benchmarking ceasiom software to predict flight …

21
BENCHMARKING CEASIOM SOFTWARE TO PREDICT FLIGHT CONTROL AND FLYING QUALITIES OF THE B-747 A. Da Ronch , C. McFarlane ∗∗ , C. Beaverstock ∗∗ , J. Oppelstrup , M. Zhang , A.Rizzi Department of Engineering, University of Liverpool, Liverpool, UK ∗∗ Department of Aerospace Engineering, Bristol University, Bristol, UK Dept. of Aeronautical & Vehicle Engineering, Royal Institute of Technology (KTH), Stockholm, Sweden Keywords: Aircraft design, aerodynamics, flight dynamics, flight control, CFD. Abstract CEASIOM, the Computerized Environment for Aircraft Synthesis and Integrated Optimiza- tion Methods, is a framework that integrates discipline-specific tools for conceptual design. At this early stage of the design it is very use- ful to be able to predict the flying and handling qualities of the aircraft. In order to do this for the configuration being studied, the aerody- namic database needs to be computed and cou- pled to the stability and control tools to carry out the analysis. This paper describes how the adaptive-fidelity Computational Fluid-Dynamics module of CEASIOM computes the aerodynamic database of an aircraft configuration, and how that data is analyzed by the Flight Control Sys- tem Designer Toolkit module to determine the flying qualities and the control laws of the air- craft. The paper compares the predicted flying qualities with the flight-test data of the Boeing B747 aircraft in order to verify the goodness of the overall approach. 1 Introduction There is much current interest in using Compu- tational Fluid-Dynamics (CFD) to compute the static and dynamic forces and moments acting on an aircraft as a complement to the usual practice of measuring them in a wind-tunnel. In verify- ing the validity of this approach, the usual exer- cise is to compute the static and dynamic coeffi- cients or derivatives and compare them with the corresponding values measured in the wind tun- nel. Such a comparison reveals the differences and similarities between the two approaches, but says little about the sensitivity of the stability and control characteristics to these differences. A fur- ther step must be taken to provide insight on the overall accuracy of the CFD predictions of air- craft behaviour in flight. One way is to process the CFD-generated aero-data by linear and non- linear stability and control analysis and extract flying or handling qualities that can be compared with flight-test data. This is the approach taken in this paper. The tool used to do this is CEASIOM, the Computerized Environment for Aircraft Syn- thesis and Integrated Optimization Methods, cur- rently being developed within the frame of the SimSAC Project 1 , Simulating Aircraft Stability And Control Characteristics for Use in Concep- tual Design, sponsored by the European Com- mission 6th Framework Programme. CEASIOM uses adaptive-fidelity CFD to create the aerody- namic dataset and then uses it to analyze flying qualities using its stability and control analysis tools. The predicted flying qualities of the Boe- ing B747-100 aircraft are compared to those ob- served in flight tests [1]. 1 http://www.simsacdesign.eu 1

Upload: others

Post on 17-Mar-2022

3 views

Category:

Documents


0 download

TRANSCRIPT

BENCHMARKING CEASIOM SOFTWARE TO PREDICTFLIGHT CONTROL AND FLYING QUALITIES OF THE B-747

A. Da Ronch∗, C. McFarlane∗∗, C. Beaverstock∗∗, J. Oppelstrup†, M. Zhang†, A.Rizzi†∗ Department of Engineering, University of Liverpool, Liverpool, UK

∗∗ Department of Aerospace Engineering, Bristol University, Bristol, UK† Dept. of Aeronautical & Vehicle Engineering,

Royal Institute of Technology (KTH), Stockholm, Sweden

Keywords: Aircraft design, aerodynamics, flight dynamics, flight control, CFD.

Abstract

CEASIOM, the Computerized Environment forAircraft Synthesis and Integrated Optimiza-tion Methods, is a framework that integratesdiscipline-specific tools for conceptual design.At this early stage of the design it is very use-ful to be able to predict the flying and handlingqualities of the aircraft. In order to do thisfor the configuration being studied, the aerody-namic database needs to be computed and cou-pled to the stability and control tools to carryout the analysis. This paper describes how theadaptive-fidelity Computational Fluid-Dynamicsmodule of CEASIOM computes the aerodynamicdatabase of an aircraft configuration, and howthat data is analyzed by the Flight Control Sys-tem Designer Toolkit module to determine theflying qualities and the control laws of the air-craft. The paper compares the predicted flyingqualities with the flight-test data of the BoeingB747 aircraft in order to verify the goodness ofthe overall approach.

1 Introduction

There is much current interest in using Compu-tational Fluid-Dynamics (CFD) to compute thestatic and dynamic forces and moments acting onan aircraft as a complement to the usual practiceof measuring them in a wind-tunnel. In verify-ing the validity of this approach, the usual exer-

cise is to compute the static and dynamic coeffi-cients or derivatives and compare them with thecorresponding values measured in the wind tun-nel. Such a comparison reveals the differencesand similarities between the two approaches, butsays little about the sensitivity of the stability andcontrol characteristics to these differences. A fur-ther step must be taken to provide insight on theoverall accuracy of the CFD predictions of air-craft behaviour in flight. One way is to processthe CFD-generated aero-data by linear and non-linear stability and control analysis and extractflying or handling qualities that can be comparedwith flight-test data. This is the approach taken inthis paper. The tool used to do this is CEASIOM,the Computerized Environment for Aircraft Syn-thesis and Integrated Optimization Methods, cur-rently being developed within the frame of theSimSAC Project1, Simulating Aircraft StabilityAnd Control Characteristics for Use in Concep-tual Design, sponsored by the European Com-mission 6th Framework Programme. CEASIOMuses adaptive-fidelity CFD to create the aerody-namic dataset and then uses it to analyze flyingqualities using its stability and control analysistools. The predicted flying qualities of the Boe-ing B747-100 aircraft are compared to those ob-served in flight tests [1].

1http://www.simsacdesign.eu

1

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

2 CEASIOM

CEASIOM [2]is a framework tool that integratesdiscipline-specific tools comprising Computer-Aided Design (CAD) and mesh generation tools,Computational Fluid-Dynamics (CFD) codes,Stability and Control (S&C) softwares, . . . , allfor the purpose of aircraft conceptual design [3].An overview of the CEASIOM environment isprovided in Fig. 1. Significant components em-bodied in CEASIOM are

1. Geometry module, Geo–SUMO: a cus-tomized geometry construction system thatallows users to define the geometry in aXML file by a small number, on the orderof hundreds, parameters with intuitive in-terpretation

2. Aerodynamic module, AMB–CFD: usesadaptive fidelity CFD to replace orcomplement current handbook aerody-namic methods with the steady and un-steady TORNADO Vortex-Lattice Method(VLM) and CFD codes in Euler mode forlow speed and high-speed regimes, andsupport for external RANS (Reynolds Av-eraged Navier-Stokes) flow simulators forhigh-fidelity analysis of extreme flight con-ditions [4]

3. Stability and Control module, SDSA: asimulation and dynamic stability and con-trol analyzer and flying-quality assessor;it comprises six degrees of freedom testflight simulation, performance prediction,including human pilot model, StabilityAugmentation System (SAS), and a Lin-ear Quadratic Regulator (LQR) based flightcontrol system package [5] among manyother features [6]

4. Aero-elastic module, NeoCASS: quasi-analytical structural sizing and Doublet-Lattice Method (DLM), finite elementmodel generation [7]; linear aeroelasticanalysis and structural optimization [8];low fidelity panel methods usually adopted

Fig. 1 CEASIOM framework consists of geome-try, aerodynamic, structure and stability and con-trol modules

5. Flight Control System design module,FCSDT: a designer toolkit for flightcontrol–law formulation, simulation andtechnical decision support, permittingflight control system design philosophyand architecture to be coupled in the earlyphase of conceptual design [9]

6. Decision Support System module, DSS: anexplicit DSS functionality, including issuessuch as fault tolerance and failure tree anal-ysis [10]

The paper continues with a presentation of therelevant modules involved actively in this study.Emphasis is put on the automated generation ofthe Euler grid for the B747 model geometry.Aerodynamic sources are revised and a smartprocedure to reduce the total number of CFDsimulations required to fill-in the aerodynamic ta-bles outlined. Predictions of aerodynamic loadsat low and high speed are compared to exper-imental data. Then, results obtained with theFCSDT software are illustrated.

3 Geometry

The successful generation of aerodynamic ta-bles for flight dynamic applications using high-fidelity CFD tools depends strongly on the sys-tem capabilities to create a suitable mesh for thenumerical simulation in a reasonable amount of

2

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

time. Traditionally, the generation of a compu-tational mesh for a new aircraft configuration isan expensive process. During the design phase,the aircraft geometry is often modified iterativelyto improve certain aspects or avoid undesiderableaircraft behaviour. The generation of a new meshat every design iteration, if not automated, wouldbecome a bottleneck in the design phase. Cit-ing Dawes et al. [11], the question is "whetherCFD can participate in the design process withsufficient speed to drive down the design cycletime". Progress has been made in this direction,as exemplified by the customized geometry mod-eller for aircraft design, Vehicle Sketch Pad [12],which supports export of meshable surface mod-els and/or meshes in standard formats such asIGES, STEP, and CGNS.

3.1 SUMO

A very attractive feature of theCEASIOM framework is offered by the SUrfaceMOdeler, referred to as SUMO [13], for therapid generation of three-dimensional, water-tight aircraft geometries that are adequate forhigh-fidelity analysis using CFD. SUMO2 is agraphical tool aimed at rapid creation of aircraftgeometries and automatic surface mesh gener-ation. It is not a full-fledged CAD system, butrather a easy-to-use sketchpad, highly special-ized towards aircraft configurations. SUMO isbased on a C++ library for geometric primitivessuch as b-spline surfaces, and currently providesa graphical frontend to a small subset of thelibrary. It is actively developed in order tostreamline the workflow as far as possible tothe intended use: rapid surface modelling ofaircraft configurations. Unstructured surfacemeshes can be generated completely withoutuser intervention. Heuristics determine defaultparameters for the mesh generation code, whichusually yield a satisfactory mesh. Manualtuning of these parameters is possible as well.Triangulations are based on a three-dimensionalin-sphere criterion, which yields better mesh

2http://www.larosterna.com/dwfs.html

quality than Delaunay methods for stronglyskewed surfaces, such as thin, swept delta-wings.Geometric refinement criteria produce a finermesh in regions of strong curvature, while a limiton the minimum element size can be imposed toavoid resolution of irrelevant geometric detail.Unstructured volume meshes can be generatedfrom the surface mesh, using the tetrahedralmesh generator TetGen [14]. The volume meshcan be saved as CGNS, a bmsh file for the CFDsolver EDGE [15], or TetGen’s plain ASCIIformat.

Unstructured Euler grids for different de-flection of the canard wing were generated us-ing SUMO for the investigation of aerodynamiccharacteristics of a TRansonic Cruise, TCR, windtunnel model [16]. A recent study [17] presenteda viable route to generate high-quality RANSgrids using SUMO . The approach used in thisstudy stems from the geometry aircraft definitionbased on the CEASIOM XML file. The represen-tative geometry of the B747 model is illustratedin Fig. 2(a). The description is used to generatea triangular mesh on the surface, and at the nextlevel, the volume mesh is generated using tetra-hedral elements filling the space between the walland the farfield. This is shown in Fig. 2(b).

4 Aerodynamics

A prerequisite for realistic prediction of the sta-bility and control behaviour of an aircraft isthe availability of complete and accurate aerody-namic data. Traditionally, wind-tunnel measure-ments are used to fill look-up tables of forces andmoments over the flight envelope. These wind-tunnel models only become available late in thedesign cycle. To date, most engineering tools foraircraft design rely on handbook methods or lin-ear fluid mechanics assumptions. These methodsprovide low cost reliable data as long as the air-craft remains well within the limits of the flightenvelope. However, current trends in aircraft de-sign towards augmented-stability and expandedflight envelopes require an accurate descriptionof the non-linear flight-dynamic behaviour of theaircraft. CFD can be used as complement and

3

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

(a) SUMO geometry

(b) SUMO mesh

Fig. 2 Representative geometry of the B747model in SUMO

replacement of expensive experimental measure-ments.

The CEASIOM aerodynamic module, re-ferred to as the Aerodynamic Model Builder(AMB), has functionality to allow the generationof tables of aerodynamic forces and momentsrequired for flight dynamics analysis. The ap-proach is to use sampling and data fusion to gen-erate the tables, with a variety of sources of aero-dynamic data. These aspects are now described.

4.1 Semi-Empirical Method

The Data Compendium (DATCOM ) is a doc-ument of more than 1500 pages covering de-tailed methodologies for determining stabilityand control characteristics of a variety of air-craft configurations. In 1979, DATCOM wasprogrammed in Fortran and renamed the USAFstability and control digital DATCOM . Digi-tal DATCOM is a semi-empirical method whichcan rapidly produce the aerodynamic derivativesbased on geometry details and flight conditions.DATCOM was primarily developed to estimateaerodynamic derivatives of conventional config-urations [18, 19]. For a conventional aircraft,DATCOM gives all the individual components(body, wing, horizontal and vertical tail), and air-craft forces and moments. Digital DATCOM hasbeen implemented in AMB. A DATCOM inputfile is produced by interpreting and formatting theXML aircraft data. In addition the flight condi-tions of interest are added to the DATCOM file.

4.2 Potential Solver

TORNADO 3, a Vortex Lattice Method (VLM),is an open source Matlabc© implementation ofa modified horse-shoe vortex singularity methodfor computing steady and low reduced frequencytime-harmonic unsteady flows over wings. It canpredict a wide range of aircraft stability and con-trol aerodynamic derivatives [20]. The lifting sur-faces are created as unions of thin, not necessar-ily flat, quadrilateral surface segments. Effectsof airfoil camber are modeled by surface normal

3www.redhammer.se/tornado/

4

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

rotation. Leading edge control surfaces are mod-eled likewise. The modification to the horseshoevortices allows trailing edge control surface de-flection by actual mesh deformation. The steadywake can be chosen fixed in the body coordi-nate system or following the free stream. Over-all effects of compressibility at high Mach num-ber are assessed by the Prandtl-Glauert similarityrole (see Anderson [21]), and zero-lift drag esti-mates are obtained by Eckert’s flat plate analogy.

The fuselage might be modeled in the VLMusing several techniques. The simplest approxi-mation is modelling the aircraft fuselage as twoflat elements with shapes corresponding to itsplanar projections. It is found [22] that for ageneric aircraft configuration this model providesconsiderably different values of normal forceand pitching moment coefficient slopes with re-spect to experimental results. One alternativeconsists in using the slender body theory de-veloped by Munk [23]. The main assumption,which is met in traditional aircraft fuselage, isthat the diameter of the body is far less thanits length. However, some restrictions are con-nected to this theory, i.e. the body should havethe shape of a body of revolution which is a con-dition met with difficulty in a real aircraft. Then,the fuselage surface can be modeled by a num-ber of flat panels approximating its shape andthe strength of sink/source distribution in eachpanel is the solution of a system of linear equa-tions. This technique was tested for a three-dimensional representation of fuselage geome-try [22] and is adopted herein. The basic flowsolver is wrapped by user interfaces to create ta-bles of aerodynamic coefficients and derivativesfor export to flight simulators and flight controlsystem design software. To calculate the firstorder derivatives, TORNADO performs a centraldifference calculation using the pre-selected stateand disturbing it by a small amount. With thedistorting wake, non-linear effects will be visiblein some designs, especially in those where mainwing/stabiliser interactions are important. Thepanelling for TORNADO is automatically pro-duced within AMB from the XML geometry de-scription.

4.3 Euler Solver

EDGE 4 is a parallelized CFD flow solver sys-tem for solving 2D/3D viscous/inviscid, com-pressible flow problems on unstructured gridswith arbitrary elements [15]. EDGE can be usedfor both steady state and time accurate calcula-tions including manoeuvres and aeroelastic sim-ulations. The flow solver employs an edge-basedformulation which uses a node-centered finite-volume technique to solve the governing equa-tions. The control volumes are non-overlappingand are formed by a dual grid, which is com-puted from the control surfaces for each edge ofthe primary input mesh. In any Edge mesh, allthe mesh elements are connected through match-ing faces. In the flow solver, the governingequations are integrated explicitly towards steadystate with Runge-Kutta time integration. Conver-gence is accelerated using agglomeration multi-grid and implicit residual smoothing. Time accu-rate computations can be performed using a semi-implicit, dual time stepping scheme which ex-ploits convergence acceleration techniques via asteady state from inner iteration procedure. Sev-eral turbulence models [24, 25] are implementedin the solver.

A key feature in the present investigation isthe analysis with deflected control surfaces. Theway EDGE calculates the aerodynamics of con-trol surface deflections is based on the use of tran-spiration boundary conditions. In this approach,instead of moving the grid, the wall velocity com-ponent normal to the actual deflected surface isprescribed. Such approach eliminates the needof mesh deformation, thus all calculations can berun virtually on the clean configuration grid, buton the other hand, this imposes a limitation on theamount of maximum and minimum deflections.An alternative approach would be the generationof a different grid for every new configuration ofdeflected control surfaces. This approach is notfeasible for the intended use of creating aerody-namic tables given the number of possible com-binations of control surfaces and corresponding

4http://www.edge.foi.se/

5

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

deflection angles.In this paper, EDGE is used in Euler mode

with pre-set reasonable values for the maximumand minimum deflection of each defined controlsurface. All calculations were obtained using onesingle grid generated with SUMO .

4.4 CFD Interface

A Matlab c© graphical interface was developedto enhance the available aerodynamic sources in-cluded in the AMB, which were previously re-stricted to DATCOM and TORNADO . The CFDinterface [26] is extremely general and flexible,as illustrated in Fig. 3. The interface prepro-cesses the flow of informations generated in theAMB module and creates suitable input files tothe particular CFD solver used. The status of thejob is monitored and when successfully finished,the CFD output is postprocessed and converted ina format fully compatible with the Stability andControl tools embodied in CEASIOM . The sur-face and flowfield solution are then translated ina format compatible with TecplotR© [27]. Theoption of parallel calculation is available to re-duce the overall computational time. The itera-tive processAMB → CFD → AMB is likely tobe repeated several times during the generationof aerodynamic tables. A second CFD code, inaddition to EDGE , was interfaced to the AMB.This is the Parallel Multi-Block (PMB) solver, aresearch code under development at the Univer-sity of Liverpool [28]. When the flow conditionsfeature the onset and development of vorticalstructures, shock-induced separation and otherviscous-related phenomena, the task can be tack-led using the RANS equations. The PMB solvercan be used both in inviscid and viscous mode.The CFD interface has been tested extensivelyand used in several works [29, 30, 31].

A variety of calculations were obtained us-ing the CFD interface, including the flow simu-lation around a fixed aircraft configuration witha generic deflection for any available control sur-face (i.e., landing configuration with deploymentof multiple controls) and unsteady time-accurateanalysis for the model undergoing forced har-

AerodynamicModelBuilder

Flow SolverExternal CFD

set inputs

read outputs

Interface

Fig. 3 Interfacing AMB to a generic CFD solver

monic motions in the three body axes for the pre-diction of dynamic derivatives.

4.5 Sampling and Data Fusion Methods

The application of the aerodynamic predictionmethods in the current work is for the genera-tion of tables of forces and moments. This po-tentially entails a large number of calculations,which will be a particular problem if CFD is thesource of the data. This issue was addressed bysampling, reconstruction and data fusion in a pre-vious work [32]. Two scenarios were postulated,based on a requirement for tables for a com-pletely new design and for updating tables for ex-isting designs which are being altered. Fig. 4 il-lustrates the sampling and data fusing algorithms.

In the first scenario it is assumed that the re-quirement is for a high fidelity model and thatthis can be generated offline (i.e. the calcula-tion can be done overnight without a user waitingfor the model during an interactive session). Inthis scenario the emphasis is on sampling find-ing nonlinearities in the forces and moments.Two approaches to this sampling, based on theMean Squared Error criterion of Kriging and theExpected Improvement Function [33, 34], wereconsidered [32]. The second scenario is whena designer is involved in an interactive session.It is assumed that the aircraft geometry is incre-mented from an initial design, perhaps selectedfrom a library, and that a high fidelity modelis available for the initial design from the firstscenario. Data fusion (based on co-Kriging) isthen used to update this initial model, based ona small number of calculations at an acceptablecost (which at present rules out RANS). In this

6

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

Fig. 4 CEASIOM framework system illustratingthe sampling and data fusing algorithms

scenario it is assumed that the flow topology re-sulting from the initial geometry does not changeduring the geometry increments. If this is not thecase (i.e., the wing sweep angle increases so thatvortical flow starts to dominate at high angles),then either a new initial geometry needs to be se-lected, or the interactive session needs to be sus-pended so that a new high fidelity model can begenerated under scenario one.

Using these techniques it was shown that ta-bles could be generated in the order of 100 cal-culations under the first scenario and 10 calcula-tions under the second scenario.

4.6 CEASIOM Aerodynamic Table Format

The aerodynamic model considered within theCEASIOM framework is tabular in form. Themodel consists of tables of forces and momentsfor a set of aircraft states and controls which spanthe flight envelope. The aerodynamic table con-sists of as many rows as the number of possi-ble permutations of the aircraft states and con-trols, usually on the order of 103 or even 104 en-tries. Each entry in the table contains the 6 aero-dynamic coefficients in wind axis as function ofthe states and controls. The aircraft states fea-ture the angles of incidence (angle of attack,α,and sideslip,β), the Mach number,M, and theangular rates referred to the body axis,p, q andr, respectively, for the roll, pitch and yaw rates.CEASIOM was actively developed to handle un-conventional configurations with multiple controlsurfaces [9], enhancing the traditional aerody-namic model based on three conventional control

surfaces (ailerons, rudder and elevator). A vari-able number of columns in the table is reservedto accommodate the aircraft controls.

On the order of tens of CFD calculations arerequired to fill-in the aerodynamic tables for thestatic dependencies due to the angles of inci-dence and the deflection of control surfaces. Achallenging task is the computation of the dy-namic dependencies of the aerodynamic loadson the angular rates. Several options were in-vestigated. Quasi-steady dynamic derivativescan be obtained with DATCOM or, on the nextlevel of fidelity, TORNADO . Although cheapto compute, the validity of these estimationsis the primary concern when analyzing criticalflow conditions and unconventional configura-tions. CFD has the potential to provide accu-rate predictions of dynamic derivatives. The ac-curacy retained in the solution is obtained at theprice of large requirements in terms of com-putational time and resources. Unsteady time-accurate solutions with grid deformation at eachphysical time step to conform the applied har-monic motion are typically on the order of hun-dreds times more expensive than a steady statesolution. A well-established framework was pre-sented in Da Ronch et al. [35] for the estimationof dynamic derivatives using time-accurate solu-tions. Nonetheless, the described approach is nota viable option in the design phase and for thepresent study. A valid alternative to the time-marching method is offered by the HarmonicBalance (HB) method where the non-linear CFDsolution is truncated up to a predescribed num-ber of Fourier modes. As demonstrated in DaRonch et al. [36], two harmonics are generallyadequate to predict dynamic derivatives even atcritical flow conditions. Compared to the CPUtime required for an unsteady time-accurate solu-tion, the solution obtained using the HB methodis on the order of tens times faster [37], and about3 to 4 times more expensive than a steady stateanalysis. This option is of interest to fill-in theaerodynamic tables.

The approach followed in this study usesTORNADO for the prediction of quasi-steadydynamic derivatives. This was considered a rea-

7

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

sonable approximation because the interest is onthe linear and quasi-linear aerodynamics. Athigher Mach numbers, compressibility effectscan be significant. However, the range in the an-gle of attack at cruise speed is limited to about6.0◦. The idea of exploiting the potential ofthe HB approach was not adopted because thismethod was implemented within the structuredframework of the PMB solver. All the CFD-based calculations herein included were obtainedusing the unstructured EDGE solver.

5 The FCSDT module

The Flight Control System (FCS) is a multidisci-plinary design space that involves control effectorsizing and topology, as well as the systems archi-tecture that actuates these effectors. The FlightControl System Designer Toolkit (FCSDT) is asoftware framework to develop flight control sys-tems. The environment integrates modules to ad-dress each of the aforementioned design streams.The module Flight Control System Architecture(FCSA) is used to develop the effector topologyand systems architecture. This environment en-ables the user to edit the control effector topol-ogy, by adding and removing effectors, in ad-dition to design of the systems architecture foreach respective effector. The Stability and Con-trol Analyser Assessor (SCAA) module imple-ments the aerodynamic data from the AMB intoa SimulinkR© flight mechanics model, which isthen used to analyse the aircraft flight dynam-ics. Analysis variables include the trim values,associative linear results (which include time andfrequency domain results) and simulation results.This module also includes an Eigen structure as-signment algorithm, for design of a state feed-back controller. Eigen structure assignment isa control synthesis method which the input isa desired Eigen structure, or Eigen vectors andvalues, specifying the reduced state and controlinput matrices and control structure. The out-put is the control gains that yield the desiredEigen structure. More details on the mechanicsof Eigen structure assignment algorithm can befound in Liu and Patton [38]. There is an addi-

tional control design environment in LTIS, whichusesH∞ control synthesis to generate a robustfeedback controller.

6 B747 Model Example

The Boeing 747 is a very large four-enginedturbofun transport aircraft designed to ferrymore than 350 passengers on medium–long haulflights, which includes transatlantic operations.The aircraft, designed in the 1960s to meetgrowth demands for mass long haul transport, hasspanned several decades of operations, and is stillin service today. The aircraft surviving severalgenerations has spawned a number of derivativeswith varying operational requirements. The Boe-ing 747-100 employs a large number of individ-ual control surfaces, beyond the conventional el-evator, ailerons and rudder.

To obtain the necessary low speed character-istics the wing has triple–slotted trailing flaps andKrueger leading edge flaps. The Krueger flapsoutboard of the inboard nacelle are variable cam-bered and slotted while the inboard Krueger flapsare standard unslotted. Longitudinal control isobtained through four elevator segments and amovable stabilizer. The lateral control employsfive spoiler panels, an inboard aileron betweenthe inboard and outboard flaps, and an outboardaileron which operates with flaps down only oneach wing. The five spoiler panels on each wingalso operate symmetrically as speedbreaker inconjunction with the most inboard sixth spoilerpanel. Directional control is obtained from tworudder segments [39]. An aircraft model using amodel generated for the Boeing 747-100 has beenselected for investigation of its flight dynamicswithin FCSDT. This is to validate FCSDT for useon multiple surface aircraft examples. The geom-etry of the B747 model is represented in Fig. 5 fordifferent fidelity-level approximations. The sim-plest approximation of the model, which is suit-able for VLM-based calculations, consists of thelifting surfaces only. The TORNADO geometryis shown in Fig. 5(a). Lifting surfaces weresized to closely match the B747 geometry andits control surfaces. Comparison between the

8

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

TORNADO geometry and the B747 geometry ofthe real aircraft, as illustrated in Fig. 6, showsthat the simplified model reproduce well the mainfeatures of the model geometry. Then, the base-line geometry was enhanced to include the influ-ence of the fuselage which is modelled as a slen-der body in TORNADO calculations. The cor-responding geometry is illustrated in Fig. 5(b).Lifting surfaces were not modified to accommo-date the fuselage. On the top of the fidelity–stair,the watertight geometry used in CFD calculationsshown in Fig. 5(c) was generated with SUMO .The surface geometry was simplified by remov-ing the engine nacelles and adopting a simplerepresentation of the fairings. These were consid-ered reasonable simplifications because the inter-est is on the flow development on the upper liftingsurfaces.

Both TORNADO and EDGE geometriesconsist of control surfaces which were sizedfrom the aircraft model, including the innerand outer ailerons, all-movable stabilizer, two–segment elevator and two–segment rudder. Thesecontrol surfaces are highlighted in Fig. 5(c).

Details of the surface grid are shown in Fig. 7.The surface mesh does not conform to hingelines, but surface elements affected by the rota-tion are mapped into the CFD solver EDGE , andthese are coloured. The geometry has sharp trail-ing edges which are adequate for inviscid flowmodels. For a RANS model, the proper mod-elling of trailing edges of lifting surfaces is re-quired. Also, a detailed RANS model would in-clude the resolution of the opening gaps and ex-posed edges of a deflected control device.

6.1 Validation of Aerodynamic Sources

This section summarizes predictions of aerody-namic loads based on several aerodynamic op-tions with wind tunnel experimental data pro-vided by the manufacturer [1]. At low speed,predictions obtained with TORNADO are con-sidered. At higher speed featuring compressibil-ity effects and formation of shock waves, low–order and CFD-based results are presented.

(a) TORNADO geometry

(b) TORNADO geometry with fuselage

(c) Watertight geometry for CFD analysis

Fig. 5 Geometry of the B747 model for severalaerodynamic options for increasing fidelity

9

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

(a) Side view

(b) Front view

(c) Top view

Fig. 6 Comparison of the TORNADO geometryin Fig. 5(a) with a three-view drawing of theB747 model

(a) Tailplane and control surfaces

(b) Fairings and control surfaces on main wing

Fig. 7 Surface mesh for the B747 model

10

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

6.1.1 Low Speed Aerodynamics

Predictions of lift and pitching moment coef-ficients are illustrated in Fig. 8. A signifi-cant change in the force slope is observed be-tween 10.0◦ and 15.0◦ in the experimental dataset. The moment coefficient is also non-linearabove the static stall. Numerical predictions ofthe lift coefficient in Fig. 8(a) were obtainedwith TORNADO . The data set labelled as"TORNADO Baseline" is based on the originalformulation of the VLM, which predicts a lin-ear increase in the aerodynamic coefficients forincreasing angle of attack. Then, the option toaccount for a viscous correction was switchedon and the corresponding data set is referred toas "TORNADO Viscous". The stall prediction isbased on informations ofdCL /d α andCL0. Thetwo numerical data sets start to diverge at around10.0◦, with the viscous calculation underpredict-ing the static stall. Fig. 8(b) illustrates for thepitching moment coefficient the effect of mod-elling the fuselage with sink/source singularitiesin the TORNADO calculations. The resulting ef-fect of including the fuselage is to offset the dataset obtained for the Wing-Vertical-Horizontal tail"WVH" configuration (see Fig. 5(a) for the ge-ometry) by a certain amount. Although theoverall agreement is improved over the rangein the angle of attack for the Fuselage-Wing-Vertical-Horizontal tail "FWVH" configuration(see Fig. 5(b) for the geometry), a poorer agree-ment is achieved at low angles of attack.

One example describing the authority ofthe control surfaces deflection is provided inFig. 9. Numerical calculations were obtainedusing TORNADO with the viscous correctionswitched off and the clean configuration of liftingsurfaces. The outboard elevator was deflected.The comparison between numerical and experi-mental data sets is reasonable. The slope of theforce coefficient is slightly underpredicted, whilea good prediction is provided for the moment co-efficient.

AoA [deg]

CL

-5 0 5 10 15 20 25-0.4

0

0.4

0.8

1.2

1.6

2TORNADO BaselineTORNADO ViscousExp Data

(a) Lift coefficient

AoA [deg]

Cm

-5 0 5 10 15 20 25-1.2

-0.9

-0.6

-0.3

0

0.3

0.6TORNADO FWVHTORNADO WVHExp Data

(b) Pitching moment coefficient

Fig. 8 Prediction of lift and pitching moment co-efficients for several values of angle of attack atMach number of 0.15

11

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

δelev [deg]

CL

0 2 4 6 8 100.03

0.04

0.05

0.06

0.07

0.08 TORNADOExp Data

(a) Lift coefficient

δelev [deg]

Cm

0 2 4 6 8 10-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12 TORNADOExp Data

(b) Pitching moment coefficient

Fig. 9 Prediction of lift and pitching moment co-efficients for several values of outboard elevatordeflection at Mach number of 0.15

6.1.2 High Speed Aerodynamics

Fig. 10 shows the lift curve slope producedby the various fidelity-level aerodynamic codes.The Euler results show the closest correlationto the Boeing 747-100 published data. The ac-tual values and the curve slope are compare wellto experimental values. DATCOM shows com-parable lift-curve slope, with the actual valuesslightly less than experimental data. However,as Mach number increases, DATCOM resultsbegin to fall away from experimental re-sults. TORNADO results without compress-ibility correction remain constant through allMach numbers, and the offset from experimen-tal data increases because the more influen-tial effect of compressibility at higher speeds.TORNADO results with the Prandtl-Glauert sim-ilarity role overpredict the lift-curve slope, di-verging at higher Mach numbers. A reason forthe poor correlation to experimental data is likelyto be caused by the simple compressibility cor-rection which begins to break down at the higherend of the transonic regime.

Fig. 11 shows the drag polars computedby the various codes at the Mach numbersspecified. It is observed that DATCOM resultshows the best correlation with the NASA ex-perimental data. This is not unexpected be-cause DATCOM was developed using conven-tional configurations and the semi-empiricalmethods calibrated using, among others, the Boe-ing 747. Thereby, the parasitic drag term,CD0, isestimated using parameters such as wetted area.The Euler results differ from the viscous ex-perimental data in the absence of any estima-tion of the drag due to friction. As the angleof attack is increased, drag divergence is pre-dicted. TORNADO results achieve a poor agree-ment with other data sets.

Fig. 12 illustrates the pitching moment coef-ficient for several values of the angle of attack,which is a good indicator for the aircraft staticstability. At lower Mach numbers, a good corre-lation of numerical data sets is achieved in termsof stability. However, the numerical values de-viate by a constant offset, which suggests dis-

12

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

AoA [deg]

CL

0 1 2 3 4 5 6 7 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1DATCOMTORNADOTORNADO CompEDGEExp Data

(a) M = 0.75

AoA [deg]

CL

0 1 2 3 4 5 6 7 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1DATCOMTORNADOTORNADO CompEDGEExp Data

(b) M = 0.80

AoA [deg]

CL

0 1 2 3 4 5 6 7 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1TORNADOTORNADO CompEDGEExp Data

(c) M = 0.85

Fig. 10 Prediction of lift coefficient for severalvalues of angle of attack in transonic regime

CL

CD

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0

0.02

0.04

0.06

0.08

0.1

0.12DATCOMTORNADOTORNADO CompEDGEExp Data

(a) M = 0.75

CL

CD

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0

0.02

0.04

0.06

0.08

0.1

0.12DATCOMTORNADOTORNADO CompEDGEExp Data

(b) M = 0.80

CL

CD

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0

0.02

0.04

0.06

0.08

0.1

0.12TORNADOTORNADO CompEDGEExp Data

(c) M = 0.85

Fig. 11 Prediction of lift versus drag coefficientin transonic regime

13

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

crepancies in theCm0 value. This is expectedbecause the term is highly dependent on the air-foil section and fuselage geometry used in thecomputations. TORNADO with compressibilitycorrection overpredicts the curve-slope, amplify-ing the compressibility effect on the static stabil-ity. The Euler results achieve a good compari-son with experimental data for the Mach numbersconsidered. At this point the trend line for the ex-perimental data set has non-linear behaviour, thestatic stability migrating toward neutral stability.This could be attributed to an effect associatedto non-linear effects induced by viscous phenom-ena.

The overall performance of the Euler re-sults is satisfactory and shows the need of usingCFD to obtain good predictions in the transonicregime, where results based on traditional meth-ods become suspect.

The surface pressure distribution on the B747model is illustrated in Fig. 13 for several flowconditions and geometry configurations. Thesetest cases were taken from the database of CFDcalculations that were run for the generation oflook-up tables. A solution obtained on the cleanconfiguration is depicted in Fig. 13(a) at angle ofincidence of 6.0◦ and Mach number of 0.8. Astrong shock wave forms on the upper main wingand extends downstream until 50% of the localchord, not interacting with the control surfaces.The remaining two subfigures were obtained forthe same flow conditions, i.e., angle of incidenceof 1.0◦ and Mach number of 0.9, but for differentgeometry configurations. The control sign con-vention adopted within CEASIOM framework isas defined by Cook [40], which is a positive con-trol surface displacement gives rise to a negativeaeroplane response. Fig. 13(b) shows the solu-tion obtained imposing a positive deflection tothe four elevator segments (trailing-edge down).A shock wave appears near the hinge axis wherethe flow suddenly expands to conform to the con-trol deflection. The two rudder segments weredeflected by a negative angle (trailing-edge star-board) in the solution illustrated in Fig. 13(c).The resulting flow is locally accelerated near therudder hinge line.

AoA [deg]

Cm

0 1 2 3 4 5 6 7 8-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8DATCOMTORNADOTORNADO CompEDGEExp Data

(a) M = 0.75

AoA [deg]

Cm

0 1 2 3 4 5 6 7 8-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8DATCOMTORNADOTORNADO CompEDGEExp Data

(b) M = 0.80

AoA [deg]

Cm

0 1 2 3 4 5 6 7 8-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8TORNADOTORNADO CompEDGEExp Data

(c) M = 0.85

Fig. 12 Prediction of pitching moment coeffi-cient for several values of angle of attack in tran-sonic regime

14

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

(a) M = 0.8, α = 6.0◦

(b) M = 0.9, α = 1.0◦, positive elevator deflection

(c) M = 0.9, α = 1.0◦, negative rudder deflection

Fig. 13 Surface pressure distribution for theB747 model geometry at several angles of attack,α, Mach numbers,M, and configurations

6.2 Stability Analysis

The non-linear set of equations governing therigid aircraft motion are first linearized with re-spect to an equilibrium (trim) point. The solutionof a standard eigenvalue problem, expressed inthe classical form(A − λI) x, is solved for theeigenvalues,λi, and corresponding eigenvectors,xi. The modes of motion are then identified inan automatic way from the solution of the eigen-vector problem, though the identification mightfail when eigenvalues are non-conjugate for somemodes. The identified modes for the B747 exam-ple are illustrated in Fig. 14.

The short period mode is characterized byrapid oscillations in angle of attack about a nearlyconstant flight path. This mode can be excitedby rapidly deflecting the elevator, and usually itis fast and well damped. In simulated flight itcan be excited by initial disturbance in pitch rate.The time to halve the motion amplitude is plot-ted against the period of oscillation in Fig. 14(a).Numerical results are overplotted to ICAO rec-ommendations.

The phugoid mode is characterized by veryslow oscillations in pitch angle and velocity, anda nearly constant angle of attack. To excitethis mode, the airplane should be trimmed inlevel flight and next stick should be pulled backslightly and maintained. This pitches the airplaneup into a climb. As the airplane climbs, it losesspeed and lift, causing it to gradually pitch down-ward and enter a dive. During the dive, the air-plane gains speed and lift, bringing it back into aclimb. The main aerodynamic forces taking partin inducing a phugoid are lift and drag forces.The lift is the oscillation inducer and the drag isa damper. The phugoid period is illustrated inFig. 14(b).

The Dutch-roll mode is moderately involvingfast side-to-side swaying of the aircraft. It in-volves oscillations in bank, yaw, and sideslip an-gles. In a flight-state display, a Dutch roll will beindicated by the velocity vector circle oscillatingfrom side to side. This mode can be excited byrapidly deflecting the rudder, and should be fastand well damped. In simulated flight it can be ex-

15

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

cited by initial disturbing of sideslip angle. Nu-merical results for the Dutch-roll mode are shownin Fig. 14(c).

6.3 Linear Model Analysis

Fig. 15 presents the trimmed values for the an-gle of incidence and elevator angle. In general,as Mach number increases, the required angleof incidence decreases, and the required eleva-tor angle increases. This suggests that a lowerangle of incidence is required for the requiredCL, primarily generated by the main lifting sur-face, the elevator angle providing the necessarypitching moment to balance the aircraft. Thisalso implies that the pitching moment generatedby the elevators, with little influence on the lift.DATCOM results suggest that the influence ofthe elevators is significant on the vehicular lift,leading to a smaller efficiency value orCL/Cm.

Fig. 16 shows the pole plots generated. It isdifficult to interpret these without also inspect-ing the eigen vectors in tandem. The figure pri-marily shows the differences in the higher fre-quency modes, and will be the focus of this briefdiscussion. The SCAA module, using a func-tion to compare the eigen vectors, identifies thehigher frequency complex poles as the short pe-riod, with the relatively lower frequency poles asthe Dutch-roll. The results show large discrep-ancies with correlation of the TORNADO resultwith compressibility effect showing the greatestdifference in short period prediction. The migra-tion with Mach number all suggest a similar trendof increasing natural undamped frequency.

However, these can all be made to behavedynamically, within the linear region, similar toone another. The control synthesis method, eigenstructure assignment, is used to compute gainvalues for a feedback controller. The valuescomputed specify a eigen value of−2 ± 2i forthe short period mode, for a feedback controllerA + BK. A andB are known, plant and controlmatrix, respectively. Fig. 17 shows the variationsin feedback gain values using the available aero-dynamic sources.Kα refers to the gain value offeedback angle of attack to elevator, andKq is

(a) Short period characteristics

(b) Phugoid characteristics

(c) Dutch roll characteristics

Fig. 14 Flight dynamic properties of the B747aircraft at cruise speed

16

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

Mach

Ao

A[d

eg]

0.75 0.8 0.85 0.9-1

0

1

2

3

4

5

6

7

8

9

10DATCOMTORNADOTORNADO CompEDGEExp Data

(a) Angle of attack

Mach

δ ele

[deg

]

0.75 0.8 0.85 0.9-2

-1

0

1

2

3

4

5

6

7

8

9

10DATCOMTORNADOTORNADO CompEDGEExp Data

(b) Elevator deflection

Fig. 15 Trimmed angle of incidence and elevatorangles

Real

Imag

-1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25-3

-2

-1

0

1

2

3

4

5DATCOMTORNADOTORNADO CompEDGEExp Data

(a) M = 0.75

Real

Imag

-1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25-3

-2

-1

0

1

2

3

4

5DATCOMTORNADOTORNADO CompEDGEExp Data

(b) M = 0.80

Real

Imag

-1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25-3

-2

-1

0

1

2

3

4

5DATCOMTORNADOTORNADO CompEDGEExp Data

(c) M = 0.85

Fig. 16 Pole plots for various angles of attack

17

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

Mach = 0.75 Mach = 0.8 Mach = 0.85 Mach = 0.9−50

0

50

100

150

DATCOM Tornado Tornado with comp. Euler NASA

Mach = 0.75 Mach = 0.8 Mach = 0.85 Mach = 0.90

20

40

60

80

Kq

Fig. 17 Feedback gain values computed by EigenStructure Assignment control synthesis

the pitch damping, or pitch rate feedback to ele-vator. This can similarly be used to specify thecomplete eigen structure, limited by the numberof inputs available.

6.4 Single Rudder Surface Failure

The following brief example models a rudderfailure of a single rudder surface, at an altitude of11000m and a range of Mach numbers. The trimis obtained by computing the required angle ofincidence and bank angle states for steady stateconstant altitude and velocity. This is achievedthrough computing the required throttle, eleva-tors, ailerons and rudder, where the bottom sur-faces is failed and stuck at -10.0◦. Fig. 18 showsthe state values and the computed input values.The top-left subfigure plots the angle of attack insolid line and the roll angle in dashed line.

Fig. 19 presents the pole plot for this failurecase. The traditional complex modes of shortperiod, phugoid and dutch roll can be readilyidentified. However, it is not easy to automati-cally identify these by way of an algorithm, ascross coupling, or involvement of states that arenot typically associated to these modes is ob-served. The dynamics of the longitudinal com-plex modes also observes lateral modes and visa

0.75 0.8 0.85 0.90

1

2

3

α (d

eg)

0.75 0.8 0.85 0.9−0.2

0

0.2

0.4

Mach

φ (d

eg)

0.75 0.8 0.85 0.9−1

−0.8

−0.6

−0.4

−0.2

0

Mach

δ ele (

deg)

0.75 0.8 0.85 0.9−1.6

−1.4

−1.2

−1

−0.8

−0.6

Mach

ξ inne

r ai

l (de

g)

0.75 0.8 0.85 0.94

5

6

7

8

Mach

ζ low

er r

udd

(de

g)

Fig. 18 Trim values computed for Boeing 747-100 with failed lower rudder at -10.0◦

versa. This leads to behaviour in the short periodand phugoid, where variations in roll and yaw oc-cur, where the Dutch-roll might now have a littlepitching motion.

For each of these conditions, it would be pos-sible to generate a controller to remove abnormalbehaviour, and return the modes to more tradi-tional modes. This would make the aircraft be-have with normal handling in the event of fail-ures. Such a comprehensive controller is seldomdesigned due to the large number of flight con-figurations, conditions and failure modes. Ro-bust controllers are preferable, where a limitednumber of controllers are designed, for a com-plete flight envelope. A tool built by TsAGI forFCSDT integrates such a design tool, which usesH∞ control synthesis to design robust controllers.

7 Conclusions

Recent advances in the generation of aerody-namic tables for flight dynamics are described inthis paper using the CEASIOM software, whichwas developed within the frame of the SimSACProject, Simulating Aircraft Stability And Con-trol Characteristics for Use in Conceptual De-sign, sponsored by the European Commission 6thFramework Programme. Aerodynamic tables intabular form were generated using several aero-

18

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

Real

Imag

-2 -1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5 M=0.75M=0.80M=0.85M=0.90

-0.06 -0.04 -0.02 0 0.02-0.1

-0.05

0

0.05

0.1

Fig. 19 Pole plot for the Boeing 747-100 withfailed lower rudder at -10.0◦

dynamic codes, from semi-empirical and linearpotential methods to CFD. A smart procedurewas considered to fuse data obtained with differ-ent fidelity-level aerodynamics. A database wasthen generated for low and high speed aerody-namics. The test case is the Boeing 747-100. Thestudy demonstrates the establishment of a robustand automated procedure, with the aircraft ge-ometry description defined using on the order ofhundreds of parameters as the starting point andthe look-up aerodynamic tables based on Eulercalculations as the ending point. This chain em-bodies major achievements. The automated gen-eration of volume unstructured grids stems fromthe aircraft geometry description. This step is au-tomated, not requiring user’s intervention in mostof the cases. Realistic control surfaces systemwas sized based on the aircraft geometry modeland included in all numerical models. ParallelCFD calculations can be routinely launched tofill-in aerodynamic tables. Aircraft states andcontrols are selected within the flight envelope inan efficient manner, with the benefit of reducingthe total number of CFD analysis. This step isalso automated. The user can define deliberatelyadditional critical flow conditions to analyze incase the sampling algorithm failed to resolve im-portant non-linear features in aerodynamic loads.

Using the aerodynamic database, the trim solu-tion was computed for the different aerodynamicsources. In transonic regime with compressibil-ity effects and formation of shock waves, the Eu-ler results were in agreement with experimentaldata. Eigen values were calculated for severalMach numbers and results were compared to ex-perimental data, demonstrating the need of CFDto complement low-order methods. A feedbackcontroller was designed using the eigenstructureassignment to migrate the poles to a specified lo-cation. Then, a case with a failed lower segmentrudder was examined. A new trim condition wasfound and pole plots presented for several Machnumbers.

8 Acknowledgments

The financial support by the European Commis-sion through co-funding of the FP6 project Sim-SAC is gratefully acknowledged.

9 Copyright Statement

The authors confirm that they, and/or their com-pany or institution, hold copyright on all of theoriginal material included in their paper. Theyalso confirm they have obtained permission, fromthe copyright holder of any third party materialincluded in their paper, to publish it as part oftheir paper. The authors grant full permission forthe publication and distribution of their paper aspart of the ICAS2010 proceedings or as individ-ual off-prints from the proceedings.

References

[1] Rodney, C. H. and Nordwall, D. R., “The Simu-lation of a Jumbo Jet Transport Aircraft. VolumeII. Modelling Data (B747),” The Boeing Com-pany, Wichita Division, 1970.

[2] Von Kaenel, R., Rizzi, A., Oppelstrup, J.,Grabowski, T., Ghoreyshi, M., Cavagna, L., andBérard, A., “CEASIOM: Simulating Stability& Control with CFD/CSM in Aircraft Concep-tual Design,”26th International Congress of theAeronautical Sciences, ICAS, 2008.

19

A. DA RONCH∗, C. MCFARLANE∗∗, C. BEAVERSTOCK∗∗, J. OPPELSTRUP†, M. ZHANG†, A.RIZZI†

[3] Rizzi, A., “Modeling and Simulating AircraftStability & Control – the SimSAC Project,”AIAA Guidance, Navigation and Control Con-ference, AIAA–2010–8238, Toronto, Ontario,2010.

[4] Ghoreyshi, M., Vallespin, D., Da Ronch, A.,and Badcock, K. J., “Simulation of Aircraft Ma-noeuvres Based on Computational Fluid Dy-namics,”AIAA Guidance, Navigation and Con-trol Conference, AIAA–2010–8239, Toronto,Ontario, 2010.

[5] Dul, F. A., Grabowski, T., and Stefanek, L.,“Implementation of the Basic Linear QuadraticGaussian Control within the Flight Control Sys-tem,” SimSAC Official Deliverable, D 5.3-1,2008.

[6] Grabowski, T. G., Mieszalski, D., andMarcinkiewincz, E., “Stability Analysis inConceptual Design Using SDSA Tool,”AIAAGuidance, Navigation and Control Conference,AIAA–2010–8242, Toronto, Ontario, 2010.

[7] Da Ronch, A., “GUESS: Generic Unknown Es-timator for Aircraft Structural Sizing,” MasterThesis, Royal Institute of Technology (KTH),Stockholm, Sweden, October 2008.

[8] Cavagna, L., Ricci, S., and Travaglini, L., “Neo-CASS: an Integrated Tool for Structural Siz-ing, Aeroelastic Analysis and MDO at Con-ceptual Design Level,”AIAA Guidance, Navi-gation and Control Conference, AIAA–2010–8241, Toronto, Ontario, 2010.

[9] McFarlane, C., Richardson, T. S., Da Ronch,A., and Badcock, K. J., “Comparison of Con-ventional and Asymmetric Aircraft Configura-tions Using CEASIOM,”AIAA Guidance, Nav-igation and Control Conference, AIAA–2010–8243, Toronto, Ontario, 2010.

[10] Maheri, A., “Designer Decision Support System(DDSS) Development and Overview,” SimSACOfficial Deliverable, D 5.3-3, 2010.

[11] Dawes, W. N., Dhanasekaran, P. C., Demargne,A. A. J., Kellar, W. P., and Savill, A. M.,“Reducing Bottlenecks in the CAD-to-Mesh-to-Solution Cycle Time to Allow CFD to Partic-ipate in Design,”Journal of Turbomachinery,Vol. 123, No. 3, 2001, pp. 552–557.

[12] Hahn, A. S., “Vehicle Sketch Pad: A ParametricGeometry Modeler for Conceptual Aircraft De-

sign,” 48th AIAA Aerospace Sciences Meeting,Orlando, Florida, January 2010. AIAA 2010-657.

[13] Eller, D., “Mesh Generation Using SUMO andTetGen,” SimSAC Official Deliverable, D 2.3-5,2009.

[14] Si, H., “On Refinement of Constrained De-launay Tetrahedralizations,” In Proceedings ofthe 15th International Meshing Roundtable,September 2006. Software available fromhttp://tetgen.berlios.de, July 2009.

[15] Eliasson, P., “EDGE, a Navier-Stokes Solverfor Unstructured Grids,” FOI Report, FOI–R–0298–SE, 2005.

[16] Eliasson, P., Vos, J. B., Da Ronch, A., Meng-meng, Z., and Rizzi, A., “Virtual Aircraft De-sign of TransCRuiser – Computing Break Pointsin Pitch Moment Curve,”28th AIAA AppliedAerodynamic Conference, AIAA–2010–4366,2010.

[17] Tomac, M., Rizzi, A., and Oppelstrup, J.,“From Geometry to CFD Grids - An Auto-mated Approach for Conceptual Design,” AIAAGuidance, Navigation and Control Conference,AIAA-2010-8240, 2-5 August 2010, Toronto,Ontario.

[18] Razgonyaev, V. and Mason, W. H., “An Evalu-ation of Aerodynamic Prediction Methods Ap-plied to the XB-70 for Use in High SpeedAircraft Stability and Control System Design,”33rd Aerospace Sciences Meeting and Exhibit,AIAA95-0759, Reno, Nevada, 1995.

[19] Williams, J. E. and Vukelich, S. R., “The USAFStability and Control Digital DATCOM,” Mc-Donnell Douglas Astona UTICS Company, StLouis Division, St Louis, Missouri, 63166,1979, AFFDL-TR-79-3032.

[20] Melin, T., “TORNADO: a Vortex–Lattice MAT-LAB Implementation for Linear AerodynamicWing Applications,” Master Thesis, Royal Insi-tute of Technology (KTH), Department of Aero-nautics, Sweden, December, 2000.

[21] Anderson, J. D., “Modern Compressible Flow,”McGraw Hill, 1990.

[22] Khrabrov, A., Kolino, K., Glazkov, S., Sido-ryuk, M., Bolyrev, S., and Anoshin, J., “Reviewof Steady and Unsteady Panel Codes and TierI Simulation of Aircraft Stability and Control

20

Benchmarking CEASIOM Software to Predict Flight Control and Flying Qualities of the B-747

Characteristics Using Vortex Lattice Method,”SimSAC Official Deliverable, D 3.2-1, 2007.

[23] Munk, M., “The Aerodynamic Forces on Air-ship Hull,” NACA Report No. 184, 1924.

[24] Wallin, S. and Johansson, A. V., “An ExplicitAlgebraic Reynolds Stress Model for Incom-pressible and Compressible Turbulent Flows,”Journal of Fluid Mechanics, Vol. 403, 2000,pp. 89–132.

[25] Hellsten, A., “New Advancedk−ω TurbulenceModel for High Lift Aerodynamics,” AIAAJournal, Vol. 43, No. 9, 2005, pp. 1857–1869.

[26] Da Ronch, A., “CFD Start Documentation,”SimSAC Official Deliverable, D 3.2-6, 2009.

[27] “Tecplot. User’s Manual. Version 10,” AmtecEngineering, Inc., Bellevue, Washington,September 2003.

[28] Badcock, K. J., Richards, B. E., and Woodgate,M. A., “Elements of Computational Fluid Dy-namics on Block Structured Grids using Im-plicit Solvers,”Progress in Aerospace Sciences,Vol. 36, 2000, pp. 351–392.

[29] Da Ronch, A., Ghoreyshi, M., Badcock, K. J.,Mengmeng, Z., Oppelstrup, J., and Rizzi, A.,“On Time Optimal Manoeuvres of Aicraft Us-ing CFD–Based Tabular Data with MultipleControl Surfaces,” Abstract Submitted to the49th AIAA Aerospace Science Meeting, 4–7January 2010, Orlando, FL.

[30] Ghoreyshi, M., Da Ronch, A., Badcock, K. J.,Dees, J., Bérard, A., and Rizzi, A., “Aerody-namic Modelling for Flight Dynamics Analy-sis of Conceptual Aircraft Designs,” 27th AIAAApplied Aerodynamics Conference, AIAA-2009-4121, 22-25 June 2009, San Antonio, TX.

[31] Ghoreyshi, M., Badcock, K. J., Da Ronch,A. Marques, S., Swift, A., and Ames, N.,“Framework for Estabilishing the Limits of Tab-ular Aerodynamic Models for Flight DynamicsAnalysis,” Presented as AIAA-2009-4121, Un-der Review for Publication in Journal of Air-craft.

[32] Ghoreyshi, M., Badcock, K. J., and Woodgate,M. A., “Accelerating the Numerical Genera-tion of Aerodynamic Models for Flight Simula-tion,” Journal of Aircraft, Vol. 46, No. 3, 2009,pp. 972–980.

[33] Lophaven, S. N., Nielsen, H. B., and Sønder-

gaard, J., “DACE, A Matlab Kriging Toolbox.Version 2.0,” Technical Report IMM-TR-2002-12, Technical University of Denmark, August 1,2002.

[34] Lophaven, S. N., Nielsen, H. B., and Søn-dergaard, J., “Aspects of the Matlab ToolboxDACE,” Technical Report IMM-REP-2002-12.Technical University of Denmark, August 1,2002.

[35] Da Ronch, A., Vallespin, D., Ghoreyshi, M.,and Badcock, K. J., “Computation of DynamicDerivatives Using CFD,”28th AIAA AppliedAerodynamics Conference, AIAA–2010–4817,Chicago, IL, 2010.

[36] Da Ronch, A., Ghoreyshi, M., Badcock, K. J.,Görtz, S., Widhalm, M., Dwight, R. P., andCampobasso, M. S., “Linear Frequency Domainand Harmonic Balance Predictions of DynamicDerivatives,” 28th AIAA Applied AerodynamicConference, AIAA–2010–4699, Chicago, IL,2010.

[37] Mialon, B., Khrabrov, A., Da Ronch, A., Cav-agna, L., Mengmeng, Z., and Ricci, S., “Bench-marking the Prediction of Dynamic Deriva-tives: Wind Tunnel Tests, Validation, Accel-eration Methods,”AIAA Guidance, Navigationand Control Conference, AIAA–2010–8244,Toronto, Ontario, 2010.

[38] Liu, G. P. and Patton, R. J., “Eigenstructure As-signment for Control System Design,” Wiley,1998.

[39] Heffley, R. K. and Jewell, N. F., “Aircraft Han-dling Qualities Data,” NASA CR 2144, Dec.1972.

[40] Cook, B. H., “Flight Dynamics Principles,” But-terworth Heinemann, Cranfield, 1997.

21