spacecraft reference frame (symmetrical spacecraft) 2. The case 1 describes torque-free motion of the symmetrical rigid body. Now we will consider unsymmetrical spacecraft motion. Suppose the spacecraft orbit, mass, its sizes and initial values of x, y, z -components of satellite angular velocity vector are the same as defined in case 1. Fig 4 Components of angular velocity vector M (external moments) = 0 Let the spacecraft inertia tensor is following: In this situation, simulated angular motion of the spacecraft corresponds with motion of free symmetric gyroscope (z is a spin axis) – its nutational oscillations. Describable case is useful for validation of the applied simulation method because makes it possible to compare results of simulation ( with well known dynamical phenomena. • ω ,ω ) Figure 6 shows diagrams of accelerations in x spacecraft direction at the spacecraft CoM (equivalent to the drag) and at the point with 2 mm offset from CoM along the axis 6.86 0.03 5.9 J = − 0.03 8.66 − 9.5 − 5.9 9.5 5.4 Figures 7 and 8 demonstrate diagrams of spacecraft angular velocity and accelerations vectors components obtained by solution (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 14 Figure 10 demonstrates level of accelerations in x spacecraft direction at the spacecraft CoM (equivalent to the drag) and at the point with 2 mm offset from CoM along the axis in the spacecraft structural coordinate system for initial angular velocity components [1e-4; 1e-4; 1e-3(rad/s)] (green curve relates to the symmetrical spacecraft, red – to the unsymmetrical). Fig 7 Components of angular velocity vector (unsymmetrical spacecraft) Fig 10 Quasistatic accelerations along to X axis of the spacecraft reference frame Experiment Fig 8 Components of angular acceleration vector (unsymmetrical spacecraft) Figure 9 shows plots of accelerations in x spacecraft direction at the spacecraft CoM (equivalent to the drag) and at the point with 2 mm offset from CoM along the axis (green curve reflects situation when Mx = 0 and red graph for case abs(Mx) > 0). Fig 9 Quasistatic accelerations along to X axis of the spacecraft reference frame (unsymmetrical spacecraft) We have had opportunity to verify reviewed method of quasistatic accelerations estimation by analysis of data obtained by accelerometer during orbit flight. The accelerometer has been placed on board of Russian spacecraft Tatyana/ Univerast – 2. This small platform was designed for flight testing of new tri-axial stabilization system with low level of produced microaccelerations. Orbit parameters of the Universat-2 are following: circular sun-synchronous, inclination – 98 degrees and altitude is 820 km. Satellite was launched on September 2009. Unfortunately, after spacecraft separation from the launcher upper stage the satellite stabilization system did not work correctly (infrared Earth sensor failed). Communication with the spacecraft was interrupted after one month operation of the satellite in orbit. For analysis of the accelerometer flight data we have to take into account this real platform situation in combination with existence of displacement of the accelerometer relative to the center of mass of the satellite. The proof mass offset will cause that not only accelerations due to the non-conservative forces but also accelerations due to gravitational forces and angular motion act on the proof mass. Hence, the measurement model we can describe according to formula (4). As we have possibility (on base of telemetry data) to reconstruct parameters of spacecraft attitude motion (components of satellite angular velocities) than we can estimate the magnitude of microaccelerations at instrument position and compare it with measured data. It has been confirmed that during observa- 15 tion windows the character of the spacecraft angular motion was primary component influencing accelerometer measurements. So, maximum predicted values of non-gravitational accelerations acting along each axis of the satellite reference frame should not be more than 3.5x10-8 ms-2, for accelerations rising due to gravitational gradient – 0.64x10-8 ms-2. The level of computed values of accelerations rising due to spacecraft rotation was defined as two order bigger (+3x10-6 ms-2). The Fig. 11 and Fig. 12 show typical observing window plots of raw data of the accelerometer (green, ms-2) versus angular velocities along X and Y satellite axes (blue, rad/s). Finally the Fig. 14 demonstrates model (red) describing dependence of measured smoothed data (blue) on angular velocities of stabilization system (high level interventions to stabilize the spacecraft along direction to Earth) and temperature registered at sensor. Green curve reflects absolute values of residuals of model predictions related to measured data. The time resolution on all figures is one minute. L e t e c k ý z p r avo da j 1/2011 Fig 14 Accelerometer measurements as function of components of angular velocity and sensor temperature (during one revolution) Conclusion The paper describes estimation method of level of quasistatic accelerations on board of a spacecraft. Performed simulations demonstrate variation of accelerations level in dependence on dynamical characteristics of the spacecraft and its angular motion. This method has been validated by analysis of microaccelerometer data received during space flight on board of the Russian spacecraft Universat-2. The base expression (4) for microaccelerations calculation can be applied for non-gravitational accelerations determination from measurements (aM) obtained by microaccelerometer mounted at the point C’ (C’ (x, y, z) of V). • ΓC = aM − (ω × ∆r + (ω × ∆r ) × ω + µe rC 3 [3(e ⋅ ∆r )e − ∆r ]) In the case, uncertainties relating to accuracy of the sensor accelerometer location and spacecraft angular velocity determination should be analyzed. Fig 12 Spacecraft angular velocity along X axis and accelerometer measurements (one revolution) References [1] Polezhaev, V.I. and other: Konvektivnyje processy v nevesomosti. Moscow, Nuaka, publishing of Russian Academy of Sciences, 1991 [2] Herrick, S.: Astrodynamics. Orbit determination, Space navigation and Celestial Mechanics University of California, Los Angeles,1971 [3] Narimanov, G.S.: Osnovy teorii poleta kosmicheskych apparatov, Mashinostroenie, Mocsow, 1972 [4] Beliakov, I., Borisov, D.: Osnovy kosmicheskoy technologii. Moscow, Mashinostroenie, 1985 [5] Polezhaev, V.I., Ermakov, M.K, Nikitin, N.V., Nikitin, S.A, Yaremchuk, V.P.: The Use of Microaccelerations Data for Convection Modeling & Analysis of the Microaccelerations Limits, 7th Annual Microgravity Environment Interpretation Tutorial, V.2, Section 24, 2004 [6] Sehnal L., Pospisilova, L.: Thermospheric Model TD88. Astronomical Institute Czechoslovak Academy of sciences. Preprint N67. [7] Beletsky, V.: Motion of an Artificial Satellite about its Center of Mass. Moscow, 'Nauka', 1965 (in Russian), English translation: Israel Program for Scientific Translation, Jerusalem, 1966 Fig 13 Spacecraft angular velocity along Y axis and accelerometer measurements (one revolution) [8] Preresty, R., Chvojka, M., Fedosov, V.: Use of the high sensitive electrostatic accelerometer for orbit perturbation effects investigation on board of LEO spacecraft. Proceedings of International Astronautics Congress 2010, Prague 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 The influence of structural damping on flutter characteristics of a small sport aircraft Vliv strukturálního tlumení na flutterové charakteristiky malého sportovního letounu Ing. Martin Zejda / Aerospace Research Centre, FME CTU in Prague; Ing. Aleš Kratochvíl / Aerospace Research Centre, FME CTU in Prague; Doc. Ing. Svatomír Slavík, CSc. / Czech Technical University in Prague, Faculty of Mechanical Engineering, Department of Aerospace Engineering Authors present the basic mathematical models of damping and principles of experimental determination of structural damping from the system frequency response. The specific method was applied on measurements of few types of small sport aircraft. The flutter calculation of tail surfaces with consideration of structural damping influence is shown. The applicability of the used method is discussed for the need of further flutter calculations. Jsou představeny základní matematické modely tlumení a experimentální zjišťování velikosti strukturálního tlumení z frekvenční odezvy systému. Vybraná metodika byla aplikována na měření několika typů letounu. Je uveden příklad výpočtu flutteru ocasních ploch s uvážením vlivu strukturálního tlumení. Je diskutována aplikovatelnost použité metodiky pro potřeby aeroelastického výpočtu. Keywords: structural damping, logarithmic decrement, flutter analysis, half-power point method Introduction One of the research topics of Aerospace Research Centre at Department of Aerospace Engineering – Faculty of Mechanical Engineering, CTU in Prague is focused on a flutter characteristics analysis of small sport aircrafts, with maximal take-off weight up to 600 kg. The calculation model, which is used for flutter test, is based on standard “p-k” model where the damping characteristics and frequency are dependent on speed [4]. Thus, the actual real value of damping for each mode of the airplane structure is very desirable physical quantity to know. Damping as one of the structural property significantly influences the aeroelastic behaviour of airplane structure. This characteristic can be substituted by appropriate mathematical model, when practical calculations are needed, or it can be directly gained from ground frequency tests. The shown method is based on determination of structural damping from FRF (Frequency response function) of the forced structure. The values of damping are collected to obtain an adequate database of how the typical airplane structure behaves under dynamic loading. Structural damping Damping of mechanical structures can be approximately divided into these three categories • material damping – related to the molecular structure of the material • structural damping – usually caused by friction between parts of the structure • external damping – caused by interaction between the structure and environment We can see that relations between damping and its causes make it more problematic to measure. As a result the indirect methods are often used, to determine the value of system damping. Considering the influence of structural damping in computation of dynamical structural behaviour is usually accomplished by usage of the simpler model where is expressed by suitable mathematical expression. The damped motion is more frequently described by these mathematical models. The simplest one is the viscous, where the damping forces are linearly dependent to velocity of the motion. The governing equation of motion is mx + bx + kx = F or x + 2ζΩx + Ω 2 x = F m where Ω means the undamped natural frequency and is the damping ratio defined by Ω2 = k b ; 2ζΩ = m m Hysteresis mathematical model is used for cases, where damping force is proportional to elastic force. The phase shift between forces is π/2. The motion of system is described with equation mx + (1 + jη )kx = F 17 L e t e c k ý z p r avo da j Using the expression for Ω leads to x + (1 + jη )Ω 2 x = F m F 2 x + (1 + jη )Ω x= The term in brackets represents the complex stiffm ness. If the value of damping is proportional to stiffness and mass of the structure, the reasonable mathematical model to use is a proportional model, characterized by governing equation 1/2011 The practical evaluation of damping is very tough and therefore this method can be recommended only as a control tool. The half-power point method is reasonable, when FRF (Frequency response function) of structure is measured. Method is based on relationship between the damping ratio and position of half-power points. The half-power points are those in a FRF diagram in which the amplitude decays 1 to of a peak value (Fig. 2). 2 m + (a1m + a2 k )x + kx = F ( ) x + a1 + a2 Ω 2 x + Ω 2 x = F The terms a1 and a2 represents the coefficients of proportional damping. Mathematical models are described in more details in [1]. Experimental evaluation of structural damping The classical approach, how to evaluate the structural damping of the system, is to determine the decay of amplitudes, from time flow of the oscillating motion. Assuming the linear behaviour of the system applies to ratio of two successive amplitudes Figure 2 x(t +T ) / xt = e −ζ ΩT Structural damping can be simply expressed in form of logarithmic decrement – shown in the Fig.1. This results to be the half peak value in power spectra diagram, so that’s why the half-power points. Assuming a linear behaviour and small damping the relation is ω2 − ω1 = 2Ωζ ω2 − ω1 =ζ 2Ω The main advantage of this method is the possibility to determine the value of structural damping directly from measured frequency response function. The results are valid only for modes with linear behaviour and sufficient frequency shift differing them from other modes. The method is discussed in more details in [3]. Data acquisition Figure 1 x 2πζ = ϑ´= const ln 2 = 1− ζ 2 x1 In most cases the value of logarithmic decrement is assu med to be , so than we can write ϑ = 2πζ Data were gathered during the ground frequency tests of several small aircrafts. The measurement system consisted of an 11 channel analyser TL-5412_CDD, 2 electrodynamics exciters, 2 channel signal generator and power amplifiers. The piezoelectric accelerometers of IEPE standards were used. The measured data were analysed in ME scope software which analyses frequency characteristics by means of Fast Fourier Transformation of signal data from a time domain in to a frequency plane. The airplane with fixed control surfaces was softly hung on adjustable frame, when natural frequencies of body were measured. Vice versa the modal parameters of the control surfaces were measured, when body structure was fixed. C z e c h A e r o s pa c e P r o c e e d i n g s Contr. surf. HT Wing mode type 18 type of airplane SD - 4 Viper f [Hz] d [-] PiperSport f [Hz] d [-] VL-3 Sprint f [Hz] d [-] Samba f [Hz] TL-3000 Sirius d [-] f [Hz] d [-] Skylane f [Hz] d [-] 1. sym. bending 11,2 0,072 11,3 0,055 8 0,040 5,8 0,075 10,8 0,115 12,1 0,104 1. sym. torsion 49,4 0,092 40,8 0,114 31,1 0,195 31,6 0,069 21,6 0,109 34,1 0,184 1. antisym. bending 13,45 0,113 23,7 0,172 19,7 0,103 15,6 0,212 17,3 0,089 19,4 0,210 1. antisym. torsion 48,3 0,084 39,6 0,043 30,5 0,206 31,9 0,057 30,4 0,151 34,8 0,253 1. sym. bending 18,8 0,167 22,9 0,217 17,7 0,125 16,1 0,176 13,3 0,083 33,6 0,178 1. antisym. bending 15,9 0,100 13,1 0,144 13,5 0,183 14,2 0,121 12,2 0,129 11,8 0,295 FRE* 30,6 0,386 20,1 0,328 25,2 0,536 6,06 0,476 19,2 0,278 15,6 0,550 FRA** 20,4 0,482 14,4 0,486 25,6 0,508 7,25 0,597 11,4 0,579 14,9 0,591 * Fundamental rotation of elevator ** Fundamenta rotation of ailerons Tab. 1 - Frequencies and values of logarithmic decrement Tab. 1 - Frequencies and values of logarithmic decrement The structure was excited by two shakers. These were powered by a sweep sine signal with corresponding or opposite phase, to obtain symmetric or antisymmetric excitation. The frequency range was set from 2 to 100 Hz. Natural frequencies were specified by means of amplitude values in real and imaginary components. Results The values of structural damping were measured on six types of sport aircrafts. These were all-metal low-wing monoplanes, composite low-wing monoplanes and strut braced high-wing monoplanes. Frequency test were performed with a complete equipped airplanes at the light-mass test configuration, so it means that only one light pilot and empty fuel tanks were simulated. The half-power point method was used to determine the logarithmic decrement of typical modes of both structure and control surfaces. Modes were chosen according to conditions of used method, to obtain significant results. These are shown in Table 1. For each mode the frequency and negative logarithmic decrement are given. Fig. 3 – Map of logarithmic decrement vs. frequency for modes of structure and control surfaces The methodology FAR 23.629 allows to use the damping value of g = +0,03 as an inherent structural damping. This is adequate to value of negative logarithmic decrement d = 0,1. However, this value should be used with caution if the damping of the mode decreases very rapidly with an increase in airspeed. On the Figure 3 the relationship between frequency “f” and value of negative logarithmic decrement “d” is shown. As we can see, there are four areas corresponding to modes of the structure or control surfaces. As a benchmark the value of damping according to FAR 23.629 is given. Modes of the structure more or less correspond to the value given by FAR 23.629, with average value of d=0,122 for wing modes and d=0,160 for horizontal tail modes. The modes of control surfaces show considerably higher values of damping d=0,483 in average, which is nearly five times higher than FAR 23.629 value. Example of calculation The example calculation was performed on all-metal down-wing airplane of foreign manufacturer with design speed of VD = 275km/h see Fig. 4 – 5. Fig. 4 19 L e t e c k ý z p r avo da j The modes of tail structure, rudder and elevator were calculated. Force effects of a pilot were simulated by 1kg of mass, placed on a control stick. Flutter equations were based on Lagrange’s energy equations of a structure elastic system with control surfaces. The “p-k” computational model was used to find eigenvalues of complex flutter matrix. The structural damping values were added into generalized stiffness matrix [4]. The results are presented in Fig. 6 - 8 which show the dependence of damping (in form of negative logarithmic decrement) and frequency versus equivalent air speed. The degrees of freedom stand with five eigenmodes of primary structure with fixed elevator and rudder and three eigenmodes of elevator with fixed stabilizer and fin. 1/2011 In the second case (see Fig. 7), respecting FAR 23.629, the value d=0,1 was used for all eigenmodes of control surfaces and primary structure. The speed VFL=267,1km/h and speed ratio VFL/VD=0,971 lead to the same conclusion as in the previous case. d [1] d [1] VEAS [km/h] f [Hz] VEAS [km/h] f [Hz] VEAS [km/h] Fig. 7 – V-d and v-f diagrams (Structural damping according to FAR 23.629) VEAS [km/h] In the last figure (Fig. 8) the real measured structural damping of critical mode d=0,512 was added to the computation. The critical second mode of elevator still shows a decline of damping at speeds over 170km/h but the flutter speed goes above 1,2VD. The result is, that the tail structure was said to be flutter free, respecting the real structural damping. In addition to improve the inherent flutter stability, the manufacturer was advised to rebalance the elevator. Fig. 6 – V-d and v-f diagrams (Without structural damping) d [1] In Fig. 6 the structural damping was not used in calculation. The second mode of elevator is supposed to be critical. The value of negative logarithmic decrement falls to zero at speed of VFL=170,6km/h, which is the critical flutter speed and gives the speed ratio VFL/VD=0,620. Considering the inherent structural damping of d=0,1 , given by FAR 23.629, the flutter speed goes up to approximately VFL=273,6km/h and the speed ratio raises to VFL/VD=0,995. According to FAR 23.629 the airplane structure should be flutter free up to 1,2VD, Therefore it is supposed that some undamped dynamical effects can be observed. VEAS [km/h] C z e c h A e r o s pa c e P r o c e e d i n g s f [Hz] VEAS [km/h] Fig. 8 – V-d and v-f diagrams (Measured structural damping) Conclusion The values of structural damping were evaluated for typical eigenmodes of six aircraft structures. The half-power point method was used. The results show that modes of control surfaces are always remarkably higher damped than modes of primary structure. The possible causes of this phenomenon could be the effect of control system path especially its stiff- 20 ness and friction between rotating and moving elements. The question is how the structure behaves, when effects of real pilot are considered. This should be the topic of subsequent research. Although the half-power point method seems to be very useful for determination of structural damping, it is valid only for linear eigenmodes with sufficient frequency shift from other modes, otherwise it gives incorrect results. This is a main disadvantage, since the airplane structure modes can be generally complex with characteristics of nonlinearity. The effects of structural damping were shown on example flutter calculation of tail structure. It was proved, that knowledge of real damping values is very reasonable, especially for control surfaces modes, which are supposed to be the most critical. References [1] Miláček, S.: Měření a vyhodnocování mechanických veličin; Vydavatelství ČVUT, Praha 2001. ISBN 80-01-02417-2. [2] Stejskal, V., Okrouhlík, M.: Kmitání s Matlabem; Vydavatelství ČVUT, Praha 2002. ISBN 80-01-02435-0. [3] Bilošová, A.: Experimentální modální analýza; VŠB Technická univerzita Ostrava, Ostrava [200?] [4] Slavík, S.: Flutter calculation model with isolated modal characteristics of control surfaces for small sport airplanes; Czech aerospace proceedings, Vol. 2, 2008, ISSN 1211-877X. Small aircraft ECS optimization using 1D tool Optimalizace ECS malých dopravních letadel pomocí 1D simulačního nástroje Ing. Jiří Hejčík, Ph.D. / Institute of Aerospace Engineering, Brno University of Technology; Prof. Ing. Miroslav Jícha, CSc. / Department of Thermodynamics and Environmental Engineering, Brno University of Technology The paper details evaluations of a heat exchanger performance influence on a small airplane and/or helicopter ECS. The ECS performance platform’s operation under different pressure losses and heat exchanger effectiveness is numerically simulated using a multi-domain 1D simulation tool LMS Imagine.Lab AMESim. Příspěvek se zabývá studií vlivu parametrů tepelného výměníku na výkon klimatizačního systému malých letounů a vrtulníků, na základě provedených numerických 1D simulací chování klimatizačního systému, s využitím simulačního nástroje LMS Imagine.Lab AMESim. Keywords: ECS, Heat exchanger, Air conditioning Introduction The Environmental Control System (ECS) is a key component for flight safety. It is basically an air conditioning system, which main task is to deliver air at the required pressure, temperature and humidity to the cockpit and cabin. It has to work during all phases of flight and for any occupancy. The ECS is also responsible for the cabin pre- ssurization control with the possibility to regulate pressurization speed (time required for cabin pressure increase or decrease) in case of an aircraft with a cruising ceiling of over 3000 m. Since the benefit of the ECS to flight safety is undisputable, it is also being installed to small aircrafts, but only when the reliability, efficiency and weight of the ECS fulfill designer’s requirements. 21 L e t e c k ý z p r avo da j 1/2011 Environmental Control System ECS behaviour simulation The ECS is in principle different from air-conditioning systems known from the automobiles or household appliances because it works with no refrigerant. Only air is used in the ECS either as a coolant or a heating medium. Because of the required high overpressure (the pressure in the cabin is up to 62 kPa above ambient) and huge volume of the delivered air to the ECS the ventilating air is delivered from the aircraft’s or auxiliary power units (APU’s) engines [1]. This “bleed” air cannot be used directly for ventilation because it is too hot (about 200 °C) hence it is necessary to cool it down to a suitable temperature before it enters the cockpit or the cabin. The balance between heat exchanger, turbine and compressor parameters is mandatory for the effective ECS performance. Probably the easiest way to reach the parameter balance is to tune up heat exchanger parameters, because its design is more flexible than that of the others. The 1D model of the system shown in Fig. 1 was created in a multidisciplinary 1D simulation tools LMS Imagine. Lab AMESim (Fig. 2) to understand how the ECS behaviour is influenced by the heat exchanger parameters. This model is based on some of experimental data e.g. turbine and compressor characteristics, power losses in bearings and temperature drops in connecting pipes, so that it simulates commercially available ECS system behaviour. Commercial Aircraft’s ECS is a quite complicated device where the bleed air goes through an ozone converter, cooling unit and water separator to the mixing chamber where it is mixed with the recirculating air. Subsequently, the required air temperature and humidity is set and ventilating air enters to the distribution system. On the other hand the small aircraft’s and helicopter’s ECS is quite simple and it usually contains an air cooler and water separator only. It is caused by the lower ceiling and flying range of small aircrafts so that some parts are not necessary (e.g. ozone converter, air recirculation etc.), and the directly treated bleeding air is suitable for effective cabin ventilation. And how does the small aircraft’s ECS work? The bleed air from the engine goes through the cooling unit - usually an air to air heat exchanger - which pre-cools the bleed air by ambient air. After that it flows to the expansion turbine where it is finally cooled. Certain amount of the bleed air is added afterwards to set the required air temperature. The last operation before the air is distributed to the cabin or cockpit is moisture removing. The simplest arrangement of cooling units, as it is used in small planes, is shown in Fig. 1. Fig. 2 ECS model in Amesim The heat exchanger efficiency and pressure drops were used as simulation inputs. The obtained values of the bleed air and ambient air parameters (flow rate, temperatures and pressures at specific points), turbine and compressor RPM and power losses in bearings were the outputs. The simulation were done for two operational regimes (see Table 1) with heat exchanger parameters varying in the range of 50 to 100 % for the heat exchanger efficiency and for the relative pressure drops from 0,9 to 9 % for the bleed air and from 0,5 to 12 % for the ambient air. Table 1 ECS regimes Fig. 1 ECS of small aircraft or helicopter Although there are no all control valves depicted in Fig. 1, it is evident that the thermal performance of small ECS is set by the bleed air flow rate, expansion turbine + compressor characteristics and heat exchanger parameters. Regime 1 Bleeding Air Inlet pressure 331 900 Pa Inlet temperature 165,6 °C Expansion pressure 110 300 Pa Ambient Air Inlet pressure 95 100 Pa Inlet temperature 16,4 °C Table 1. ECS regimes Regime 2 300 500 Pa 190,9 °C 109 700 Pa 95 100 Pa 43,9 °C C z e c h A e r o s pa c e P r o c e e d i n g s 22 Conclusion Probably the most important value for the cooling part of the ECS is the expansion turbine outlet temperature. This is shown in Fig. 3, where there is the influence of the outlet temperature on the operational regime and pressure drops or heat exchanger efficiency (constant pressure drops). The simulation results confirm the expectation that the heat exchanger efficiency has the greatest influence on the outlet air temperature. On the contrary, the most surprising was the finding that the pressure drops on the bleed air side of heat exchanger have only minor influence on the outlet temperature, where the 10 times higher pressure drops of the bleed air caused the outlet temperature change about 5 °C. Further analysis of results shows that this is due to the turbine characteristic, where the expansion is changing in accordance to the air mass flow. The influence of the ambient air pressure drops seems to be negligible on the Fig. 3, where for the constant heat exchanger efficiency and constant pressure drops of the bleed air the outlet temperature from the expansion turbine is almost constant. But one has to take into account that the heat exchanger efficiency is the input for the simulation and the mass flow rate of the ambient air is almost 2 times greater than for the bleed air, hence the bleed air temperature at the outlet of the heat exchanger is constant. It shows the weak point of simulations, because there is no dependency between mass flows and heat transfer (heat exchanger efficiency). The simulations show the basic influence of the ECS behaviour on the heat exchanger parameters. However they show the procedure is good enough to compare different ECS but not for their optimisation, because input parameters are independent what is not true in reality. Fig. 3 Heat exchanger influence on ECS behaviour The improved ECS model containing the thermal hydraulic behaviour of the heat exchanger is currently under development. Because this improved model is more complex than the described one, it should be used not only for the ECS optimization but also as an input generator for the CFD simulation of cabin air quality. Acknowledgement This work was supported by CLKV (task A6 Prediction of indoor environment in aeroplane cabin) and by MPO project FI-IM5/217 Air conditioning system for helicopters and small airplanes. References [1] HUNT, Elwood H., et al. Commercial Airliner Environmental Control System : Engineering Aspects of Cabin Air Quality [online]. c1995-2009 [cit. 2009-05-10], <http://www.boeing.com/ commercial/cabinair/ecs.pdf>. Intentionally left void. 23 L e t e c k ý z p r avo da j 1/2011 Development of series production technology for stator and rotor blades Vývoj technologie výroby statoru a rotoru pro opakovanou výrobu. Ing. Karel Barák; Ing. Michal Řehák/ Department of Aerospace Engineering, Czech Technical University in Prague, Faculty of Mechanical Engineering This paper relays the soundness of technology pretaining to the structural componants of a fan propulsion demonstrator developed for small airplanes. This demonstration was intended for experimental verification of constructional and technological solution and acquiring data for subsequent construction optimization. The paper‘s intent is especially focused on the technologies for the production of the composite rotor blades and stature of ducted fan propulsion for repeated production. Příspěvek se zabývá technologií výroby částí demonstrátoru ventilátorového pohonu, určeného pro experimentální ověřování konstrukčních a technologických řešení pohonu a získávání dat pro následnou optimalizaci konstrukce. Příspěvek je zaměřen zejména na vývoj technologie výroby kompozitních součástí rotorového systému a předstatoru pro opakovanou výrobu. Keywords: UL aeroplanes, cold jet, ducted fan, ultra light rotor technology, composites material. Introduction 1 Rotor blade Demonstrator of fan propulsor is a system that serves for verification of correct function of a new propulsion unit with canal fan. This propulsor is developed for small airplanes. This test equipment was described in articles published in lit [1]. Construction of original rotor blades and stator was described in lit [2]. The described parts of fan propulsor are made of glass fiber and carbon fiber composit with foam cores. This makes possible to achieve high stiffness. Rotor blade of fan propulsor is composite sandwich structure based on low viscosity polymer resin, carbon fibers and a light-weight foam. On the basis of provided information was done the balance of vacuum infusion processing (see figure 1). Generally this technology is suitable for making of smaller products, whose shape is complicated, requiring high geometric precision and surface quality [2]. Selected technology meets all requirements for series production. The part was made in three-parts mould. This one allowed to make leading edge of the blade without mould joint (especially in term of aerodynamics). Visual analysis was made after curing of each of single specimen. Then eventually changed characteristics of infusion process for improve surface quality by removal of air voids (of defects). In the figure is fifth sample of rotor blade after curing and removal out of the mould (see figure 2). The goal was rotor blade with minimum defects mainly caused by air voids and with approved structure. Conception of the rotor blade The rotor blade consists of foam core with titanium bushing for a hinge, carbon flange and a coat. The shell of the blade is made from carbon reinforcement that is assimilated to transfer of load (tensile load and torsion) from the blade to the titanium bushing [1]. In the figure is three-dimensional CAD part of the blade (see figure 3). Materials The offer of epoxy systems, that are primarily used for vacuum infusion process,is large. Two component epoxy C z e c h A e r o s pa c e P r o c e e d i n g s resin was chosen after previous experience in producing simple flat table parts. Carbon reinforcements was used in blade making. For the flanges of the blade is used high strength unidirectional carbon fabric (see figure 4). The shell of the blade is manufactured from balanced high strength carbon fabric with plain weave (see figure 5). The foam core was made from one - component polyurethane foam. Its dentsity is about 25 kg/m3. Manufacturing moulds The blade was made with using of composite mold, that was assimilate to vacuum infusion technology. The rotor vane mold was made with using of milled duraluminium model (see figure 6). This one was made by NC tooling. Own composite mold was made by contact lamination on positive master model. The materials of the mould are glass fabric and basic epoxy resin. Predefine chosen technology does not pose claim to thermal resistivity of resin system. At the beginning of making mould was used gelcoat. A gelcoat is a material used to provide a high-quality finish on the visible surface of the mold. Then the surface is also hardier to mechanical damage. The gelcoat eminently affected final quality of blade surface. The mould was equipped with steel bushings and quidepins for easier right assembling of all its parts. Compagination mold parts was realized by screws and nuts (see figure 7). Sealing of the mold was maked through vacuum bag and sealing tape. Sharp ends of screws was necessary cover up with silicone rubber before vacuum processing, in order to not for vacuum leakage (see figure 8). Molds of blade cores were made on milled duralumin model. The moulds were casted from silicone rubber (see figure 9). 2 STATOR Stator is located at the end of entry duct. Load of stator blades is minimal. The load is carried by the curved entry duct. The original stator consists of these basic parts: outer ring, inner ring and fourteen blades (see figure 10). Any part consists of fourteen parts. 24 Whole stator was made from 42 parts. This number of parts made production and assembly lengthy. It also makes difficult to fix or exchange stator. For these reasons, we were looking for the possibility of simplifying the production technology. For simplified design was chosen solution which integrates the three essential elements of the original stator to the one elements. Stator segment is composed of the stator blade and part of the inner and outer ring. Stator assembled from these segments can be glued directly to the airplane structure. If there is need to replace the entire stator, it can provide seating face. Seating face can be made in separate mould. For assembling segments there are shaped lock on edge of rings (see figure 11). Composite mould For making mould of this difficult shape of rotor segment was made master model with rapid prototyping method (see figure 12). The stator three-part mould was made by contact lamination on positive master model. The materials of the mould are glass fabric and basic epoxy resin (see figure 13). The three-part mould allowed a manufacturing of this difficult shape of rotor segment. Dividing of the mould was designed for easy product removal from the mould. Making of the stator is provided by contact lamination in negative forms. For manufacturing was selected epoxy resin MGS L285 with hardener L285. Individual parts of stator are formed by glass fiber surface layer. Firstly is made blade part of the stator, then are assembled individuals parts of mould and finish glass fiber surface layer (see figure 14). For joining blade parts was used laminating epoxy with cotton flake filler as is usual in practices. Weight of stator segment is 65g. Curing was with the increased temperature with slow start-up. Total time curing cycle with the slow start-up to 40 °C was 5.5 hours. Net time required to produce one segment stator is 1.5 hours, total production of one segment takes 8 hours. Simple preparation was made where are individual segments of stator placed together. For easy placing is the preparation equipped by hinges. Bushings for hinges in ring parts of segment are in them already from production in forms (see figure 15). 25 L e t e c k ý z p r avo da j 1/2011 up the production and reduce production costs. The designed manufacturing process of stator and its composition was successfully verified. Designed technology reduced the time required for stator production about 30%. Weight savings of over 18% and reduced difficulty of manufacture. 3 Conclusion The right manufacturing process leading to elimination air voids from surface of the blade was found, thereby quality progression of blade surface was achieved during prototyp production. Production productivity was satisfactory. Five blades have been manufactured with one set of molds in ten days, thus one rotor blade in two days. Design concept and selected materials are suited to serial production. The quality of three-part mold surface was preserved after curing of six specimens as well as moulds for cores. For the next development is supposed adjustment and simplification of mold sealing system. In term of serial production more suitable mold sealing system could speed References [1] Barák K.: Sendvičová kompozitní lopatka ventilátoru pro pohonnou jednotku malých letadel; Sborník TRANSFER, č. 11/2010, 2010, p. 50-57 [2] Martaus F.: Aplikace technologie RTM při vývoji kompozitního vrtulového listu; Sborník TRANSFER, č. 5/2007, 2007, p. 5-12 [3] Barák K., Malásek T., Brabec J.: Ducted Fan Power Unit Demonstrator for Ultralight Airplanes — Part Three; Czech AEROSPACE Proceedings, No.3/2008, 2008, p. 7-9 [4] Růžička, P. Ducted Fan Power Unit Demonstrátor for Ultra Ligh tweight Airplanes. Czech Aerospace Proceedings, Letecký zpravodaj. November 2007, No3/2007, p. 49-52. ISSN 1211-877x Microaccelerometer Error Analysis Rozbor chyb mikroakcelerometru Ing. Milan Chvojka Completed cosmic microaccelerometer cannot be fully tested on ground due to its extreme sensitivity. Its final parameters must be calculated from data measured on discrete blocks. Article describes results of accelerometers blocks measurements and on this basis it estimates the expected random error of the complete device. Article evaluates contributions of separate blocks to the overall random error and sketches out fundamental ways of future improvement. Kosmický mikroakcelerometr nelze otestovat jako celek v pozemních podmínkách. Výsledné parametry musí být určeny výpočtem z dat naměřených na dílčích blocích. Článek popisuje výsledky měření jednotlivých bloků kosmického mikroakcelerometru a na jejich základě odhaduje náhodnou chybu celého přístroje. Porovnává příspěvky jednotlivých bloků k celkové náhodné chybě a nastiňuje nejdůležitější oblasti jeho dalšího vývoje. Keywords: Cosmic microaccelerometer, error estimation, ground testing, sources of errors, random and systematic errors, ways of improvement. Introduction Basic parameter of each measurement device is its relative resolution. It is the ratio of its own noise divided by measurement range, expressed usually in percentage. By explicandum (defined notion) own noise we understand sum of random disturbances of the device output measurements not caused by the change of input measured phenomena through the whole frequency range. Own noise is usually defined as parameter of the Gaussian curve of measurement results when measurement device measures zero (or constant) input value. Two sigma criterion is used in the ESA projects. This criterion covers about 95% of measured values. Some other applications can use different criterion derived from the nature of the purpose. Most of measurements described in this article were done on the blocks or boards prepared for the microaccelerometer version 4, model PFM (abbreviation MAC). Completed device without right and upper wall is visible on the Figure 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 26 Review of tests Figure 1 - View to inside electronics of MAC-04 PFM Model. On the left side there are the electronic boards; on the right side there is the sensor covered by position detectors. Microaccelerometer short description This chapter serves to clarify the terms used in this article. Microaccelerometer block diagram is visible on the Figure 2. Microaccelerometer is based on the glass cubic proof mass placed in the glass cubic cavity with six pairs of electrodes. One pair is placed on each wall. The proof mass is kept by the electrostatic servomechanism in the centre of the cavity. Block diagram shows the fundamental functional blocks. Mechanical details except the sensor are not included. But their behaviour, especially temperature deformations, has significant influence to the microaccelerometer parameters. Output voltages from the position detectors are marked P11, P12, P21, P22, P31 and P32. They are proportional to the deviation of the cubic proof mass from the geometrical centre of the sensor cavity. Position control voltages, which are applied to the cavity electrodes are marked A11, A12, A21, A22, A31 and A32. These control voltages are proportional to the external acceleration which acts to the cavity. It is due to the fact that electrodes area, cube mass and gap length between the electrodes and the cubic proof mass are practically constant. Deviations of the proof mass cube for the centre of the cavity are less than 1% of the gap length when feedback loop is working. Cubic sensor has six degree of freedom that is why the control loops are six. This sensor doesn’t measure the gravitational acceleration, because this acceleration acts with the same intensity to the proof mass and also to the cavity and all others parts of the device. Figure 2 - Microaccelerometer Block Diagram All tests described in this article are concentrated to the estimation of parameters of the microaccelerometer. From the sensitivity analyses we know that the biggest influence to the low frequency noise have position detectors POSDETs, based on the capacitive principle. From the previous projects we also know that these position detectors have two major external sources of disturbances. Temperature is the first and about second one the big discussion was several times. Part of experts had hypothesis that this external influence is humidity, part of them had hypothesis that this part of bias is caused by electronics long term instability, without external influence. Analysis of used components and sensitivity analysis of schematics of the POSDETs result in the most probable hypothesis that this second major influence is the humidity. There fore the combined test was realized. This test confirms this hypothesis. Results of this test are plotted on the Graph 1. Position detectors POSDETs needs except supply voltages also two auxiliary voltages Sin230k and PLDCLK for their function. Auxiliary POSDET testing device POSDET Testing Jig is necessary for their testing. This POSDET Testing Jig (abbreviation PTJ) includes also two preamplifiers for the amplification of two POSDET output voltages. These amplifiers have amplification factor 18x. The output voltages of POSDET are eighteens smaller than the values plotted in the Graph 1 Graph 1 - Amplified output voltage of position detector POSDET after humidity change from 70% to 20% (at 5100 sec. time) and after temperature change from 25°C to 30°C (at 8700 sec. time) Result of this test significantly changes the view of management staff to the way of next tests. Commonly was accepted idea, that precise tests of microaccelerometer (MAC) electronics boards, especially position detectors POSDETs, must be realized in vacuum conditions. After this initial test there were realized these following tests: Test of the long term stability of POSDET, test of the temperature bias of POSDET, test of the own noise of A/D converters and their supportive circuits, test of the temperature dependence of A/D converters and their auxiliary circuits, test of the influence of temperature changes of the distance between POSDET boards and MAC cover, test of 27 L e t e c k ý z p r avo da j Relation between POSDET and complete sensor error Position detector POSDET has two outputs for each axe: Translation output voltage proportional to the translation deviation of the cubic proof mass from the central position and rotational output voltage proportional to the rotational deviation of the cube. If the translational POSDET output voltage has error then complete sensor has error which is expressed by the Equation 1. Equation 1 Voltage (V) Where: C m Pv0.04 D 0.02 0 Lps -0.02 Thermo-vacuum test was realized, with temperature time profile simulating temperature cycling of microaccelerometer on the satellite during several orbits. POSDET board was equipped by the four temperature sensors T5, T6, T7 and T8 in the corners. Temperature time profile measured by these sensors is plotted on the Graph 2. The POSDET board outputs were trimmed to zero before this test. Translation and rotation output voltages of the POSDET board during this test are plotted on the Graph 3. POSDET temperature sensors time profile 50 40 30 Temperature (deg C) the influence of temperature dependence of Sin230kHz auxiliary sine voltage to the POSDET output voltages and test of the influence of the temperature dependence of the phase between PLDCLK synchronization voltage and the Sin230kHz sine voltage to the POSDET outputs. All these tests checks external influences to the microaccelerometer systematic and random errors.. is the absolute measurement error in translation channel (in ms-2) is the capacitance between one electrode nad the cube (in pF) Linear and rotational channel voltages is the mass of cubic proof mass (in kg) is the polarization voltage (in V) is the distance between electrode and the cube wall in central position (in m) is the offset error of the POLDET position detector in transl. channel (in V) is the sensitivity of the POSDET in translation channel (in m/V) Where r Lpr Jcube C,Pv,D 10 T5, top left T6, top right T7, bottom left T8, bottom right -10 -20 22/05/2007 12:00:00 23/05/2007 00:00:00 23/05/2007 12:00:00 24/05/2007 00:00:00 24/05/2007 12:00:00 25/05/2007 00:00:00 25/05/2007 12:00:00 26/05/2007 00:00:00 Date and time Graph 2 - Temperature time profile during the POSDET thermo-vacuum test Rotational channel Linear channel Linear and rotational channel voltages 0.04 If the rotational POSDET output voltage has error -0.06 UDRot then the complete sensor has error which is expressed by Equation 2. 23/05/2007 00:00:00 23/05/2007 12:00:00 24/05/2007 00:00:00 24/05/2007 12:00:00 25/05/2007 00:00:00 25/05/2007 12:00:00 Date and Time is the absolute measurement error in rotation channel (in rad.s-2) is arm of force (in m) r = 7.294 52.10-3 m is the sensitivity of the POSDET in rotation channel (in rad/V) = 1.370 89 . 10-3 rad/V is the moment of inertia of the proof mass cube is the offset error of the POLDET position detector in rotation channel (in V) symbols indicates the same physical quantity as in the previous Equation Results of ground tests Initiatory tests were concentrated to the POSDET parameters checking. For POSDET board testing was designed electronic auxiliary device named POSDET Testing Jig (abbreviation PTJ) and mechanical testing jig simulating the mechanics of microaccelerometer sensor and its cover. 0.02 0 26/05/2007 00:00:00 Voltage (V) Equation 2 20 0 -0.04 -0.08 22/05/2007 12:00:00 1/2011 Rotational chann Linear channel -0.02 -0.04 -0.06 -0.08 22/05/2007 12:00:00 23/05/2007 00:00:00 23/05/2007 12:00:00 24/05/2007 00:00:00 24/05/2007 12:00:00 25/05/2007 00:00:00 25/05/2007 12:00:00 26/05/2007 00:00:00 Date and Time Graph 3 - POSDET output voltag during the thermo-vacuum test Significant are the jumps in both channels during lowest temperature cycling. Origins of these steps are the rigid connections of the capacitor trimmers to the printed circuit board (PCB). In next version of POSDET there were these rigid connections replaced by the soft elastic ones. Graph 4 shows dependence between temperature and POSDET translation output voltage and Graph 5 shows this dependence for the rotational output of POSDET. C z e c h A e r o s pa c e P r o c e e d i n g s 28 Significant are the jumps in both channels during lowest temperature cycling. Origins of these steps are the rigid connections of the capacitor trimmers to the printed circuit board (PCB). In next version of POSDET there were these rigid connections replaced by the soft elastic ones. Graph 4 shows dependence between temperature and POSDET translation output voltage and Graph 5 shows this dependence for the rotational output of POSDET. Graph 6 - Temperature dependence of auxiliary sinus voltage Sin230kHz. Red curve comes from MAC-03. Blue curve: Improved generator on MAC-04. Green line is the linear approximation. Graph 4 - Translational POSDET output voltage vs. temperature during temperature cycling Graph 5 - Rotational POSDET output voltage vs. temperature during temperature cycling Graph 7 - Temperature dependence of phase between PLDCLK and Sin230kHz on MAC-04 microaccelerometer model. Outputs from PTJ on the left axes are recalculated to the POSDETs output voltages (divided by factor 18). On the Graph 4 we see the phase shift between temperature cycling and the translational output of the POSDET. Its cause is the fact that translation measurement has origin in the opposite electrodes, which have significant distance. This fact causes in significant temperature gradient between them. Rotational measurement is derived from adjacent electrodes, which are close each to other. In this case the phase shift is significantly reduced. Temperature biases of both POSDET outputs derived from the Graphs 4 and 5 are: Another source of microaccelerometer error is long term instability of POSDET outputs. In the ground conditions is very difficult collect some representative data. Graph 8 shows results of the measurements, where external terrestrial influences were relatively highly reduced. Fluctuation of translation channel is about and in rotation channel about during 14 hours. Adequate errors in acceleration channels calculated according Equation 1 and 2 are: Temperature bias of translation output is: Temperature bias of translation output is: Temperature bias of rotational output is: Temperature bias of rotational output is: 29 L e t e c k ý z p r avo da j POSDET Output Voltage Time Stability 4 3 POSDET Output Voltage [mV] 2 1 0 0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000 -1 -2 -3 -4 POSDET Output Voltage Time Stability -5 4 3 On the Graph 10 there are plotted outputs of A/D converters when the input voltages and external conditions are constant and only own noise of A/D converters and their auxiliary circuits is significant. Statistical analysis of the data plotted on the Graph 10 shows that dispersion of the translation data recalculated to the voltage A11 is and for the rotational data expressed as voltage A12 Sensitivities of microaccelerometer with respect to input voltages of A/D converters in translation and rotation channels are: for translation channel and Time [sec] Tran Output Rotat Output for rotation channel. POSDET Output Voltage [mV] Graph 28: Long term outputs of POSDET board with stabilized input and external conditions 1 1/2011 Graph 8 - Long term outputs of POSDET board with stabilized input and external conditions Latest significant source of microaccelerometer disturbances is the own noise of A/D converters with Resolution is usually defined as dispersion of summa0 their auxiliary the random part temperature dependence of45000 these circuits. A/D 0 circuits 5000 and 10000 15000 20000of the 25000 30000 35000 40000 50000 rized noise. Then own noise of A/D converters converted to converters -1Latest are the Sigma-Delta types. These types have output word 24 bitesdisturbances long, but lowest 4 bites significant source of microaccelerometer are noisy, resolution is about 20 bits. Full measurement range of the converters is from 0.0 to +5.0the V. adequate acceleration is: is -2the own noise of A/D converters with their auxiliary circuits Graph 9 represents the response of A/D converters together with their auxiliary circuits to the and-3 the part thewith temperature dependence of these cirtemperature jumprandom from 30.8°C to of 11.4°C constant input voltage P11=P12=+0.2V. P11 is the for translation channel and translational output of the POSDET measuring X axes. P12 is its rotational output. A11 is the output cuits. A/D converters are the Sigma-Delta types. These types -4 of translational output of the X axes position control circuit. A12 is rotational output of it. for rotation channel. have output word 24 bites long, but lowest 4 bites are noisy, resolution is about 20 bits. Full measurement range of the converters is term fromoutputs 0.0 toof +5.0 V.board Graph 9 represents theexternal response of Temperature dependences of A/D converters including their Graph 8: Long POSDET with stabilized input and conditions A/D converters together with their auxiliary circuits to the tem- auxiliary circuits derived from data plotted on the Graph 9 Latestperature significant source microaccelerometer disturbances is the own noiseinput of A/D voltage converters with and expressed as adequate accelerations are: jumpoffrom 30.8°C to 11.4°C with constant their auxiliary circuits and the random part of the temperature dependence of these circuits. A/D P11=P12=+0.2V. P11 is thetypes translational output of the converters are the Sigma-Delta types. These have output word 24 bites long,POSDET but lowest 4 bites are noisy, resolution is 20 P12 bits. Full rangeoutput. of the converters fromoutput 0.0 to +5.0 V. measuring Xabout axes. is measurement its rotational A11 isisthe Graph 9 represents the response of A/D converters together with their auxiliary circuits to the of translational output of the Xconstant axes position control circuit. A12 temperature jump from 30.8°C to 11.4°C with input voltage P11=P12=+0.2V. P11 is the for translation channel and translational output of the POSDET measuring X axes. P12 is its rotational output. A11 is the output is rotational output of it. -5 Time [sec] Tran Output Rotat Output of translational output of the X axes position control circuit. A12 is rotational output of it. for rotation channel. Graph 9: Response of A/D converters outputs to the temperature jump with constant input voltage. Temperature dependence is systematic error and can be partly reduced by calibration of this dependence and its mathematical extraction. Summarized microaccelerometer error Each device has random and systematic errors. Each systematic error, for example like temperature dependence, has deterministic and random part. For estimation of the random part of all temperature dependencies it is necessary to know what the error of the temperature measurements is Graph 9:Graph Response 9 of -A/D converters outputs to the temperature jump with constant Response of A/D converters outputs to input and how much percent of systematic error is its random voltage.constant input voltage. the temperature jump with part. These values depend on concrete conditions in each project. Here we will suppose that random part is 30% of systematic error and the random error of temperature measurements is = 0.5°C. Under these conditions the most significant accelerometer errors are: Random errors: 1. Long term instability of POSDETs 2. Own noise of A/D converters and their auxiliary circuits Systematic errors: 3. POSDETs outputs Temperature bias 4. Influence of temperature depen- dence of Sin230k amplitude to POSDETs outputs 5. Influence of temperature depen- dence of phase between Sin230k and PLDCLK to POSDETs outputs Graph 10: Noise of10 A/D-converters outputs constant input voltage and temperature. 6. Temperature dependence of A/D Graph Noise of A/Dwithin converters outputs within On the Graph 10 there are plotted outputs of A/D converters when the input voltages and external converters and their auxiliary circuits constant input voltage and temperature. conditions are constant and only own noise of A/D converters and their auxiliary circuits is significant. Statistical analysis of the data plotted on the Graph 10 shows that dispersion of the translation data recalculated to the voltage A11 is = 5.29 V and for the rotational data expressed as voltage A12 is = 4.21 V. Sensitivities of microaccelerometer with respect to input voltages of A/D converters in translation and rotation channels are: 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 Estimation of overall random error of the complete microaccelerometer based on the results of tests of basic microaccelerometer blocks and taking into account the conditions described in the beginning of this chapter is summarized in the Table 1. development of electronic circuits.converters In the microaccelerometer data we also find some few Outputs of A/D duringoutput thermo-vacuum small jumps. See Graph 11. Big jumps like ones on the Graph 3 were steamed off by the redesign of tests trimmer Smallconnection. jumps Very in small output caused the also flash capacitive jumps,data, visible on the Graphby 11, have origin in the behaviour of the mechanics. That is the reason, why it is necessary to redesign partly also the mechanical tension releasing mechanics of the microaccelerometer, especially fixing elements of the glass sensor inside the cover. Outputs of A/D converters during thermo-vacuum tests Small jumps in output data, caused by the flash mechanical tension releasing Error No. 2 Errors expressed in Volts Systematic Errors expressed in acceleration 2 Random errors expressed in acceleration 1 Long Term POSDET stability Translation 20 V --- 5.732 . 10-10 m.s-2 1 Long Term POSDET stability Rotation 70 V --- 1.002 . 10-7 rad.s-2 2 Own Noise of A/D Converters Translation 10.58V --- 1.102 . 10-9 m.s-2 2 Own Noise of A/D Converters Rotation 8.42 V --- 4.383 . 10-8 rad.s-2 3 POSDET Output Temperature Bias Translation 376V/K 1.078 . 10-8 m.s-2/K 1.617 . 10-9 m.s-2 3 POSDET Output Temperature Bias Rotation 686V/K 9.821 . 10-7 rad.s-2/K 1.473 . 10-7 rad.s-2 4 Influence of Temperature Dependence Amplitude of Sin230kHz Translation 46.39 V/K 1.330 . 10-9 m.s-2/K 1.995 . 10-10 m.s-2 4 Influence of Temperature Dependence Amplitude of Sin230kHz Rotation Negligible --- --- 5 Influence of Temperature Dependence of Phase between Sin230k and PLDCLK Translation 36.5V/K 1.046 . 10 m.s /K 5 Influence of Temperature Dependence of Phase between Sin230k and PLDCLK Rotation 36.5V/K -8 6 Temperature Dependence of A/D converters Tra 417V/K 4.345 . 10-8 m.s-2/K 6.518 . 10-9 m.s-2 6 Temperature Dependence of A/D converters Rot 422V/K 2.197 . 10-6 rad.s-2/K 3.296 . 10-7 rad.s-2 -9 -2 1.569 . 10-10 m.s-2 -2 -9 5.223 . 10 rad.s /K 7.835 . 10 rad.s -2 Overall Random Error of Translation 1.017 . 10-8 m.s-2 Overall Random Error of Rotation 6.288 . 10-7 rad.s-2 Graph 11: Outputs of A/D converters during thermo-vacuum tests (variable temperature) Graph 11 - Outputs of A/D converters during thermo-vacuum tests All the fixing elements of the glass sensor must(variable be statically temperature) determinate or elastic with permanent contact. Fulfilment of these two conditions significantly reduces the internal forces in the mechanics caused by temperature changes and gradients. Conclusion All the fixing elements of the glass sensor must be statically Initial estimation of 2random error for microaccelerometer version 4 in translation channel was determinate or elastic with permanent contact. Fulfilment of 1.364 . 10-8 m.s-2. See [2]. Calculation based on the test results of realized blocks gave value these significantly reduces the-2 for internal forces 1.017 . 10-8two m.s-2 conditions for the translation channels and 6.288 . 10-7 rad.s the rotation channels. This The estimation given in [2] of the translation channel error made at the beginning of the shows that anticipated results were reached. On the other hand these tests disclosed several microaccelerometer version 4 design was that 2error will be 1.364 . 10 m.s . Recent result, based results in the mechanics caused by temperature changes and gradion the test of developed and realized aboutof 25% the less than initial estimation. From this it small insufficiencies, especially influence of mechanical construction to electronics. This new The estimation givenblocks, in is[2] translation channel experience ents. show the way of future redesign. follows that parameters of newly developed blocks satisfy established goal. error made at the beginning of the microaccelerometer ver-8 Main errors reduction future design m.s-2. References sion 4sources designof was that and their error will be in 1.364 . 10 Conclusion From the Table 1 results, that most significant are two sources of errors: Instability and temperature [1] Chvojka,estimation M. T.: Analýza zdrojů nestabilit kosmického mikroakcelerometru; Recent result, based on the test of developed and realized Initial of chyb a random error for microaccele-Výzkumná dependence of position detector POSDET and own noise and noise and temperature dependence of zpráva R-4862, VZLÚ, a.s., Praha – Letňany 2010 A/D converters and their auxiliary circuits. There are two most significant ways of future blocks, is about 25% less than initial estimation. From this rometer version 4 in translation channel was 1.364 . 10-8 development of electronic circuits. In the microaccelerometer output data we also find some few -2 it follows of newly developed blocks . SeeM.[2]. Calculation based test results of reali-VZLÚ, a.s., small jumps. See that Graph parameters 11. Big jumps like ones on the Graph 3 were steamed off by thesatisfy redesign of [2]m.sChvojka, T.: ACC Performance Analysis Ver.on 2.6;the Výzkumná zpráva VZL-AC-0003, capacitive trimmer connection. Very small jumps, visible on the Graph 11, have also origin in the -2 established goal. Praha – Letňany 2009 for the translation zed blocks gave value 1.017 . 10-8 m.s behaviour of the mechanics. That is the reason, why it is necessary to redesign partly also the mechanics of the microaccelerometer, especially fixing elements of the glass sensor inside the cover. channels and 6.288 . 10-7 rad.s-2 for the rotation channels. Main sources of errors and their reduction in future This results shows that anticipated results were reached. On Outputs of A/D converters during thermo-vacuum tests design the other hand these tests disclosed several small insufficiSmall jumps in output data, caused by the flash mechanical tension releasing From the Table 1 results, that most significant are two encies, especially influence of mechanical construction to sources of errors: Instability and temperature dependence electronics. This new experience show the way of future of position detector POSDET and own noise and noise and redesign. temperature dependence of A/D converters and their auxiliary circuits. There are two most significant ways of future development of electronic circuits. In the microaccelerometer output data we also find some few small jumps. See References Graph 11. Big jumps like ones on the Graph 3 were stea[1] Chvojka, M. T.: Analýza zdrojů chyb a nestabilit kosmického med off by the redesign of capacitive trimmer connection. mikroakcelerometru; Výzkumná zpráva R-4862, VZLÚ, a.s., Praha – Letňany 2010 Very small jumps, visible on the Graph 11, have also origin in the behaviour of the mechanics. That is the reason, why [2] Chvojka, M. T.: ACC Performance Analysis Ver. 2.6; Výzkumná it is necessary to redesign partly also the mechanics of the zpráva VZL-AC-0003, VZLÚ, a.s., Praha – Letňany 2009 microaccelerometer, especially fixing elements of the glass sensor inside the cover. Table 1 - Estimation of Overall Random Table 1: Estimation of Overall Random Microaccelerometer Error Microaccelerometer Error -8 -2 Intentionally left void. 31 L e t e c k ý z p r avo da j 1/2011 Analysis of dynamic characteristics of gas turbine Analýza dynamických vlastností spalovací turbíny Tomáš Jamróz, Jiří Had, Vladimír Dániel / VZLÚ, Plc., Praha The research conducted explains the dynamic analysis of a generator rotor that is loaded and its relation to stress under excessive vibration. The generator rotor consists of a gas turbine, compressor, and steam turbine. A connection of the compressor and of the steam turbine lead to a structural modification to the original construction. In the article we are shown basic calculations of dynamical properties of a new construction of the generator rotor (in particular to critical speed and its speed dependance, and its sensitivity analysis for unbalance). Článek se zaobírá dynamikou rotorového soustrojí, které je namáháno nadměrnými vibracemi. Rotorové soustrojí se skládá z plynové turbíny, kompresoru a parní turbíny. Spojení kompresoru a parní turbíny vedlo ke konstrukční změně původního návrhu. V článku jsou uvedeny základní výpočty dynamických vlastností konstrukčně upravené soustavy (kritické otáčky a jejich závislost na otáčkách a citlivostní analýza na nevývahu). Keywords: turbomachinery; critical speed; unbalance Fig. 1 - 3D projection of FEM model of the Generator rotor Introduction Generator rotor is running at operating speed of 7650 min-1. The Generator rotor consists of a gas turbine, a compressor, a steam turbine, a gearbox and a generator. Dynamic model of this system involves only the gas turbines, the compressors and the steam turbine part. Above mentioned parts of the generator rotor for the dynamic analysis are supported by three sliding bearings. Radial bearing of gas turbine (L6), radial bearing of compressor (L5) and the combined radial-axial (L4r-L4a) bearing of compressor. Steam turbine is supported by two radial and one axial bearing (L3 - combined and L2- radial). The parts are connected by coupling. The gas turbine and the compressor is connected by a cone ring coupling and between the compressor and the stream turbine is used a gear coupling. The rotor system is loaded by an excessive vibration especially in position L5. Dynamic calculation of rotor system was performed to find out whether the system does not work near critical speed and to find the cause of excessive vibration. Model For the purposes of the dynamic characteristics calculation in the software MSC.Nastran is necessary the Generator rotor is discredited into finite elements. The model consists of beam elements modeling the shaft. Disk of gas turbine is done by shell elements. Blades of compressor was not modeled in details but was included in the model of the compressor rotor. Hubs coupling at the compressor rotor were done as point masses with the mass and inertia effects. Stiffness and mass was modelled separately. The stiffness of the rotor was done by the beam elements. Mass points distributed along the rotor system define the mass and inertia effects of the each parts of the rotor system. Compressor rotor and rotor of the Gas turbine are connected by beam elements and rigid elements (RBE2) that replace screw connection. The beam elements transmit axial moment. This simplification replaces of a Cone ring coupling. Compressor rotor and rotor of the Steam turbine are connected only rigid element (RBE2) that replaces function of a Gear coupling. Rotor is supported on four points which are numbered L3-L6. The supports are shown in Figure 1. The shaft supports (L6, L5, L4r and L3r) enable free motion with exception of radial displacement. The bearings are modelled with 1D elements with asymmetric stiffness properties. C z e c h A e r o s pa c e P r o c e e d i n g s Calculation Critical speed Dynamic properties of the bearing depend on external parameters, namely angular velocity of the rotor ω . That’s why is necessary to perform calculations of dynamic properties in entire operating range of speeds. The result of calculation of the natural frequencies is sets of complex eigenvalues that were calculated for various operating speed. Each complex eigenvalue consists of imaginary part Ωm (represents the natural frequency of rotating shaft with angular velocity ω ) and the real part δm (represents a damping) 32 stability, ie. for all ω > ωk. The values of critical speed and the logarithmic decrement are shown in table 1. The damping is described by a logarithmic decrement To evaluate of danger of the critical speeds was used a methodology according to the American standard API. There is defined an Amplification Factor, which is proportional to the theoretical maximum value of excess vibrations in the critical speed. The value of AF is defined by its own decay and natural frequencies By connecting corresponding natural frequencies Ωm computed for various shaft speed ω get curves expressing dependence of natural frequencies on the speed of the shaft ω called Campbell diagram. Critical speeds ωk of the rotor are speeds at which natural frequencies curves Ωm( ω ) cross the rotation frequency curve Ω( ω ). Table 1 - Critical speed of the rotor and the Amplification factor If the value of AF in critical speed meets the requirement 0<AF<2.5, according to the standard the critical speed ω k is over damped and it is not considered as critical speed. Between 2.5<AF<5.0 the critical speed is very attenuate and for AF>5.0 is critical speed less attenuate. Operating rotor speed is 127.5Hz. This value is between two critical speeds with very less attenuation. Fig. 2 - Campbell diagram One of the important information needed to assess of danger of critical speed is the value of decay δm ( ω k) in the critical speed. The critical speed will be stable if decay is δm (ω k) > 0. The results of the calculation of natural frequencies show that the first and third critical speed are unstable ie. δm ( ω k)<0. In these cases the most important influence on stability has bearings. The characteristics of sliding bearing show that the rotor motion will be unstable for all the speed limits of Fig. 3 - Amplification factor 33 L e t e c k ý z p r avo da j 1/2011 Fig. 4 - Natural frequencies modes of rotor shapes of 1th, 3th, 5th and 7th. Sensitivity analysis for unbalance Several planes were chosen to define of unbalance. The first two planes were defined at faces of the compressor and the other two were placed in coupling. Due to the operation mode shape only planes with couplings has been tested. In a hub of the Gear coupling on the compressor rotor was added a mass 0.08kg, 0.16kg and 0.24kg of 87.5mm radius. These unbalances gradually formed torques 7E-3kgm, 14E-3kgm and 24E-3kgm. that was defined above. The Figure 5 shows that unbalance is only sensitive to the critical speed of the Gear coupling around 8500[min-1]. The sensitivity analysis of the unbalance in the Cone ring coupling was solved as well as in case of the sensitivity analysis for the Gear coupling. In a hub of the Cone ring coupling on the rotor of the Gas turbine was added a mass 0.16kg, 0.24kg and 0.32kg of 110mm radius. These unbalances gradually formed torques 17E-3kgm, 26E-3kgm and 35E-3kgm. Fig. 5 - Sensitivity to unbalance on the Gear coupling Fig. 6 - Sensitivity to unbalance on the Cone ring coupling In the Figure 5 is shown a significant increase of the vibration in natural frequencies without mass point for an unbalance. The second part draws sensitivity to unbalance In the Figure 6 is shown a significant increase of the vibration in natural frequencies without mass point for an unbalance. The second part draws a sensitivity to unbalance C z e c h A e r o s pa c e P r o c e e d i n g s that was defined above. The Figure 6 shows that unbalance is only sensitive to the critical speed of the Cone ring coupling and then it is quickly damped. See to table 1. mode 5. Conclusion Dynamic analysis of rotor system unfortunately did not identify the exact cause of an excessive vibration in the bearing housings. According to analysis results to find the critical speed, it was found that rotor system is working in proximity to one of the natural frequencies. This critical shape has an important damping ratio and thus it is not cause excessive vibration in bearing housing. The excessive vibration could be caused by instability of the first two natural frequencies. In future measurements will be made to prove or disprove whether the issue of oil whip or oil whirl. Another cause of excessive vibrations could be unbalance in couplings. An influence of the Gear coupling on vibration can reduce by a structural adjustment by reducing 34 the shaft end. There could be another reason of the excessive vibration and that is coaxial support of compressor rotor and the Steam turbine rotor at a standstill. This situation would cause misalignment of rotors in support of single parts at the heat state of generator rotor. An effect of misalignment of support will be dealt with in the future. References [1] Brepta, R., P;st, L., Turek, F.: Mechanické kmitání, Sobotáles, Praha, 1994. [2] Slavík, J., Stejskal, V., Zeman, V.: Základy dynamiky strojů, ČVÚT, Praha, 1997. [3] Krämer, E.: Dynamics of Rotors and Foundtions, Springer Verlag, Berlin, 1993. [4] Muszynska, A.: Rotordynamics, Taylor & Francis Group, 2005. Kinetic analysis of oxidizing decomposition of carbon fiber reinforced epoxy composite Kinetická analýza oxidačního rozkladu uhlíkového epoxidového kompozitu Ing. Zdeněk Mašek The kinetic oxidizition and decomposition epoxy composite was studied. Temperature decomposition was studied on a laboratory sample at the temperature upto 550oC in an oxidized atmosphere. The kinetic-diffusion model was proposed, following the measured values. The shape of thin composite plates of the sample was calculated. Byla studována kinetika oxidačního rozkladu epoxydového kompozitu. Tepelný rozklad byl studován na laboratorním vzorku při teplotách do 550 0C. v oxidační atmosféře. Byl navržen kineticko-difúzní model na základě experimentálních údajů. Model počítá s tvarem tenkých kompozitních desek. Keywords: Epoxy composite, carbon fiber, thermo gravimetric analysis (TGA), kinetic equation, activation energy. 1 Introduction The tempering of carbon composites concludes to exhausted gases evaporation. The production of exhausted gases is followed by mass decreasing of combusted material. The description of mass decreasing caused by heating is the object of thermo gravimetric analysis (TGA). In this work a TGA analysis of carbon fiber reinforced sample of epoxy composite was done. The sample was heated in the air region by the controlled temperature trend. The measured values aimed were analyzed at the base of reaction kinetic. During the thermal decomposition the production of basic combusted gases was observed. During the heating the temperature directly on the surface of the sample was measured. The measurement of temperature on the surface and of the hot air can detect the exothermic reaction appearance. The values measured were compared with equations of kinetic and diffusion processes proposed. The first rough process estimate involved a description with correlation according to a speed equation of Prout-Tompkins : d dt = . The whole process was divided into 35 several parts and the parameters of activation energy, pre-exponential factor and exponents of velocity equation was calculated for each part. The next approximation involved a three-step model including a kinetic of pyrolytic decomposition in the first part, surface oxidizing reaction and a diffusion in the second part and the oxidizing reaction in the third part. 2 Experimental arrangement 2.1 Processing material The tempering of carbon composites concludes to exhausted gases evaporation. The production of exhausted gases is followed by mass decreasing of combusted material. The description of mass decreasing caused by heating Epoxy composite plates of weight around 200 g and with thickness of 3 – 5 mm were tested. There is a difference compared with the classical TGA, which goes with several gram-weight samples. The plates were manufactured by RTM technology and fibre reinforcement was aligned in equal direction. The weight ratio of carbon fibers was about 50 %. 2.2 Experimental equipment Experimental equipment developed for composite degradation was used. The reacting chamber has a volume of 240 dm3 and is heated by electrical elements. The chamber is not hermetically closed and small negative pressure is kept inside by sucking of exhausted gasses. Sucking of exhausted gasses enables to keep an oxidizing atmosphere inside the reacting chamber. The program for temperature control is adjusted with regards to optimal yield and quality of processed carbon fibers composites. The curve of the temperature increase is shown on the graph 1. The speed of temperature increase is not constant. The temperature is monitored by thermo electrical thermometer fixed closely above the surface. The plate is put at the iron grind in the horizontal level. 2.3 Experimental procedure Thermogravimetric analysis (TGA) is a method often used for studying of thermal behaviour of materials. The material is heated and a mass decreasing is plotted with temperature changing depending on the time. Kinetic parameters of thermal degradation are determined from the measured values. 2.4 Analysis of results Data files of temperature trends and mass decreasing were taken and converted into table processor Excel. All numeric calculations and graphs were provided by this means. 3 Theoretical background 3.1 The velocity equation It is obvious that the chemical kinetic equations are written in this form: Velocity of conversion = Velocity constant . velocity function L e t e c k ý z p r avo da j • • • 1/2011 Velocity of conversion: d /dt is a time derivative of extent of process that is a number between 0 and 1. Velocity constant K=A.exp(E/RT) is Arrhenius equation which express the dependence of velocity constant on the temperature. A is called a pre-exponential factor, E is a activation energy, R is universal gas constant and T is absolute temperature in Kelvin. Velocity function f( ) is a function of the extent of process its shape is derived from the reacting mechanism. Many authors referred about expressing of kinetic parameters and several methods were described. The experimental methods are obviously based on the linear temperature increasing because the linear temperature dependence makes the counting easier. There are a integral and a differential approach how to solve the kinetic equations. The integral approach provides the kinetic parameters for all reach of checked values, though the differential approach can get the kinetic parameters from the one point of gravimetric analysis. The integral approach is more robust, gives more reliable values and is more protected against the influence of distant values. The differential approach is more sensitive to the distant values, but can detect the changes of reacting mechanism at specified temperature reach.1 In this paper is discussed about using of kinetic equation d /dt = A(1- )n m. This equation gives a good condition for process description. There was referred in papers2,3 about suitability of this equation for description of epoxy system degradation. 3.2 The regression task The essence of the regression task is finding of function (t) which fits to measured vales mostly. The regression task concludes to finding of the minimum of optimizing function. The optimizing function is obviously defined as a summa of squares of differences taken from measured values and values of the optimizing function. The minimum of optimizing function is looked for 4 variables: Apre-exponential factor, E-activation energy of degradation reaction, n,m – parameters of the velocity function. 3.3 Diffusion process The kinetic equations are obviously used for homogeneous reaction description which go through all bulk of reacting site and are independent of diffusion processes. The degradation process contrary runs on the specific defined surfaces and is slowed by transporting processes. In the three-steps model given a diffusion processes are calculated and some simplifying presumptions are used. Diffusion processes are described by Fick laws. 4 Kinetic model of thermal decomposition On the beginning the one-step model of decomposition was solved as a regression task with searching of parameters of the equation. The measured values of extent of reaction were numerically smoothed with using of regress polynomials of order 3. For this intention the measured values of TGA were divided into several periods and each period was smoothed by regression poly- C z e c h A e r o s pa c e P r o c e e d i n g s 36 nomial yi = ati2 + bti + c, where (ti, yi) is smoothed value yi at the time ti. As a extent of conversion value was taken value mi/M, where mi is actual mass of the sample at the time ti and M in initial mass of the sample. For numerical calculations was the equation set up to make numerical calculation easier: ted parameters for each part are shown in the figure1. TheDegree process of degradation wasfrom divided into adecomposition sequenceof Graph1 of conversion and temperature curve thermal oxidative epoxy resin of parts according to the shape of the conversion curve. The The process of degradation was divided a sequence according to theThe shape of the regression was done in theintoeach partof parts degradation. calconversion curve. The regression was done in the each part degradation. The calculated culated for each part are shown in the figure1. parametersparameters for each part are shown in the figure1. Temperature reach 0C A(1010 s-1) E activation energy (105J.mol-1) n m 268 – 325 4,20 1,35 5,48 0,50 325 – 347 2,66 1,35 3,59 0,61 347 – 379 2,68 1,35 10 0,87 379 – 409 2,68 1,50 7,91 0,20 409-553 5,90 2,17 0 -3,65 553 5,81 2,21 0 -1 553-502 5,9 2,17 -0,14 -3,26 Figure1 Kinetic parameters for a sequence of temperature intervals Figure 1 - Kinetic parameters for a sequence From the figure1 we can conclude that the model can distinguish at least two different of temperature intervals processes, the first one with the activation energy 135 kJ.mol-1 in the reach of temperatures 268 – 400 0C and the second one with activation energy 217 kJ.mol-1 above the temperature 0 0 400 C. From calculated values of theconclude parameters n,m at the temperature we can From thethefigure1 we can that the model553 canC distin- The optimizing function then gives the shape: Where the summation goes through the selected time interval, yi are values of smoothing polynomial at the time ti and are values of derived smoothing polynomial at the time ti. The function S is now taken as a function of 4 variables S=S(A,E,n,m) and its minimum is searched. The necessary condition for the minimum is zero value of partial derivatives of S function after checked parameters: guish at least two different processes, the first one with the activation energy 135 kJ.mol-1 in the reach of temperatures 268 – 400 0C and the second one with activation energy 217 kJ. mol-1 above the temperature 400 0C. From the calculated values of the parameters n,m at the temperature 553 0C we can conclude that that process is ruled by the diffusion, thus f( ) = -1. This function is obviously referred as a velocity function for the diffusively ruled process4. 5 Three-steps process model 5.1 Model schema The weak point of kinetic model is that that doesn’t operate with diffusion processes. The following text is an attempt to establish a quantitative kinetic process model, The numerical calculation was done by iteration method which would include diffusion. The model is represented where the approximation of the next step was taken the relatiby a plastic plate of composite carbon fiber reinforced on: material. The matrix is a kind of epoxy resin. The plate is put at the grind in the reacting chamber and the reacting chamber is slowly heated. Inside the chamber is very slowly motion of the hot gas. The concentration of the oxyThe numerical calculation was done by iteration method where the approximation of the next gen is approximately equal to the normal oxygen concentstep was taken the relation: S ( x n ) ration. The temperature of the plate is balanced with the x n Where 1 xn S (, x n )are values of optimizing function derived after temperature inside the chamber. The characterizing shape checked parameter. of the plate is the thickness. The exhausted gases go out Where S , S are values of optimizing function derived after checked parameter. from the plate. the thermal is conclude that that processThe is ruledrun by theof diffusion, thus f(-1.decomposition This function is obviously referred as a velocity function for the diffusively ruled process4. described by 3 following processes. 5 Three-steps process model 5.2 Process I – quick pyrolytic decomposition 5.1 Model schema The weak point of kinetic model is that that doesn’tgoes operatesimultaneously with diffusion processes.in Thethe Thermal degradation reaction following text is an attempt to establish a quantitative kinetic process model, which would all bulk of the plate. The combusted gases go out through include diffusion. The model is represented by a plastic plate of composite carbon fiber reinforced material. The matrix is a kind of epoxy resin. The plate is put at the grind in the the surface of the plate and block the acting of oxidative reacting chamber and the reacting chamber is slowly heated. Inside the chamber is very slowly motion of the hot gas. The concentration of the oxygen approximately to the atmosphere. This process is finished by iscreating ofequal porous normal oxygen concentration. The temperature of the plate is balanced 0 with the temperature structure at the temperature around C and the plate inside the chamber. The characterizing shape of the plate is410 the thickness. The exhausted gases go out from the plate. The run of the thermal decomposition is described by 3 following looses around 25% of the initial mass. The carbon fibers processes. are not revealed during this process, they are surrounded 5.2 Process I – quick pyrolytic decomposition degradation reaction goesthe simultaneously the all bulk of the plate. The combusted byThermal a porous char with rests ofinhigh-molecular hydrocargases go out through the surface of the plate and block the acting of oxidative atmosphere. 0 bon compounds. The regression for around this 410 process C and This stable process is finished by creating of porous structure at the task temperature the plate looses around 25% of the initial mass. The carbon fibers are not revealed during this was solved for temperature reach until the temperature process, they are surrounded by a porous char with the rests of high-molecular hydrocarbon Graph1 Degree of conversion and temperature curve from thermal oxidative decomposition of 0compounds. stable The regression task for this process solved at for temperature reach until 410 C. Calculated parameters are was shown the Figure2 0 Graph 1 - Degree of conversion and temperature curve epoxy resin the temperature 410 C. Calculated parameters are shown at the Figure2 from thermal oxidative decomposition of epoxy resin The process of degradation was divided into a sequence of parts according to the shape of the reach conversion curve. The regression was done in the each part degradation. The calculated Temperature 0 C The process of are degradation was divided into a sequence of parameters for each part shown in the figure1. parts according to the shape of the conversion curve. The 0 10 -1 each part degradation. -1 calcularegression reach was done in the The Temperature C A(10 s ) E activation energy (105J.mol ) n m 268 – 325 325 – 347 347 – 379 379 – 409 4,20 2,66 2,68 2,68 1,35 1,35 1,35 1,50 5,48 3,59 10 7,91 0,50 0,61 0,87 0,20 A(1010 s-1) E activation energy (105J.mol-1) 268 - 409 2,0 1,25 Figure2 Kinetic parameters of Process I n m 1,6 0 Figure 2 - Kinetic parameters of Process I 5.3 Process II – Oxidative decomposition with diffusion The motion of exhausted gases then doesn’t block the diffusion motion into the plate. The porous char core is oxidized and the carbon fibers are being revealed. The porous char core is oxidized on the plane boundary that matches with the plate surface. The plane boundary of oxidizing is being moved inside the plate and a layer of revealed fibers is growing. The growing thickness of the revealed fibers increases the diffusion resistance and decreases the 37 L e t e c k ý z p r avo da j 5.3 Process II – Oxidative decomposition with diffusion The motion of exhausted gases then doesn’t block the diffusion motion into the plate. The porous char core is oxidized and the carbon fibers are being revealed. The porous char core is oxidized on the plane boundary that matches with the plate surface. The plane boundary of oxidizing is being moved inside the plate and a layer of revealed fibers is growing. The growing thickness of the revealed fibers increases the diffusion resistance and decreases the velocity of oxidizing. The important simplifying presumption is that the layer of porous carbon char and fibers have a sharp boundary. The micro-porous layer creates a kind of solid core of the plate and the thickness of the solid plate is decreasing. The process is started about temperatures around 410 0C. The plate looses approximately 21% of its initial mass. The simplifying schema of the process II is shown on the Picture2. In the region out of the plate is a slowly motion of the hot air. The left boundary is meant to be impermeable. The boundary of the plate is surrounded by the thin diffusion air layer where the mass transport runs. The diffusion layer of revealed carbon fibers follows. The oxidizing process goes on the boundary between the carbon fiber layer and micro porous char core. Following those presumptions we can formulate equations describing the surface oxidation and diffusion. 1/2011 cA is molar concentration of oxygen, De is an effective diffusivity coefficient of oxygen in the air and porous layer of revealed carbon fibers, variable r is a distance from the left boundary of the plate and variable t is a time. The equation (1) is solved in quasi-steady-state, where is equal to zero and the reacting surface is in the distance rC from the left boundary of the plate. Then the equation (1) goes to equation (2) (2) where r goes from rC to R, R is the thickness of the plate. The solution of equation (2) must fill those boundary conditions: (3) On the surface of the char core runs the reaction: A(g) + b.B(s) - > products This expresses the value of mass flux of oxygen from the region into the plate surface, where the coefficient kAg is mass transfer coefficient, cAg is concentration of oxygen in the region of the reacting chamber and cAs is concentration of oxygen on the surface of the plate. where A is gas reacting component, B is reacting solid phase. The field of oxygen concentration is described by equatiA(g) + b.B(s) -> products on: (4) rA k AC c AC A e E / RmT c AC where A is gas reacting component, B is reacting solid phase. The field of oxygen concentration is described by equation: De 2 c A c A r 2 t (1) (1) This expresses the value of mass flux of oxygen into the solid reacting core, where the coefficient kAC is velocity constant of surface reaction and cAC is a concentration of oxygen on the char core surface. Simultaneously must be filled that the flux of the oxygen is continuous, including the phase boundaries: Picture2 Schema of Process II Picture 2 - Schema of Process II cA is molar concentration of oxygen, De is an effective diffusivity coefficient of oxygen in the air and porous layer of revealed carbon fibers, variable r is a distance from the left boundary of the plate and variable t is a time. The equation (1) is solved in quasi-steady-state, where c A is equal to zero and the reacting surface is in the distance rC from the left boundary of t (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 38 From the equation (5) can be expressed the unknown variable cAs on the plate surface: (9) (6) that expresses the moving of reacting boundary plane rC with using of extent of the reaction , can the whole equation be expressed as (10) After integration of equation (3) the variable cAs can be substituted from equation (6) and put into integration form of (3) and then can be expressed the variable cAC on the reacting core surface. (5) This equation involves the material’s constants and the shape parameter R, the thickness of the plate. The equation was solved numerically. The temperature dependence of material constants was respected. For diffusion coefficient value of oxygen in the air was used the relation: (11) With using of equation (5) can be expressed the reaction rate of the surface reaction with acting of variables cAC and constant of reaction rate on the core surface kAC. [mol.m-2.s-1] (6) The velocity of mass decrement of solid phase B, char solid core, can be expressed with using of extent of reaction , [mol.m-2.s-1] [m2.s-1] (12) (7) where is a molar density of char solid core and R is a thickness of the plate. With respect to relation between rA a rB Where DAB is a molecular diffusivity, kB is Boltzmann constant, NA is Avogadro number, MA and MB are molecule masses of Oxygen and Nitrogen, T is a absolute temperature in Kelvin and dA and dB are efficient diameters of oxygen and nitrogen molecule. Because the gas diffusion goes through the region of thin carbon fibers, the efficient diffusion coefficient was approximated by this equation5: (8) Where is a porosity of material environment and is a particle tortuosity, which is a kind of structural factor that relates to the ratio of real and ideal trajectory of molecules given by the porous structure. Particle tortuosity has a value 1,0 – 100. For the ideal cylinder is assumed the accurate value of as 1. The influence of Knudsen diffusion was neglected, as the real dimension of pores is of order but Knudsen diffusion is counted in dimensions of order 10 nm. For the numerical calculation those numbers were approximated: =0,35 and =1,4. The temperature dependence of molar concentration of oxygen cA was derived from the gas state equation: where b is a molar coefficient, and with application of relation [mol.s-3] (12) There was assumed that the parameters R, thickness of the plate, De , diffusivity and c Ag , concentration of oxygen can be used with better accuracy than the parameters b and B , that are in the connection with molar compounding of oxidized solid phase. Value of 39 The velocity constant of the surface reaction was supposed as the Arrhenius equation: [m-1s-1] (14) L ewhere tec ý zwith pr a vrelationship o d a j to the 1/2011 k As was approximated after Figure1, thekvalue most temperature 410 0C was taken. (At this temperature the Process II starts and the diffusion resistance is not efficient). cal simulation of equation 15 was done respecting that this The results of the simulation of parameter’s influences are shown in Figure3. The values of B /b, graphical which include parameters were taken according to the valuessample except the parameter process includes 21 % ofreal total mass. The the unknown molar density and stoichiometric number b. Thus, this factor was used to filling simulation was usedalso for searching the acceptable trend the simulation condition (15)review a graphical simulation of equation t dt =1. of the curve. The ofSimultaneously parameters gained by this gra15 was done respecting that this process includes 21 % of total sample mass. The graphical phical simulation is pictured in the figure 3. simulation was used also for searching the acceptable trend of the curve. The review of parameters gained by this graphical simulation is pictured in the figure 3. R (mm) In the equation (10) are used parameters that can be counted only with the limited accuracy. There are next limited occurrences. Parameter R is defined as a half of plate thickness. (A half thickness of plate convention is from symmetry reason). Though the thickness of the plate is changing during the process and the volume of the gapes is increasing. Then the conditions for diffusion are changing too. The value of coefficient kAg can be calculated very difficulty and not accurately and the value of surface velocity constant kA is unknown. The molar density of can be estimated also very roughly. For first approximation can be assumed that the char material is a carbon with a specific ratio of high-molecular hydrocarbons. The equation (10) cannot be filled with accurate constants and parameters, but it is an expression of trend that can be fitted to simulate the process of decomposition. Then the simulation with different parameters can conclude to the creation of semi-empiric model. A(108m.min-1) E (105Jmol-1) kAg (m.min-1) 4 2,0 1,44 0,014 Figure3 Review of simulation results B /b(mol.m-3) 0,35 1,4 1600 1 dt t Figure 3 - Review of simulation results 5.4 Process III – oxidizing decomposition without diffusion 5.4 Process III – oxidizing decomposition without diffusion In this process the porous layer of char is oxidized and only the reaction on the surface of the fibers goes. This process on the fibers surface goes more slowly than diffusion of oxygen, thus the diffusion isn’t assumed and the process is ruled by the kinetic of oxidation only. The process goes partially simultaneously with process II. The extent of this process is increasing proportionally to the extent of process II. When the process II is finished then the process III goes only. The oxidizing of carbon fibers goes according the equation C + O2 = CO2 on the fibers surface. The simply kinetic equation for this process is proposed: Simulation The base for simulation is the equation (15). This equation was integrated for different values of parameters. Only those parameters can be acceptable that give a value of the integral equal to 1. This condition says that extent of the reaction described is 1. The integration root was between temperatures T1=410 0C and T2=550 0C regarding the integrating variable time t. (15) This equation expresses assumption theonvelocity In this process the porous layer of char the is oxidized and only thethat reaction the surface of the fibers goes. This process on the fibers surface goes more slowly than diffusion of oxygen, of the surface reaction is slowing proportionally to the thus the diffusion isn’t assumed and the process is ruled by the kinetic of oxidation only. The process goes simultaneously withthe process II. The extent of of this oxygen process is increasing extent of partially the reaction while concentration is proportionally to the extent of process II. When the process II is finished then the process III constant. For searching of parameters of the kinetic equagoes only. The oxidizing of carbon fibers goes according the equation C + O2 = CO2 on the fibers surface. tion an extra experiment with pure virgin carbon fibers was The simply kinetic equation for this process is proposed: run and parameters from regression task were calculated. /R T The decomposition of the virgin carbon fiber is tcurve Ae Eof (1 ) This equation the assumption the velocity of thethe surface reaction slowing shown on expresses the Picture 3. Asthatcan be seen, trend ofis the proportionally to the extent of the reaction while the concentration of oxygen is constant. For 0 curve changes at the temperature 550 C. The regression searching of parameters of the kinetic equation an extra experiment with pure virgin carbon fibers was and parameters regressionoftask were calculated. The curve of task wasrunsolved for 2from intervals temperatures. The results decomposition of the virgin carbon fiber is shown on the Picture3. As can be seen, the trend 0 of the regression shown550 onC.the TheFigure regression4. task was solved for 2 intervals of the curve changes at theare temperature m of temperatures. The results of the regression are shown on the Figure4. There was assumed that the parameters R, thickness of the plate, De , diffusivity and cAg, concentration of oxygen can be used with better accuracy than the parameters b and , that are in the connection with molar compounding of oxidized solid phase. Value of kAS was approximated after Figure1, where the value with most relationship to the temperature 410 0C was taken. (At this temperature the Process II starts and the diffusion resistance is not efficient). The results of the simulation of parameter’s influences are shown in Figure3. The values of parameters were taken according to the real values except the parameter /b, which include the unknown molar density and stoichiometric number b. Thus, this factor was used to filling the simulation condition (15) =1. Simultaneously a graphi- Temperature reach 0C A(1010 m.min-1) E activation energy (105J.mol-1) 140 – 552 Regression impossible 552 5,8 2,15 Figure4 Kinetic parameters for thermal decomposition of virgin carbon fiber Figure 4 - Kinetic parameters for thermal 6. Analysis of the results decomposition of virgin carbon fiber From the Table1 is evident that the processes with activation energy around 140 kJ.mol-1 go 0 until the temperature 410 C. That is the reason that the following regression task was recalculated for this particular temperature interval. The values of parameters are summarized in the Figure2. The values of parameters n and m are then at expected levels. 6 Analysis ofII the results The process was modeled with kinetic reaction and diffusion. The model was tested b were taken with real values well known parameters with From theofTable1 is parameters. evident The thatunknown the processes B and acti- -1 as a tuning parameters of the model / b which was used as a semi-empiric vation energyparameter around B140 kJ.mol go until the temperature during 0the simulation. The appropriate option of this semi-empiric parameter results to a 410 C. That is the reason that the following regression model that has an acceptable behavior during the simulation. The simulation result of the model was is shown on the Figure3. During the simulation weretemperature varied these parameters that task recalculated for this particular intercannot be expressed accurately: efficient diffusivity, which includes the next parameters and val. valuesof of are summarized inand the andThe the thickness the parameters plate R, which changes during the process theFigure mass transfer coefficient kAg. 2. The values of parameters n and m are then at expected For the process III was proposed a kinetic model without diffusion. The proposed levels. kinetic equation was compared with real values from the decomposition of virgin carbon fiber. The process curve of the II decomposition of virgin carbon gives a reaction remarkable break The was modeled with fiber kinetic andat the temperature 550 0C. The most acceptable explanation is in an influence of fiber sizing and the diffusion. The model was tested with real values of well process of destruction of sizing or there is a kind of surface structure that has a better durability against the destruction. As a real kinetic parameters were taken these which rule the process at the temperature 550 0C. 7. Conclusion During the thermal analysis of the process was used a method of non-linear regression. This C z e c h A e r o s pa c e P r o c e e d i n g s known parameters. The unknown parameters and b were taken as a semi-empiric parameter / b which was used as a tuning parameters of the model during the simulation. The appropriate option of this semi-empiric parameter results to a model that has an acceptable behavior during the simulation. The simulation result of the model is shown on the Figure 3. During the simulation were varied these parameters that cannot be expressed accurately: efficient diffusivity, which includes the next parameters and and the thickness of the plate R, which changes during the process and the mass transfer coefficient kAg. For the process III was proposed a kinetic model without diffusion. The proposed kinetic equation was compared with real values from the decomposition of virgin carbon fiber. The curve of the decomposition of virgin carbon fiber gives a remarkable break at the temperature 550 0C. The most acceptable explanation is in an influence of fiber sizing and the process of destruction of sizing or there is a kind of surface structure that has a better durability against the destruction. As a real kinetic parameters were taken these which rule the process at the temperature 550 0C. 7. Conclusion During the thermal analysis of the process was used a method of non-linear regression. This method can gain all parameters of kinetic equations. The proposal of the process was created involving a part of quick pyrolysis with 40 activation energy about 135 kJ.mol-1 which runs until temperature 410 0C. Then the next part of the process follows involving diffusion and oxidation of porous char solid phase with activation energy 144 kJ.mol-1. The reaction rate of oxidation is comparable with the velocity of diffusion. This process runs at temperatures 410 – 550 0C. The third process has the activation energy around 215 kJ.mol-1. The reaction rate of oxidation is lower than diffusion of oxygen into the region of revealed carbon fibers. The TGA curve of degradation of virgin carbon fiber has a break in the velocity of oxidation. According to this observation there is a reason to assume that the activation energy decreases after some time of temperature exposition. References [1] Slovák, V.: Thermochimika Acta 372(2001), pp.175-182 [2] Paterson-Jones, J.C.: J Appl. Polym. Sci. 1975, 19: pp.1539-47 [3] Puglia, D.; Manfredi, L.B.; Vazquez, A.; Kenny, J.M.: Polymer Degrad. and Stab. 73 (2001), pp.521-527 [4] Pielichowski, K.; Njuguna, J.: Thermal Degradation of Polymeric Materials; Rapra Technology 2005, pp.38 - 40 [5] Missen, R.W.; Mims, C.A.; Saville, B.A.: Introduction to Chemical Reaction Engineering and Kinetics; John Wiley & Sons, Inc.,1999 , University of Toronto, pp. 200 - 201 Analysis of options for calibration of micro accelerometer MAC Analýza možností kalibrace mikroakcelerometru MAC RNDr. Vojtěch Zadražil / VZLÚ, Plc., Prague The options for calibration of a MAC micro accelerometer instrument has been analysed from the point of view of the level of assembly of the instrument. The following levels of assembly of the instrument had been considered at a component level, board level, and the level of a fully assembled instrument. Additionally, all necessary miscellaneous characterization and calibration procedures have been also considered. The component level characterization and calibration covers essential adjusting capacitor trims, key fixed capacitors, analogue to digital converters, and the proof mass. The board level characterization and calibration covers position of detection boards, translational/rotational actuators, science data measuring boards, house-keeping data measuring, and auxiliary signals generating board. The level of fully assembled instrument covers the alignment characterization, the thermal vacuum chamber characterization, calibration of the temperature sensors, and of the science data acquisition channels. The miscellaneous characterization and calibration procedures cover control electrodes, wall stops, prisms, and the dynamic transfer function of the entire control loop of the instrument. Byly analyzovány možnosti kalibrace přístroje mikroakcelerometru MAC z hlediska úrovně sestavení přístroje. Byly zvažovány následující úrovně sestavení přístroje: úroveň komponent, úroveň desek plošných spojů a úroveň úplně sestaveného přístroje. Navíc byly též zvažovány všechny nezbytné rozmanité charakterizace a kalibrace. Charakterizace a kalibrace na úrovni komponent pokrývá nezbytné seřizovací kondenzátorové trimry, klíčové pevné kondenzátory, analogově digitální konvertory a referenční těleso. Charakterizace a kalibrace na úrovni desek plošných spojů pokrývá desky detekce polohy, desky translačních a rotačních regulátorů a měření vědeckých dat a desku měření pomocných veličin a generátorů pomocných signálů. Úroveň úplně sestaveného přístroje pokrývá charakterizaci geometrie čidla a charakterizaci a kalibraci teplotních čidel a akvizičních kanálů vědeckých dat v tepelně vakuové komoře. Rozmanité charakterizační a kalibrační procedury pokrývají řídící elektrody, stěnové zarážky, hranoly a dynamickou přenosovou funkci úplné řídící smyčky přístroje. Keywords: micro accelerometer, calibration, assembly level 41 1 Preface This treatise pertains to the MAC04 Micro-Accelerometer space instrument built at VZLÚ for the orbital mission to orbit the Earth at altitudes of 450 to 550 [km] in roughly circular near polar orbits. This Micro-Accelerometer instrument has been designed to measure at the same time both translational and rotational accelerations of small magnitudes that can be encountered at the projected altitudes of the mission. This Micro-Accelerometer instrument has been designed to cover with certain margin (18 %) the measuring range of (−1*10 −4 ;+1*10 −4 )[m / s 2 ] equal in all three linear channels of Cartesian coordinates of translational acceleration vector and the range of (−0.0096;+0.0096)[rad / s 2 ] again equal in all three angular channels of Cartesian coordinates of rotational acceleration vector. 2 Introduction The process of calibration of a unique scientific measuring instrument encompasses a procedure whereby relationship is established between the instrument output data and the measured quantity. This relationship shall reflect both static and dynamic characteristics of the instrument. Regarding this particular instrument that is designed to measure very small accelerations in comparison to the gravitational acceleration in the rest reference frame on the surface of the Earth it is very difficult to facilitate empirical calibration of the electro mechanical part of the instrument namely that of the sensor proper, i.e., the mechanical motion behaviour of the proof mass as it is subjected to the electrostatic forces and the measured or emulated mechanical acceleration. For this reason the projected calibration utilizes for this particular part of the feedback loop chain just calculations using abstract theoretical model with just few complementary practical measurements. 3 Principle of operation of the micro accelerometer instrument The principle of operation of the accelerometer instrument consists in sensing the position of cubical proof mass relative to cubical cavity in which the proof mass is located and in controlling the position of the cubical proof mass in the cubical cavity in all translational as well as rotational degrees of freedom of motion of the cubical proof mass in all three Cartesian axes. The cavity is only 200 micrometers wider than the cubical proof mass on each side. The entire surface of the measuring cube (cubical proof mass) is completely coated with metal (Chromium) so that any two points on its surface are in galvanic (metallic, conductive) connection to each other. No electrical or mechanical connection exists between the cubical proof mass and the cubical cavity during normal operation of the instrument except for brief periods of discharging the electrical charge that may have accumulated on the measuring cube due to incoming stream of particles of mainly protons and electrons. During this discharge period of few seconds the L e t e c k ý z p r avo da j 1/2011 feedback loops keeping the centre of the cubical proof mass very close to the centre of the cubical cavity are switched off allowing the cubical proof mass to move freely and eventually drift to come into mechanical and electrical contact with wall stops. There are four wall stops at each wall of the cubical cavity located close to the four corners of each wall of the cubical cavity. These wall stops are 180 micrometers high protrusions of the walls of the cubical cavity that are connected to analogue electrical ground. Most of the remainders of the walls of the cubical cavity are coated with metal (Chromium) in such a way that on each wall there is a pair of electrodes. As the interaction force between two electrodes of a plate capacitor charged with nonzero charge is always attractive the control voltages including base polarization and additional control polarizations for translational as well as rotational degrees of freedom of motion are applied to each of these electrodes by means of feedback loops as to keep the cubical proof mass as much aligned with the cubical cavity as possible. The sensing of the translational and rotational positions of the cubical proof mass is facilitated by means of very stable sinusoidal (harmonic) voltage superimposed onto each of the electrode control voltages. The alternating harmonic electric currents passing through each of the pair of control electrode – cubical proof mass (electrode) are dependent on the capacitance of the electrode pair that in turn is dependent on the distance between the electrodes of the electrode pair. This distance is dependent on the actual translational as well as rotational position of the cubical proof mass. The harmonic electric currents passing through two pairs of electrodes that are opposite to one another are then subtracted via magnetic flux of a core of a transformer whose two primary windings are energized by these two currents. The secondary windings of this subtracting transformer provide harmonic voltages whose amplitudes are proportional to the difference of the sensing currents. See Diagram 1 below. The harmonic voltage that is the result of the position detection is then sampled synchronously with the excitation voltage to derive DC signal proportional to the deviation in the corresponding detected position. Distribution unit generates the appropriate control voltages applied to each of the four control electrodes of the cubical cavity to keep the cubical proof mass as closely aligned with the cubical cavity as governed by the strength of this feedback loop in one particular dimension. Further two distribution units (matrices) are used to close the feedback loop in the other two dimensions. It could be demonstrated by detailed analysis that both the equilibrium position displacement and the corresponding additional equilibrium polarization control voltages are directly proportional to the acting apparent translational or rotational acceleration to a very high degree. The input signal to the proportional-derivative amplifier-controller is measured as signal that is directly proportional to position and the output signal from the first stage of the two-stage proportional-derivative amplifier-controller is measured as signal that is directly proportional to apparent acting acceleration. See References [1] and [2]. C z e c h A e r o s pa c e P r o c e e d i n g s Diagram 1 - Circuit schematics of the core of the position detection board 42 43 4 Characterization and calibration of the micro accelerometer instrument The calibration is the process of estimation of parameters characterizing the instrument that are necessary and sufficient to establish the relationship between the instrument measurement data output and the value of the measured physical quantity. These parameters facilitate this relationship between the measured quantity and the instrument measurement data output by means of an algorithm that processes the data output using these characterization and calibration parameters as one of its (parametric) inputs. The characterization is either direct measurement of the parameters, analysis of the measurements or purely theoretical analysis, if need be, if it is the only viable or plausible way to determine the parameters. The characterization and calibration of the micro accelerometer instrument shall cover the behaviour of the instrument over specified range of temperatures from -20[oC] to + 60[oC] . Additionally the characterization and calibration shall include the static and optionally the dynamic behavior of the instrument. The dynamic behavior may optionally cover both temporal dynamics of the analog electronics proper and the effects of variation of temperature in time domain as well as in 2D or 3D space domain on the analog electronics proper. See References [3] and [4]. 5 Levels of characterization of the micro accelerometer instrument There are several levels of characterization of the micro accelerometer instrument that shall be considered, namely Component level characterization and calibration, Board level characterization and calibration, Instrument level characterization and calibration, and Miscellaneous characterization and calibration. These levels of characterization and calibration are treated one by one in the following subchapters. L e t e c k ý z p r avo da j 1/2011 the three position detection boards is just optional. The characterization and calibration of an adjusting capacitor trimmer consists in measuring its static characteristic that maps the relationship between the angle of rotation of the rotor of the trimmer and its capacitance. A well defined and conveniently reproducible angular reference position of the rotor shall be adopted. Preferably that of the rotor fully screwed into the trimmer by a specified torque that is safe to apply to the component. These data are necessary for the complete characterization and calibration of the accelerometer instrument because these trimmers are used to emulate various fixed translational and rotational positions of the proof mass cube of the instrument by setting them appropriately. The change in the angular setting of the rotor of the trimmer is related via the gathered data to the change in the capacitance which in turn is related to the change in the distance between the plates of the plate capacitor formed by a side of the proof mass cube and the adjacent control electrode on the inner surface of the measuring cavity of the instrument. This relationship is calculated from theoretical description of the system of electrodes. The following Figure 1 depicts the relationship between the capacitance change corresponding to angular position change of plus one revolution and the rotor initial angular position, i.e. the rotor angular position before the angular position is changed by +1 revolution. This characteristic has been measured and is plotted in the figure repeatedly thirty times for one and the same sample of the adjusting capacitor trimmer just to ascertain the precision and reproducibility of such measurements. The figure graphically shows the high degree of reproducibility that can be achieved in carefully prepared and executed measurements. Numerical analysis of the gathered data yields a value of standard deviation calculated from the repeated 30 measurements and subsequently averaged over all 181 measuring points of approximately 0.2 [fF] which is twice the resolution of the measuring instrument used to gather the data. 5.1 Component level characterization and calibration For the component level characterization and calibration of the micro accelerometer instrument the following components are considered: adjusting capacitor trimmers, additive fixed capacitors, analogue to digital converters, and the proof mass cube. The characterization and calibrations of these components are treated in the following subchapters of second level one at a time. 5.1.1 Characterization and calibration of the adjusting capacitor trimmers There are four adjusting capacitor trimmers per each of the three position detection boards that should be characterized and calibrated although an approach to the overall characterization and calibration of the accelerometer instrument may be adopted for which only one of them at chosen fixed position has to be necessarily characterized and calibrated. The characterization and calibration of the remaining three adjusting capacitor trimmers per each of Figure 1 - Capacitance change per one revolution − ∆C (ϕ ) versus rotor initial angular position: (1) ∆C (ϕ ) = C (ϕ + δ ) − C (ϕ ), δ = +2π [rad ] = +1[revolution], ϕ ∈< 0;9 > [revolution] C z e c h A e r o s pa c e P r o c e e d i n g s The following Figure 2 shows the standard deviation of capacitance change per one revolution calculated from 30 repeated measurements plotted for various rotor initial angular positions. 44 manufacturer assuming homogeneity of the material of the item. 5.2 Board level characterization and calibration For the board level characterization and calibration of the micro accelerometer instrument the following boards are considered: the three position detection boards X, Y and Z, the three translational and rotational actuator and science data measuring boards X, Y, and Z, the house keeping data measuring and auxiliary signals generating board. The characterizations and calibrations of these boards are treated in the following subchapters of second level one at a time. 5.2.1 Characterization and calibration of the position detection boards Figure 2 - Standard deviation of capacitance change per one revolution versus rotor initial angular position calculated from 30 repeated measurements of one and the same sample of capacitor trimmer 5.1.2 Characterization and calibration of the additive fixed capacitors The capacitances of the additive fixed capacitors to be populated on the position detection boards shall be measured in order to able to find pairs of them that have capacitances as close to one another as possible because it is desirable to populate the position detection boards with these pairs of matched capacitors so that the consequent necessary trimming is as small as possible. 5.1.3 Characterization and calibration of the analogue to digital converters The calibration of the analogue to digital converters essentially consists in verifying its linearity. This operation is optional. During the operation a known voltage precisely measured by another independent instrument is applied to the analogue input of the converter and subsequently data are read out of the converter. The relationship between the data read out of the converter and the voltage read out of the independent precise measuring instrument constitutes the static characteristic of the converter. For good converters this relationship is practically linear with deviations from the linearity of the order of few parts per million. 5.1.4 Characterization of the proof mass cube The characterization of the proof mass cube consists in establishing its mass and moment of inertia. Proper care should be taken to account for buoyancy correction as mass measurements are usually done at ambient conditions including ambient atmospheric pressure while the calibration weights used to calibrate the balance have specific density, usually 8.00 [g/cm3], generally different from the average density of the item weighted. The moment of inertia is calculated using the established mass and known geometry and dimensions of the item guaranteed by the The characterization and calibration of the position detection board consists in establishing the relationships between the translational and rotational output voltages of the board as the dependent variables and the four capacitances parallel to each of the four control electrodes as independent variables. Because this relationship is practically linear to a very high degree and because, taking into account parasitic capacitances of all elements of the circuit, it is not possible to determine the capacitance parallel to particular control electrode absolutely so only the coefficients of partial sensitivities: (2) ∂U T , i ∈ {1,2,3,4} ∂Ci and (3) ∂U R , i ∈ {1,2,3,4} ∂Ci are determined. Here UT and UR represent translational and rotational voltages respectively and C1 , C2 , C3 , C4 represent capacitances parallel to electrodes E1 , E2 , E3 , E4 respectively. 5.2.2 Characterization and calibration of the translational and rotational actuator and science data measuring boards This step represents establishing the static relationships between the translational and rotational input voltages as two independent variables and the four output voltages on the four control electrodes as four dependent variables and also the static relationships between the translational and rotational input voltages as two independent variables and the two digital output data readings from the two analogue to digital converters as two dependent variables. The above mentioned acquired relationships can then be transformed into relationships between the two digital output data readings from the two analogue to digital converters as two independent variables and the four output vol- 45 L e t e c k ý z p r avo da j 1/2011 tages on the four control electrodes as four dependent variables. All of the above mentioned relationships are usually practically linear to a very high degree. and the science channels. These two distinct data acquisition channel categories are treated in the following subchapters of third level one at a time. Additionally the characterization and calibration of the translational and rotational actuator and science data measuring board also entails establishing the time delays of propagation of low frequency signals through the anti aliasing filters that are part of the board structure. 5.3.2.1 Characterization and calibration of the temperature sensors 5.2.3 Characterization and calibration of the housekeeping data measuring and auxiliary signals generating board This characterization and calibration represents establishing the relationships between the auxiliary quantities (namely three linear positions detected by the position detection boards X, Y, and Z, three angular positions detected by the position detection boards X, Y, and Z, two temperatures at two temperature sensors per each of the three position detection boards X, Y, and Z, one temperature at one temperature sensor per each of the following six boards: power source board, CPU board, the tested housekeeping data measuring and auxiliary signals generating board itself, and the translational and rotational actuator and science data measuring boards X, Y, and Z and ten auxiliary voltages) versus the corresponding reading from the analogue to digital converter. The aim of this characterization and calibration is also to measure and set the proper phase relationship between the detected 230400 [Hz] sine wave signal and the detecting rectangular signal of the same frequency of duty cycle ratio 1:1. 5.3 Instrument level characterization and calibration For the instrument level characterization and calibration of the micro accelerometer instrument the following tests of the completely assembled micro accelerometer instrument are considered: the Alignment test, and the Thermal vacuum chamber test. These tests are treated in the following subchapters of second level one at a time. 5.3.1 Alignment characterization The aim of the alignment characterization is to establish the angular relationship between the direction of normals to the specified surfaces of the optical reference alignment cube and the principal directions of the instrument defined by the normals to the internal surfaces of the measuring cavity expressed as transformation matrix. Additionally the radius vector of the centre of the measuring cavity relative to the mechanical reference frame is established during this characterization. 5.3.2 Thermal vacuum chamber characterization and calibration For the thermal vacuum chamber characterization and calibration of the micro accelerometer instrument the following two distinct data acquisition channel categories of the completely assembled micro accelerometer instrument are separately considered: the temperature sensors, This characterization and calibration establishes the relationship between two temperatures at two temperature sensors per each of the three position detection boards X, Y, and Z as independent variables versus the corresponding readings from the analogue to digital converter as the dependent variables using etalon measurement of temperatures. This measurement is complimentary to that carried out during tests described in subchapter 5.2.3. The actual calibration procedure implements a sophisticated method of accounting for and eliminating the effects of the dissipation of the electronics of the micro accelerometer instrument itself. 5.3.2.2 Characterization and calibration of the science channels This characterization and calibration represents establishing the relationships between the science quantities (namely three translational accelerations X, Y, and Z, and three rotational accelerations X, Y, and Z) versus the readings from the corresponding analogue to digital converters including temperature effects in these relationships. These temperature effects can be classified into two categories: short-term effects and long-term effects. 5.4 Miscellaneous characterization and calibration This chapter covers the characterization and calibration of the following topics: control electrodes, wall stops, cavity forming prisms, and analytical characterization of the dynamic transfer function of the entire control loop. These topics are treated in the following subchapters of second level one at a time. 5.4.1 Characterization of the control electrodes The characterization of the control electrodes entails establishing the size of the area of the electrode and the position within the plane of the electrode of its two-dimensional centre. 5.4.2 Characterization of the wall stops The characterization of the wall stops entails establishing their heights to which they protrude above the corresponding surface of the prism comprising the measuring cavity. This height must conform to stringent limits imposed on them. 5.4.3 Characterization of the prisms comprising the measuring cavity The characterization of the prisms comprising the measuring cavity entails establishing the measure of flatness of the working surfaces thereof. 5.4.4 Analytical characterization of the dynamic transfer function C z e c h A e r o s pa c e P r o c e e d i n g s 46 This characterization entails the calculation of the dynamic transfer function of the entire control loop beginning at the analogue input signals of accelerations, continuing via the proof mass cube translational and rotational mechanics, via the position detection board (input circuits, preamplifiers, and amplifiers, synchronous detectors, output circuits), via the PD (Proportional Derivative) regulator first stage amplifier, then continuing in the loop to the PD regulator second stage amplifier, via the distribution unit closing the control loop at the control electrodes back at the proof mass cube, and also including the branch outside the control loop at the PD regulator first stage amplifier output to the anti aliasing filter and ending at the analogue to digital converter that performs the actual measurements and delivers the digital output signal measuring the science data. Denoting the input signal x(t) and the output signal y(t) we have the following definition of the transfer function T(s) of a linear time invariant system of single input and single output. The transfer function T(s): T (s) = (4) L{ y (t )}( s ) , s = σ + iω , σ ∈ R, ω ∈ R, R = {Set _ of _ all _ real _ numbers} L{x(t )}( s ) 6 Conclusion The options for calibration of the micro accelerometer instrument MAC have been analyzed with the emphasis on the level of assembly of the instrument. The structure of the calibration options on each assembly level has been elaborated detailing significant calibration steps that may be applied for actual calibration process. In addition to calibration procedures carried out in previous projects this work specifies methods that extend the repertoire of applicable methods and are completely new. See subchapters 5.1.1, 5.1.3, 5.2.1, 5.3.1, 5.3.2.1, and 5.4.4. So far these methods have not been used either for their exacting demands on instrumentation, experimental laboratory setup and the actual execution of the process of calibration or because they have not been previously even considered. These new methods enhance the possibilities of the calibration of the instrument to previously unforeseen levels. The application of these new methods may result in substantial improvements of precision and reliability of calibration parameters derived from the calibration data gathered during the calibration measurements. This results in overall improvement and enhancement of the calibration process. Hence the performance of the micro accelerometer instrument MAC itself is enhanced. References Where the unilateral Laplace transform L(s) is defined by the following formula: (5) ∞ L{ f (t )}( s ) = ∫ f (t ). exp(− s.t ).dt [2] Chvojka, M.: ACC Instrument Technical Description, Issue 2.0; VZLU, Prague, February 9, 2010 [3] Zadrazil, V.: ACC Characterization and Calibration Test Procedure, Issue 1.3; VZLU, Prague, April 2, 2010 0 [1] Chvojka, M.: ACC Performance analysis, Issue 2.7; VZLU, Prague, April 2, 2010 (6) [4] Zadrazil, V.; Vondrak, M.: Alignment Test Procedure, Issue 1.1; VZLU, Prague, March 25, 2010 This is usually denoted as F ( s ) = L{ f (t )}( s ) Intentionally left void. 47 L e t e c k ý z p r avo da j 1/2011 Worst Case Analysis Methods Implementation in an Accelerometer Measuring Loop Analysis Aplikace postupů Analýzy nejhoršího případu (WCA) při analýze měřicí smyčky akcelerometru Jozef Zakucia, Milan Merkl / VZLÚ, Plc., Prague Worst case analysis belongs to the most important in the analysis section that is performed to document space product quality assurance. It is important from the viewpoint that WCA procedures were described in the most concise way together with provision of a guide to its successful implementation linked to other analyses of product assurance documents system. The objective of the presented article is that such description is reliable in both a general and at an implementation level. V oblasti analýz kosmického vybavení prováděných k průkazu zajištění jakosti produktu patří analýza nejhoršího případu k nejdůležitějším. Proto je důležité, aby její postupy byly popsány způsobem co nejvýstižnějším a zároveň poskytujícím návod pro její úspěšné provedení v návaznosti na ostatní analýzy systému dokumentů pro zajištění produktu. Cílem prezentovaného článku je takovýto popis v rovině obecné i implementační. Keywords: WCA, tolerance, sensitivity, RSS, MCA, accelerometer, product assurance. 1 Introduction The purpose of the WCA analysis in general is to serve as a tool for design centering, i.e. modifying of nominal system design to receive optimum system parameters [1]. Particular WCA target is providing an assurance by analysis that system design complies with requirements. The article describes WCA implementation for assessment of accelerometer measuring loop design. General principles of acceleration measuring with an electrostatic accelerometer system were described by Fontana in [4]. Basic information to the problems of tolerance analysis presents Géher in Chapter 4 of his monograph [2]. Three different methods of the tolerance effects analysis are generally named as: • Extreme Value Analysis (EVA); • Root-Sum-Square (RSS) Analysis; • Monte Carlo Analysis (MCA). The RSS Timing (and Delay) Analysis is performed side by side with classical WCA procedure for digital electronics circuits. Analyzing of an instrument circuits generally points to interfaces analysis using EVA and RSS methods, and MCA method is preferred in case of signal path analysis. In this article we confine to WCA methods description on an example of the signal path analysis and we compare results received using both RSS and MCA method. The worst case analysis works with the circuit structure, tolerances of component characteristics, their temperature changes and time and radiation consequence changes. Grouping of tolerances from individual sources and detail tabular comparison of the three different WCA methods is presented in [5]. It is suitable for WCA to use a specialised methodology with computer program support. It combines processes of tolerance and sensitivity analyses to complete the task. The methods published by Robert R. Boyd in [3] are very useful for completing the task. 1.1 Acronyms and abbreviations AKTRAD ECSS EVA FMECA GNRHSK MC MCA POSDET RSS WCA One of the three identical electronic boards of accelerometer with sensor moving part position control as the main function European Cooperation for Space Standardization Extreme Value Analysis Failure Modes Effects and Criticality Analysis Electronic board of accelerometer with precision amplitude sinusoidal generator and polarization voltage source (besides other functions) Monte Carlo (method) Monte Carlo Analysis One of the three identical electronic boards of accelerometer with sensor moving part position detection as the main function Root Sum Squared, (Root Sum of the Squares) Worst Case Analysis 1.2 Standards Space systems design WCA is supported by standards that describe its methods with different level of detail. Initial information to WCA could be found in Task 206 of MIL-STD-785B standard [6] (formerly cancelled). The task is placed there to the Configuration analyses group SpaceFMECA systems design WCA is The supported bythere standards containing and Sneak circuit analysis as well. analysis is named as Electronic parts/circuits tolerance analysis. 1.2 Standards that describe its methods with different level of detail. IniThe list of active and discontinued ECSS Standards the WCAin standard [7] as tial information to WCA could presents be found Task 206 of discontinued with requirements transferred to [8] and the remaining data to [9]. The standard MIL-STD-785B [6] in (formerly refers to an ECSS technicalstandard memorandum ([10], WG draft version cancelled). to the date only) forThe aging data and to [9] for the WCA methodology. task is placed there to the Configuration analysis group Other relevant ECSS standards are placed in Spacecircuit Engineering region [11],as [12]. End-of-life containing FMECA and Sneak analysis well. The parameters important for the WCA implementation were presented in [13] that was formerly replaced withis derating standards (the actual of which is [14])parts/circuits without EoL parameters. As an analysis named there as Electronic toleranaccessible source of EoL data could be used instead a valid DoD handbook [15] that is free ce analysis. available. The detail WCA guidelines are contained in Appendix B of Jet Propulsion Laboratory handbook [16]. The is to perform the analysisStandards before the designpreis The list of standard activerecommendation and discontinued ECSS frozen. The analysis should comprise all functionally important circuits of the equipment and it shouldthe consider AC, DC, and transients. sents WCA standard [7] as discontinued with require- ments transferred to [8] and the remaining data to [9]. The 2 WCA Methodology standard refers to an ECSS technical memorandum ([10], 2.1 Extreme Value Analysis (EVA) in WG draft version to the date only) for aging data and to EVA procedure [16] starts with total parametric variation calculation for every used part:for the WCA methodology. [9] 1+dP = (1+dX)(1+dS)(1+dT)(1+dE)(1+dR) (1) where the individual parametric variations (normalizedare to 1,placed i.e. 1% in 0.01) denote: EngiOther relevant ECSS standards Space dP total parametric variation dX part initial tolerance neering region [11], [12]. End-of-life parameters important dS aging and drift variation for the WCA implementation were presented in [13] that was formerly replaced with derating standards (the actual of which is [14]) without EoL parameters. As an accessible source of EoL data could be used instead a valid DoD C z e c h A e r o s pa c e P r o c e e d i n g s 48 handbook [15] that is free available. The detail WCA guidelines are contained in Appendix B of Jet Propulsion Laboratory handbook [16]. The standard recommendation is to perform the analysis before the design is frozen. The analysis should comprise all functionally important circuits of the equipment and it should consider AC, DC, and transients. 2 WCA Methodology 2.1 Extreme Value Analysis (EVA) EVA procedure [16] starts with total parametric variation calculation for every used part: 1 1 + dpf 1 1 + dpf Q= 1 1 1 1 1 + dpf Step 6: Perturbate consecutively each time one from part parameters / input values using counter p: (3) (4) (1) Vrp = G(c1 • Qp,1, c2 • Qp,2, ..., cN • Qp,N) 1+dP = (1+dX)(1+dS)(1+dT)(1+dE)(1+dR) where the individual parametric variations (normalized to ) denote: dP total parametric variation dX part initial tolerance dS aging and drift variation dT temperature influence variation (in worst case direction) dE variation due to applied voltage dR variation due to radiation degradation After the cycling of p = 1.. N receive the vector: Vr = [Vr1 , Vr2 , ..., VrN] Step 7: EVA procedure in steps: (details see in R. R. Boyd: Tolerance Analysis of Electronic Circuits [3], the procedure is modified here) Step 1: Specify nominal input values and nominal part parameters. Step 2: Specify nominal output value as function G(c1, c2, ..., cN) of nominal input values and nominal part parameters. N free variables in total. Step 3: Assign the function to output variable Vo. Step 5: Set up Q = dpf I + 1, where: I is unit (diagonal) matrix; dpf is directional perturbation factor; Q is perturbation matrix with dimension N x N. (2) (6) Vrp 1 Sen p = − 1 Vo dpf and set up the vector: (7) Sen = [Sen1, Sen2, ..., SenN]. Step 8: Assign tolerances to part parameters using data sheets and create the tolerance array T with the dimension 2 x N: Step 4: Create p = 1..N as counter of input values and part parameters. (5) −Tr T = 1 Tr1 (8) −Tr2 −TrN Tr2 TrN Step 9: Create the new tolerance array M with tolerances swapped in the columns which parameter sensitivity values are negative: (9) 1 − sgn( Sen1 ) ⋅ Tr1 1 − sgn( Sen2 ) ⋅ Tr2 1 − sgn( SenN ) ⋅ TrN M = 1 + sgn( Sen1 ) ⋅ Tr1 1 + sgn( Sen2 ) ⋅ Tr2 1 + sgn( SenN ) ⋅ TrN 49 L e t e c k ý z p r avo da j Step 10: 1/2011 Standard deviation of the R1 resistor for normal (Gaussian) distribution and tolerance T1 of the resistor is: (10) Receive: Vevk = G(c1 • Mk,1, c2 • Mk,2, ..., cN • Mk,N) for extreme values using row index k = 1, 2 (11) σR = 1 (14) R1 ⋅ T1 3 Let us denote: The column vector with dimension 2 x 1: (15) ∆Vrs = 3σ Vo Vev Vev = 1 Vev2 Then we can derive (see [3]) next general equation that will be refined in the next steps: is the result of extreme value analysis (EVA) and the relation: (16) (12) ∆Vrs = Vo ⋅ Vev − Vo ∆Vev = Vev − Vo = 1 Vev2 − Vo N ∑ ( Sen p =1 p ⋅ Tp ) 2 RSS procedure in steps: Step 11a: For symmetric tolerances and row index k (=1, 2) compute (for Senp values use steps 1 to 7 of the EVA part of the procedure or determine the sensitivities analytically): is valid for deviations from the nominal output. 2.2 Root-Sum-Square (RSS) Analysis For variations with random component, every variation consists from two parts. One of them presents bias with predictable direction, the other is the random component contribution [16]. Biased portions are summarized to resultant bias. Random components are processed in an RSS procedure and the result is added to resultant bias to receive the resultant output value variation. For detail description of such procedure see [5]. RSS definition: Variance (root sum squared, dispersion) of any function is described by the relation (see [3], Appendix: Derivation of the RSS Equation): ∆Vrssk = ( −1) ⋅ Vo ⋅ k N ∑ ( Sen p =1 p ⋅ Tk , p ) Step 11b: For asymmetric tolerances and row index k: σ 2 (Vo ) = ∑ 2 where: Xi ... characteristic parameter of the i-th component, σ 2 = Var(Xi) ... variance, σ ... standard deviation. −Tr Ta = 11 Tr12 2 ∂Vo 2 ⋅σ ( X i ) i =1 ∂X i N (18) (13) ∂Vo Var (Vo ) = ∑ ⋅ Var ( X i ) i =1 ∂X i or N 2 Go to step 13. (17) −Tr21 −TrN 1 Tr22 TrN 2 create the new matrix swapping the tolerance row position values at places with negative sensitivities: ( −u ⋅ Tr + v ⋅ Tr ) Ma - 1 = 1 11 1 12 ( +u1 ⋅ Tr12 − v1 ⋅ Tr11 ) where: (u (u p p ( −u2 ⋅ Tr21 + v2 ⋅ Tr22 ) ( +u2 ⋅ Tr22 − v2 ⋅ Tr21 ) = 1, v p = 0 ) for Sen p ≥ 0, = 0, v p = 1) for Sen p < 0, (19) ( −uN ⋅ TrN1 + vN ⋅ TrN 2 ) , ( +uN ⋅ TrN 2 − vN ⋅ TrN1 ) C z e c h A e r o s pa c e P r o c e e d i n g s 50 for Senp values use steps 1 to 7 of the EVA part of the procedure or determine the sensitivities analytically. Step 12: Receive the respective RSS deviation from the formula: ∆Vrssk = ( −1) ⋅ Vo ⋅ k N ∑ Sen ⋅ ( Ma − 1) p =1 p (20) k, p 2 Step 13: Create the deviation 2x1 column matrix: (21) ∆Vrss1 Vrss1 − Vo ∆Vrss = = = Vrss − Vo ∆Vrss2 Vrss2 − Vo or output limits 2x1 column matrix: (22) Vrss1 Vo + ∆Vrss1 Vrss = = Vrss2 Vo + ∆Vrss2 The column vector with dimension 2 x 1 is the result of root sum square (RSS) analysis. 2.3 WCA – Monte Carlo Method One of the most realistic estimation of the worst case performance is Monte Carlo analysis. Inputs for MCA are probabilistic distributions of the part values variability. And output of the MCA is probability distribution of circuit output values. The result of the MCA is commonly presented in the form of a histogram. The commonly used probability distribution of the circuit components parameters are: • Uniform distribution function (the most conservative for parameter variations). • Gaussian distribution function (the most realistic for parameter variations). • In the next we will show, in step by step procedure, how to perform MCA [3]. Step 4: Specify N – number of samples, and generate N Gaussian (or Uniform) random numbers (with mean = 0 and standard deviation = 1) for each of the M components. Assume that the numbers are spread between -3 and 3 (99.74% of the numbers lie in this range). Step 5: Consider: upper total tolerance of the part = +up% lower total tolerance of the part = -lo% range = 3-(-3) = 6 Tm – tolerance multiplier z – single random number from N than Tm = ( up − ( −lo ) ) 100 range ( z + 3) + (23) ( −lo ) + 1 100 Convert the N random numbers (from step 4) to random tolerance multipliers Tmi for each of the M components using equation (23). Step 6: Create N random values of each component in its given tolerances (by multiplying nominal value of the component and N random tolerance multipliers). Step 7: Calculate N random output values for the N random values of each component. Step 8: Create histogram of the output values. 3 Example solution Accelerometer measuring loop analysis: MCA and RSS analysis were performed on measuring loop of the accelerometer. Simplified block scheme of the accelerometer feedback loop, created for the WCA purposes, is shown in Figure 1. Step 1: Define nominal circuit component values (M components). Step 2: Create the circuit (transfer) function (or matrix) and calculate nominal output of this function. Step 3: Create an array with tolerance of each part. Figure 1 Block scheme of accelerometer measuring loop Input for the feedback system presented in the Figure 1 is acceleration Γ and output is voltage UT. General formula for acceleration was derived in the form (meanings of the symbols 51 L e t e c k ý z p r avo da j are described below): Γ = (24) C0 D U 42 U 32 − m ( D + x )2 ( D − x )2 Next simplified formula was derived for translation acceleration value computation from the known sensor parameters and both internal and measured characteristic values of accelerometer: Γ = (25) 4C0 x 2 1 2 2 PVU T − PV + U 0 + U T mD D 2 Substituting for x from the relation presented in explanation of UT in Table 1 into equation (25) we receive the resultant simplified equation: PV U 0 C0 ACNTR LPS polarization voltage between the sensor electrodes; (+11) Pv = fde (U10) [V] amplitude of sensing sinusoidal voltage with 230.4 kHz frequency; (1.0) [V] capacitance of an control electrode-cube electrode pair with cube placed in the central position; [F] (13.1•10-12) AKTRAD amplifier’s gain; (8) [-] POSDET transfer from displacement to output voltage; (105) LPS = fp (Kca, Kfa, …..) [Vm-1] Functions fd1, fd2, fde, fp present simplified notations of the functional relations of their arguments. Detail insight into the system functions and equations is not introduced (it is comprised into computations) because of the limited length of this presentation. Above mentioned description of the system suffices for our considerations in this paper. 3.1 EVA The procedure works with the simplified formula for acceleration (26) value. Γ = 4C0 1 U T PV − mD DACNTR LPS (26) 2 1 2 2 PV + 2 U 0 + U T Neglecting x2 against D2 in start of treatment of the precision formula (24) that led to receiving of formula (25) implies acceleration value error not over 2 % of the precise value and simplifies essentially the next computations for RSS analysis. Formula (26) was applied for both EVA and RSS computation. The precise formula (24) was applied for the MCA computation. Meanings of the symbols in Equations (24) to (26) and Figure 1 are listed in Table 1. Table 1 Explanations of symbols used in Equations (24) to (26) translation component of on board non-gravitational acceleration [ms-2] equilibrium translation of cube position from the central position caused by constant non-gravitational acceleration [m] gap between electrodes when the sensor cube is placed in the central position; (2 •10-4) [m] sensor electrode No. 3 control composite voltage U3 = fd1 (UT, PV, UT, Ri) [V] sensor electrode No. 4 control composite voltage U4 = fd2 (UT, PV, UT, Rj) [V] translation component of regulation voltage between the sensor electrodes U T ( x ) = x ⋅ ACNTR ⋅ LPS [V] (27) Γ = F{xi } = F{C0 , PV ,U 0 ,U T , ACNTR , LPS }, Both sensor mass m and gap distance D are supposed to be constant, translation x, and consequently UT voltage are taken as parameters and the EVA computation is performed for one of the parameter possible values (x = 3.1•10-6 m, UT = 2.48 V). Individual independent variables are functions of circuit elements, mainly of pertinent POSDET and AKTRAD circuit boards elements. (28) n dx dF = d (ln F ) = ∑ 1S Γ xFi i ; where xi ∈ {C0 , PV , U 0 , U T , ACNTR , LPS } F xi i =1 and where we evaluate the relative differential first order sensitivities generally according the formula: Γ x D U3 U4 UT 1/2011 S Γ xF = 1 (29) ∂ ln F ( X , P ) x ∂F ( x, p ) = ⋅ F ( x, p ) ∂ ln X ∂x without considering dependence on complex frequency p in our case. The relative differential sensitivities were determined both analytically in the next procedure (procedure of the hand computation is presented below and the results are in the fourth column of Table 2) and according the Steps 1 to 7 of EVA procedure in 2.1 clause (results of the sensitivities computation performed in Mathematica are in the fifth column of Table 2). The relative acceleration tolerance expressed using relative differential sensitivities of the determined acceleration value to individual circuit element values: C z e c h A e r o s pa c e P r o c e e d i n g s n dx ∆Γ = ∑ 1S Γ xFi i = Γ xi i =1 = 52 3.2 RSS (30) , dC0 + C0 dP dU 0 1 {( PV DACNTR LPS − 2 PV2 ) V − U 0 2 1 PV U0 PV DACNTR LPS − PV2 + U 0 2 + U T 2 2 1 dU T 2 1 2 dA + PV DACNTR LPS − PV2 + U 0 2 + 3U T 2 + PV + U 0 + U T 2 CNTR + 2 2 U T ACNTR For the next acceleration RSS tolerances evaluation: We will use the results of particular RSS symmetric tolerances evaluations (see the Table 2). The root of sum of quadratic terms is extracted from the Table 2 data according to an RSS template Step 11a and Step 13: 1 dL + PV2 + U 0 2 + U T 2 PS } 2 LPS 5 We will replace dUT/UT with expression gained by differentiation the formula for UT (see Table 1): U T = x ⋅ LPS ⋅ ACNTR ⇒ (31) dU T dx dLPS dACNTR = + + UT x LPS ACNTR and we will receive after a sequence of simplifications and substituting nominal parameter values from the Table 2: ∑ ( Sen ⋅ T ) i =1 i i 2 (34) � 3.14 ⋅10−1 [ % ] + 3.72 ⋅10−7 [ % ] + 8.57 ⋅10−7 [ % ] + 2 2 2 + 2.37 [ % ] + 9.76 [ % ] � 12.444 [ % ] � 12.4 [ % ] 2 2 2 2 The result for relative tolerance value of Γ is: ∆Γ = ± Γ RSS ∑ ( Sen ⋅ T ) i i i 2 (35) � ± 3.53% (32) 3.3 MCA n ∑ i =1 S Γ xFi 1 dxi dC dU dP � 1 ⋅ 0 − 6.1 ⋅10−4 ⋅ 0 + 9.26 ⋅10−1 V + xi C0 U0 PV + (1.07 − 1.83 ⋅10−3U T 2 ) dA dx + (1.07 − 1.22 ⋅10−3U T 2 ) CNTR + x ACNTR + (1.07 − 1.22 ⋅10−3U T 2 ) dLPS dC dU � 1 ⋅ 0 − 6.1 ⋅10−4 ⋅ 0 + LPS C0 U0 + 9.26 ⋅10−1 Ti approximate value [%] Relative tolerance symbol C0 C0 U 0 U0 PV PV ACNTR ACNTR LPS LPS x x dPV dA dL dx + (1.07 ) + (1.07 ) CNTR + (1.07 ) PS PV x ACNTR LPS Source of tolerance data / Comment accelerometer performance Seni approximate value – see (32) Seni approx. value – RSS proc. in steps (clause 2.2) 1 1 1 accelerometer performance -6.110-4 -6.1310-4 0.001 accelerometer performance 9.2610-1 9.3010-1 1.44 accelerometer detail analysis 1.07 1.07 0.56 3.2 RSS We will proceed with Steps 8 to 12 for the next EVA comFor the next acceleration RSS tolerances evaluation: We will use theand results of particular evaluations (see the Table 2). putations receive in RSS the symmetric end thetolerances following relative toleranThe root of sum of quadratic terms is extracted from the Table 2 data according to an RSS ces forStep the11aacceleration: template and Step 13: 5 2 i i 2 1 7 2 7 2 2.37 % 9.76 % � 12.444 % � 12.4 % 2 2 The result for relative tolerance value of is: ∆2 i � 3.53%= Seni T RSS EVA i Γ Γ 2 −5.16 % +5.30 % 2 (33) (34) (35) 3.3 MCA MCA was performed on measuring loop of the accelerometer. Simplified block scheme of the accelerometer feedback loop, created for the WCA purposes, is shown in Figure 1. MCA considers the tolerances of substantial resistors in the signal measuring path and particular (35) Γ = f (U T ( x ) ) accelerometer 1.07 1.07 detail analysis we suppose x as 0 being a 0 0 parameter We will proceed with Steps 8 to 12 for the next EVA computations and receive in the end the followingTable relative tolerances for thetolerances acceleration: 2 Relative and corresponding 5.16 % sensitivities values (33) EVA 5.30 % i 1 where 2.92 Sen T � 3.14 10 % 3.72 10 % 8.57 10 % MCA was performed on measuring loop of the accelerometer. Simplified block scheme of the accelerometer feedback loop, created for the WCA purposes, is shown in Figure 1. MCA considers the tolerances of substantial resistors in the signal measuring path and particular POSDET units. Input for this feedback system, presented in the Figure 1 is acceleration Γ and output is voltage UT. The aim of this analysis is to compute tolerances of the acceleration Γ . Mentioned system is described by the function (36) U T ( x ) = x ⋅ ACNTR ⋅ LPS corresponding with (24) where x (parameter) is the position of the sensor cube (see Table 1 for other details). MCA was performed following the step by step procedure presented in clause 2.3. The computation was realized by running MATLAB script. Inputs for the script were array of the accelerometer part tolerances and function of the system stated in clause 3. The result of this computation (for parameter x set to e.g. 3.1 µm) is histogram of the acceleration Γ shown in Figure 2. The 2 σ value of the final acceleration distribution stands for 4.24% tolerance of the acceleration value. U T3 x x ACNTR LPS (36) 5 corresponding with (24) where x (parameter) is the position of the sensor cube (see Table 1 for other details). (37) MCA was performed following the step by step procedure presented in clause 2.3. The computation was realized by running MATLAB script. Inputs for the script were array of theMCA accelerometer part tolerances and function of the system stated in clause 3. 2σ = 4.24 % A histogram of the acceleration Γ was created and comThe result of this computation (for parameter x set to e.g. 3.1 µm) is histogram of the puted forΓ shown eachin value a ofset of the parameter (in itsfor acceleration Figure 2. xThefrom 2σ value the final acceleration distribution stands 4.24% tolerance of the acceleration value. between voltage U T and acceworking range). Dependence 2 4.24 % (37) leration Γ with its probabilistic distribution is presented A histogram acceleration Γ was4created and and computed each value x from a setreof the in Figureof 3theand Figure in 3D 2Dforrepresentation parameter (in its working range). Dependence between voltage UT and acceleration Γ with its spectively. probabilistic distribution is presented in Figure 3 and Figure 4 in 3D and 2D representation MCA respectively. 600 -2 +2 Number of samples 500 L e t e c k ý z p r avo da j 1/2011 4 Conclusions Comparison of used WCA methods illustrates, that EVA/ RSS methods present data with great effort even in the case of using dedicated software. MCA method encores graphical outputs as a bonus for better presentation of the results. For presented example of accelerometer measuring feedback loop the resulting tolerance values for the three used methods are: EVA {-5.16 %, +5.30 %}; RSS {-3.53 %, +3.53 %}; MCA {-4.24, +4.24 %}. Regardless of the results differences, all the methods give well corresponding results and their parallel using supports assurance of the results correctness. 400 300 References 200 100 0 -1.5 -1.45 -1.4 -2 [ms ] -1.35 -1.3 -1.25 [1] Novák, M.: Theory of Network Tolerances. Academia Praha 1987, (published in Czech: Teorie tolerancí soustav) 340 pp. [2] Géher, K., Theory of Network Tolerances. Akadémiai Kiadó, Budapest 1971, 184 pp. [3] Boyd, R.R.: Tolerance analysis of electronic circuits using MATLAB. CRC Press, 1999, 147 pp. [4] Fontana, G.: High Performance Electrostatic Sensors and Actuators for LISA Proof Mass Control. University of Trento, 2002, 57 pp., http://arxiv4.library.cornell.edu/ftp/physics/papers /0111/0111006. pdf -4 x 10 Figure 2 Histogram of acceleration PDF in 2D Figure 2 - Histogram of acceleration PDF in 2D [5] Worst Case Circuit Analysis (WCCA). System Reliability Center, http://src.alionscience.com/pdf/WorstCaseCircuitAnalysis.pdf, 2004 Alion Science and Technology. Figure 3 - Three-dimensional view of acceleration PDF histograms for variable UT value [6] MIL-STD-785 Reliability program for systems and equipment development and production [7] ECSS-Q-30-01A, First issue, Worst case circuit performance analysis. 31. March 2005, Discontinued Requirements transferred to ECSS-Q-ST-30C, remaining material to ECSS-Q-HB-30-01A [8] ECSS-Q-ST-30C Space product assurance – Dependability. 6 March 2009, 54 pp. [9] ECSS-Q-HB-30-01A (working draft) [10] ECSS-Q-TM-30-12C (working draft) [11] ECSS-E-ST-10-04C Space Engineering - Space Environment. 15 November 2008, 198 pp. [12] ECSS-E-ST-10-12C (15 November 2008) Space Engineering Methods for the calculation of radiation received and its effects, and a policy for design margins [13] ECSS-Q-60-11A (7 September 2004) Space product assurance – Derating and end-of-life parameter drifts – EEE components [14] ECSS-Q-ST-30-11C Space product assurance – Derating – EEE components. 31 July 2008, 59 pp. [15] MIL-HDBK-1547A (06 July 1998): DoD Handbook – Electronic Parts, Materials, and Processes for Space and Launch Vehicles; 253 pp. [16] JPL D-5703, Reliability Analyses Handbook. NASA JPL, Jul 1990,169 pp. Figure 4 - Acceleration PDF for variable UT value in 2D with colour scale of probability density C z e c h A e r o s pa c e P r o c e e d i n g s 54 Experimental Assessment of the Fragments Effect to the Fuselage of the Aircraft Experimentální určení účinku střepin na drak letounu Ing. Miroslav Lošťák, Doc. Ing. Miloslav Petrásek CSc. VUT Brno The article is focused on the assessment of the most probable way of an attack to the civilian aircraft. The assessment of the effect of this attack and capability of the aircraft to continue flight after damage. The attack with missile with a fragmentary warhead was assessed as the most dangerous one. The influence of this attack can be simulated using finite element method and the validation for its functionality of this model was performed using a suitable shooting experiment. The specimen was made as a semi-monocoque skin panel and it was impacted by the fragments of known mass, velocity and direction of the stroke to the impacted area. The experiment proved that single fragment causes only minor damage and has a little influence to the capability of the panel to carry loads. Also more fragments, with proper distance between each other, damage more single parts, but they don’t cause significant damage to the panel which can be explained as a sum of partial influences of single fragments. The results of this experiment were used for the improvement of the numerical calculation of the damage. Předkládaný příspěvek je zaměřen na stanovení nejpravděpodobnějšího způsobu útoku na civilní letadlo, na hodnocení jeho důsledků a schopnosti letounu pokračovat po poškození v letu. Jako nejnebezpečnější způsob útoku byl vyhodnocen útok tříštivou bojovou hlavicí rakety. Důsledky takového útoku byly simulovány s využitím metody konečných prvků, pro ověření funkčnosti tohoto modelu byl proveden vhodný střelecký experiment. Zkušební vzorek byl vyroben jako poloskořepinový potahový panel a byl ostřelován jednotlivými střepinami se známou hmotností, rychlostí a směrem dopadu na zasaženou plochu. Experimentem se prokázalo, že jednotlivá střepina způsobí jen velmi malé poškození a má malý vliv na schopnost panelu přenášet zatížení. I více střepin dostatečně vzdálených od sebe poškodí spíše jednotlivé prvky, ale nevedou k většímu poškození, které by se dalo vyjádřit jako součet dílčích účinků jednotlivých střepin. Výsledky experimentu byly následně použity ke zpřesnění numerického výpočtu poškození. Introduction We can meet with more and more international terrorism in today news media. There are also attacks to the targets very vulnerable which are mostly civil transport aircrafts. Whatever attack to the aircraft has mostly tragic results. If there is a crash of the aircraft then it results in a massive loss of lives and also in big material losses. This article is focused on the assessment of the dangerous effects of the terroristic attack. Terroristic attacks could be divided from the point of view of the execution into two categories. The first category includes attacks on the board of the aircrafts which can be made using firearms or explosives. The second category then represents attacks made to the aircrafts from the outside. Into this category have been put attacks with firearms of the bigger caliber and with warheads of the missiles. If there are used missiles then it is mostly missiles with fragmentation warhead which are commonly used as anti-aircraft weapons in the most of the war conflicts. The security of the airports was significantly increased after last terroristic attacks on the boards of the civil aircrafts. Therefore there has been recently no successful terroristic attack on the board of the aircraft. For this reason we have assessed as the most dangerous one attack to the aircraft from the outside. Analysis of the mechanism of damage of the aircraft by the firearm and by the fragmentation warhead shows that these mechanisms are very similar. The projectile from the firearm affects the airframe of the aircraft by its kinetic energy and causes damage with it. By the fragmentation warhead the damage is caused by the effect of bigger amount of the fragments which are made by the stretching of the skin of the warhead. These fragments use their kinetic energy for the damage of the airframe too. The character of the damage with single fragment could be different then damage caused by one projectile. This is caused mainly by the shape of the fragments. Effects of the damage of airframe to the immediate static strength are, with comparison of damage with single projectile and with single fragment, the same. If we are talking about damage with fragmentation warhead then of course it is damage made by multiple fragments. The effect of shooting the aircraft with the fragmentation warhead could be divided into three categories according to the range of the damage. The first category is direct stroke to the aircraft. Here a fatal damage of the structure could be expected which leads to the catastrophe of the aircraft. In the second case the warhead explodes in the distant to the aircraft smaller then critical distance. That means continuous damage of the airframe by the very thick group of the fragments and the result is similar to the previous case. If the aircraft is impacted in the area with no crucial parts of the structure it is possible to continue in flight with partial operation limitation. In the third case the warhead explodes in the distance bigger then critical distance so there will be only partial damage of the structure caused by the single fragments. It is being assumed that single fragment is not able to completely destroy part of primary structure of the airframe. It is useless to define range of the damage in the first 55 L e t e c k ý z p r avo da j case. It is absolute. In the second case it will be possible to relatively easily localize area of the continuous damage of the airframe, to assess its range and to estimate degree of the structure damage. In the third and the last case a deeper analysis must be made to assess the range and the effect of the damage. This was theoretically made before here presented experiment. It was discovered by the theoretical analysis that for the assessment of this problem is the most suitable analysis the one which uses finite element method for simulation of the damage process. We chose typical structure part of airframe which is in this case reinforced skin panel and simplified model of the fragment. Then was assessed interaction between panel and moving model of the fragment. Experiment preparation and realization No theoretical solution could be held as correct if it is not at least partially validated by the experiment. Therefore we prepared and realized experiment which should validate our theoretical expectations. The most suitable and the most realistic would be an experiment with the real warhead and the part of the civil aircraft. This solution brought a lot of security and financial complications. There would be also limited possibilities for detail description of the damage caused by the particular fragments by this experiment because it would not be possible to define their exact shape, mass and velocity and therefore their kinetic energy. The variant of the experiment was chosen with shooting to the specimen with projectiles. These projectiles simulate effect of the fragments and their parameters are known. The part of the real structure of the aircraft was finally not chosen as the specimen because it would be difficult to obtain exact material characteristics of this structure. Therefore the new specimen was manufactured for this experiment. It fulfills all required parameters. Designed and manufactured specimen corresponds with the structure solution of the modern civil aircrafts, for example aircrafts made by company Airbus. It is part of the semi-monocoque 1/2011 The base of the specimen was the skin sheet with dimensions 1000x1000mm. Four stringers and perpendicular to them two frames were riveted on this skin. The frames are connected with stringers and the skin with connection elements. This structure design allows using frames with undivided flanges. We used cross section shape of the letter Z for the frames and the stringers. Completed specimens are shown on fig. 1. The space structure form L profiles were also manufactured. To this structure the specimen was connected during the experiment. There were used specially designed projectiles for the shooting. These projectiles very exactly simulate single fragments. It was steel projectiles FSP12,7 with mass 13,39±13g. Their shape is on fig. 2. The experiment was made on the shooting assembly which consists of several parts. The base is shooting stand where the weapon is fixed. There are two measure points for measuring of the velocity in the trajectory of the fragment. Using these two velocities we were able to calculate final impact velocity of the fragment. The last part of the shooting assembly was the specimen. As the weapon was used ballistic barrel with caliber 12,7mm. Complete shooting assembly is shown on fig. 3. As noted before, the experiment consists of shooting of the single fragments to the specimen. At the single attempts specimen was impacted with various velocities. We tried to hit also various combinations of the structure parts of the specimen. Various angles of the fragments strokes to the impacted area were also used. The base position was one when trajectory of the fragment was perpendicular to the plane of the specimen. For some shots were specimen turned 30 and 60 degrees. The impact velocity alternate by the different amount of gunpowder in the cartridge case from 609m/s to 1645m/s. Results of the experiment There were hit various node points of the structure with various velocities and directions of the strokes during the Figure 1 - The specimen Figure 3 - The shooting assembly structure which consists of skin reinforced by the stringers and frames. As a material still the most commonly used material in aircraft design was chosen, duralumin. experiment. Then were compared degrees of the structure damage. On the fig.4 is shown full penetration of the panel skin. There is obvious that fragment affected only small C z e c h A e r o s pa c e P r o c e e d i n g s area at this damage. Velocity which is expected by the real terroristic attack (700m/s) was used at this attempt. The size of the opening is equal to the size of the fragment. Also deformed area around the impact is very small. The next part that was intentionally hit by the fragment was the stringer. Flange of the stringer riveted to the skin was hit at this attempt. The result is similar to the one on the fig.7 also here. In the skin is opening equal to the size of the fragment and in the flange of the stringer but the damage is little bit bigger. This could be made by the influence of the material that is ripped out from the skin by the fragment. Affected area has now the size of three diameters of the fragment. There is destroyed and divided one flange of the stringer. The web of this stringer and the second flange remains untouched. The stringer was not completely damaged and it didn´t lost all its capabilities to carry loads. The similar damage as in the previous case could be observed also by the hit of the other structure parts as is shown on fig. 5. The penetration of the connection element and the frame is shown on this figure. Also only directly hit structure parts are damaged here. There is almost no influence to the rest of the structure caused by this fragment. The affected areas and degree of the damage are small. We discovered interesting phenomenon at the one attempt which we did not realized by the previous numerical calculation. This phenomenon is on fig. 6. Almost clear penetration of the various structure parts can be seen here. However, there is also damaged flange of the frame that was not directly hit by the fragment. This damage was caused by the material that was ripped out from the skin by the fragment. We can call this material as secondary fragment. This effect is very random. This has to be also incorporated in the range of the damage assessment. The experiment caused greatest damage to the structure by the hits of the beam webs (frames or stringers) at the first sight. This damage is on fig. 7. But it is only damage of the directly hit parts of the structure. There is no bigger damage of the other parts in this figure. Therefore it is the same type of the damage as in the previous cases. One of the tested parameters which influence impact of the fragments was velocity of the fragments in the moment of the stroke. Fig. 8 shows damage made by the fragments with the biggest velocity which allows our test equipment. The velocity was 1645m/s. If we compare this damage with the one shown on the fig. 5 (fragment velocity 822m/s) we can observe here obviously much bigger damage of the parts that are hit after the penetration of the skin. More severe damage was caused by bigger amount of the material ripped out of the skin and its bigger velocity obtained by this material from the fragment. If we compare these two damages from the point of view which describes ability of the structure to carry loads, we can conclude that the influences of the damages are basically the same. The flanges of the beams were destroyed in both cases. The other parts of the structure are untouched and they didn’t lose their ability to carry loads. Comparison with numerical model The results of the experiment were compared with the 56 numerical model created using finite element method. We improved numerical model using our experimental data. The visualization of the calculation result is on fig. 9. This example shows calculation of the skin penetration. If we compare this figure with fig. 7 there is obvious coincidence. The model shows opening with size similar to the size of the fragment and relatively small affected area around the penetration. With the coincidence of the calculated result with the experimental result we can say that the numerical model is sufficiently correctly made. The experimentally proved coincidence between theory and reality gives us guarantee of the correct calculation which will be with this model realized. Figure 9 - The Finite element model Conclusion During this experiment was proved that single fragment causes only very small damage and has only small effect to the ability of the structure to carry loads. Also more fragments which have impacted in some distance to each other damage more single parts. This doesn’t cause bigger damage that would result in aircraft crash. With the arrangement of the single fragments was proved that overall damage cannot be solved by the superposition of the partial damages of the single fragments. There was also discovered that range of the damage does not depend on the velocity of the fragments. However, degree of the damage always depends on the area of the impact and distance between impacted area and the warhead explosion. These parameters determine if there is continuous damage or separated damages made by single fragments. Results of this experiment were used for the improvement of the numerical calculation of the damage when big coincidence between theory and experiment was proved. References [1] Ludvík, F. - Konečný,P.: Rakety, VA Brno, 1998 [2] Ball, R.E.: The Fundamentals of Aircraft Combat Survivability Analysis and Design, AIAA Education Series, 1985. Experimental Assessment of the Fragments Effect to the Fuselage of the Aircraft Colour illustrations to the article published on pages 54 to 56. Figure 2 - The fragment Figure 6 - The secondary fragments Figure 4 - The damage of the skin Figure 7 - The Damage of the structure Figure 5 - The damage of the frame Figure 8 - The maximum fragment velocity Development of series production technology for stator and rotor blades Colour illustrations to the article published on pages 23 to 25. Fig. 2 - sample of rotor blade after curing and removal out of the mould Fig. 3 - three-dimensional CAD part of a blade Fig. 5 - high strength carbon fabric with plain weave Fig. 4 - unidirectional carbon fabric Fig. 6 - The rotor vane mold Fig. 7 - Compagination mold Fig. 10 - Original stattor parts Fig. 9 - silicone rubber Fig. 11 - Model of stator segment © A s s o c i at i o n o f t h e A v i at i o n M a n u fac t u r e r s