You are on page 1of 15

AIAA SciTech Forum 10.2514/6.

2018-0386
8–12 January 2018, Kissimmee, Florida
2018 AIAA Guidance, Navigation, and Control Conference

A Rapid-prototyping process for Flight Control Algorithms for


Use in over-all Aircraft Design

Richard O. Kuchar ∗ and Gertjan H. N. Looye †


German Aerospace Center (DLR), Institute of System Dynamics and Control,
82234 Weßling, Germany

This paper describes the development and current status of a fully automated aircraft model
integration and preliminary control law design process for use in aircraft over-all design. This
process allows addressing flight control algorithms as a prime design variable from the earliest
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

design stages, and facilitates multi-disciplinary aircraft over-all dynamic design analyses in
for example Multi-disciplinary Design Optimization. The approach taken in this work is to
address the flight control algorithms at a specification level (e.g. dynamic response parameters)
and newly synthesize according algorithms adapted to the current aircraft design configuration
by means of a rapid-prototyping process. This in turn enables the use of dynamic analyses
in all disciplines involved (flight loads, systems, performance) from the earliest design stages,
increasing accuracy, reducing conservativeness in multidisciplinary optimization, and reducing
risk by increasing compatibility with dynamic analyses performed in detailed design stages. In
order to ensure the latter, the proposed control law prototyping process has been developed to
naturally evolve into detailed design of the flight control system.

Nomenclature

3DOF = 3-Degree of Freedom model (point-mass)


6DOF = 6-Degree of Freedom model (rigid body)
CCV = Control Configured Vehicle
CP ACS = Common Parametric Aircraft Configuration Schema
DLR = German Aerospace Center
EOM = Equations of Motion
f lightSim = Integrated flight dynamics tool (open- and closed loop), as implemented by DLR-SR
I N DI = Incremental Nonlinear Dynamic Inversion control method
M DO = Multi-Disciplinary Design Optimization
MOPS = Multi-objective Parameter Synthesis tool, as implemented by DLR-SR
N DI = Nonlinear Dynamic Inversion control method
RCE = Remote Component Environment

I. Introduction
Recent developments in aircraft design, such as new (hybrid-electric) propulsion systems, multi-functional control
surfaces, more electrical systems architectures, novel construction materials, and the increasing role of active control
functions pose considerable challenges on methods, tools and processes applied during the aircraft conceptual and
preliminary stages. These new developments widely open up the available design space. At the same time, even for more
conventional designs, further improving performance and reducing environmental impact, challenges assumptions behind
current design approaches. These assumptions are valid for conventional and future aircraft geometric configurations,
but may be either too conservative or overly optimistic for other technologies listed above.
Design analyses during conceptual and preliminary design, or as used in Multi-disciplinary Design Optimization
(MDO) processes, and detailed design tend to be quite different. In early design phases, analyses are mostly based on
statistical regression methods, as described in Roskam and Torenbeek - see [1], [2], and static analyses (e.g. in flight
∗ Research Engineer, Dept. of Aircraft Systems Dynamics, richard.kuchar@dlr.de
† Head of Aircraft Systems Dynamics, gertjan.looye@dlr.de

Copyright © 2018 by R.
Kuchar, G. Looye, German Aerospace Center (DLR). Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
loads and performance). During the detailed design and certification phases, focus shifts towards dynamic analyses.
The ability to perform early dynamic analyses in relevant disciplines could considerably help to improve compatibility
between preliminary and detailed design analyses, reduce reliance on the aforementioned assumptions, and therefore
reduce over-all risk.
Dynamic analyses require representative analysis models of both, the aircraft configuration, as well as its control
algorithms. This necessarily positions control algorithms at the level of a primary design discipline. This was depicted
by Anderson [3] in figure 1(b) and 1(a). Anderson proposed a method to assess risk of the control system for a so-called
"Control-configured Vehicle" (CCV), based on aircraft characteristics and expected control law complexity.
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

(a) Classical control synthesis (b) Control synthesis embedded in the aircraft over-all
design loop

Figure 1 Control synthesis within the aircraft design loop (from ref.[3])

The objective of the work presented in this paper is to actually enable closed-loop dynamic analyses in various
disciplines (flight dynamics / stability and control analysis, flight loads, systems, flight performance) within aircraft
conceptual and preliminary design stages. To this end, a fully automatic modeling and preliminary control law design
process has been set-up that is able to provide closed-loop analysis models for each given design iteration. The modeling
involves pre-processing of available and/or computation of missing data using appropriate methods, and integration in
flight dynamics models as appropriate for follow-on analyses (e.g. 3DOF for mission simulation, linear 6 DOF models
in operating points of interest for handling qualities assessment).
Control algorithms are designed using a so-called rapid-prototyping process, based on given design specifications
and automatic adaptation to the current design configuration. These control specifications, rather than control law
parameters are effectively design parameters at an aircraft level. In this way for example, demanded agility may be traded
off against actuation and control surface parameters. For automatic control law design, inversion-based techniques such
as Nonlinear Dynamic Inversion [4] (NDI) or Incremental Nonlinear Dynamic Inversion [5] (INDI) are used. These
methods have dedicated design parameters, that specify closed loop dynamic response behavior (reference models),
adapt to the specific aircraft configuration (inversion, allocation), or enable disturbance rejection (linear feedback
control). This not only allows for the required design automation, but also for natural propagation into detailed design
stages. Especially INDI results in control laws that are of comparable complexity to classical ones and have been
demonstrated in various flight tests [6, 7].
This paper describes the current status of the above processes and the underlying methodologies in more detail.
Besides the software implementation, also an example for an unmanned military aircraft will be discussed.

II. Automated flight dynamics modeling in over-all aircraft design


Design analyses in aircraft conceptual and preliminary design are required to be fast and, in case of MDO, automatic.
For this reason, the modeling and design process as described in this work is fully automated, from model data processing
and model generation to analyses and/or provision of analysis models to follow-on disciplinary tools as described in
figure 2.
The top half of the figure depicts the over-all aircraft optimization, based on customer and airworthiness requirements.

2
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

Figure 2 Aircraft design analysis process

In the middle, the actual disciplinary design analyses take place, based on design degrees of freedom set by and providing
analysis results to the optimizer. The process described in this paper (below) is part of a large number of disciplinary
design analyses.
In the early stages of the overall aircraft design process, the available data sources are usually sparse. Taking this into
consideration, the German Aerospace Center (DLR) started developing an inter-institutional, distributed computational
framework in order to connect available disciplinary tools on various fidelity levels and create a unified database for
further design investigation and MDO purposes. The various DLR-institutes provide respective disciplinary methods -
ranging from handbook methods for initial configuration setups, to mid- and high-fidelity methods (e.g. from lifting-line
and panel codes to CFD aerodynamic codes). In order to connect tools and collect relevant results, a dedicated xml-based
database, the "Common Parametric Aircraft Configuration Schema" (CPACS, see [8]) and an automated tool execution
framework, the "Remote Component Environment" (RCE, see [9]) has been implemented. All embedded tools require
specific inputs and produce respective output - which enforces a sequential order of computation, which differs between
respective application cases. The actual utilization of this framework and tool-sequence dependencies is extensively
discussed in [10].
The process for automatic model generation is fully integrated in the DLR MDO environment. Known as
"flightSim", the process can be integrated in a tool chain, providing a CPACS database (with content to be descri-
bed subsequently) as input and delivering analysis results or models as specified in the form of tool-parameters as outputs.

The automated modeling and control law design process (flightSim) consists of the following steps:
1) Processing of respective disciplinary model data available for the current design iteration (geometry information,
preliminary aerodynamic data, mass distribution, propulsion data, etc.; all on a feasible prediction level);
2) Generation of missing model data, based on selected computational methods or best-guess estimates. For
example, aerodynamic stability derivatives may need to be computed using a panel code;
3) Integration of runtime models, tailored towards intended analyses. For example, a 3DOF and 6DOF nonlinear
model for mission and flight dynamics analyses respectively, linear models for handling quality analyses, etc.
4) Automatic design of control algorithms
5) Provision of analysis or dedicated runtime models for use in follow-on tools.
In the following sections, model data processing and dynamic model integration will be discussed in more detail.

III. Data processing


The following preliminary parameters and databases are required to be available for the generation of a usable flight
dynamics model:
• Geometry parameters, description of the plan-form, reference length and control surface locations
• Weight and Balance, defining mass and inertia properties
• Aerodynamics, defining the actual aerodynamic derivatives - separated for a baseline configuration and the
respective control surfaces

3
• Control surface definition and allocation concept, defining the actual task-based use of aerodynamic and propulsion
based control surfaces
• Propulsion database, defining thrust, fuel flow and combustion related exhaust gas composition
• Control synthesis parameters, specifying the respective aircraft dynamic behavior
The respective data are assumed to be available within the aforementioned CPACS data structure.

A. Weight and Balance data


Besides geometrical parameters, mass properties are a key element in any dynamic aircraft model. For performance
analyses based on 3 DOF models, Operational Empty Weight (OEW) together with combinations of interest for payload
and initial fuel weight, affecting total weight and center of gravity location are sufficient. For 6DOF models, also the
inertia matrix is relevant. To this end, the DLR process contains a dedicated weight and balance module.
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

B. Aerodynamic model generation


Aerodynamic models of differing fidelity levels and respective uncertainty models can be integrated into the overall
context: Utilizing the nominal DLR work-flow would result in extensive aerodynamic performance maps of the rigid
body aircraft representation, generated either by the "Lifting Line" (vortex lattice method, see [11]) or "VSAero" tools
(3D panel method, see [12]) - depending on the actual scope of the analysis and computational budget.
In addition to the nominal aerodynamic codes, the alternative 3D panel code NEWPAN [13] is used for design
analysis, see for example figure 3. NEWPAN is an advanced 3D panel method for the computation of external
aerodynamic problems on a medium fidelity level. The utilization of NEWPAN is motivated by the uncomplicated
surface definition by direct selection of respective mesh areas, which allows alterations of the underlying control surface
setup. The work-flow in NEWPAN allows for direct computation of respective aerodynamic control surface derivatives.
Therefore only the sections of interest have to be recomputed for elaborated data generation. This allows for a mixed
fidelity approach: The baseline aerodynamic model of the aircraft under-design is computed by high-fidelity CFD
methods, whereas the aerodynamic properties of control surfaces and other lifting elements can be computed by lower
fidelity, less computational costly methods.
In accordance to the growing complexity of the underlying geometry, the effort for feasible aerodynamic mesh
generation is rising in parallel to the fidelity level of the employed computational method. With respect to NEWPAN
it is required to implement a structured 3D panel surface mesh of the aircraft and all lifting surfaces, together with
an initial wake geometry (see figure 3(a)). This process can currently be performed in a semi-automatic mode, but is
planned for a fully-automated process in the future.

(a) No controls set (b) Ailerons deflected

Figure 3 NEWPAN panel code example - DLR D150 configuration

The aerodynamic performance map defines coefficients and derivatives (totals, with respect to rigid body structure).
Forces are defined in dependency of [CX , CY , CZ ]T = f (α, β, Re, Ma), moments are formulated with [CL, CM , CN ]T =
f (α, β, Re, Ma) and the associated damping derivatives are defined by the derivative of the respective forces and
moments to the applied body angular rates [p∗, q∗, r ∗ ]T . The actual control surface derivatives are defined in a similar
way as functions of the control deflection ηi .
Corresponding to the derivation of aerodynamic datasets, control surface hinge moments can be provided - see [14].
This allows to evaluate sizing actuator loads associated with specific design maneuvers over the aircraft flight envelope.

4
C. Propulsion data
The CPACS data structure foresees the integration of multiple engines at specified locations in the structure. These
engines may be of different types.
For each engine type, an engine performance map must be included. This map can be derived by the nominal
surrogate database provided by TWdat (thermodynamic cycle tool, see [15]), resembling static thrust levels, fuel
consumption and exhaust gas composition over the entire flight envelope, based on a general thrust T = f (N1)
relationship. This allows the estimation of the overall fuel consumption, required thrust levels and the definition of a
general cost function for overall MDO processes.

D. Actuation data
Currently, data for actuators is limited to time constant or frequency/damping numbers for first and second order
dynamics approximations respectively. Also rate and position limits can be provided and actuators may be rate or
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

position controlled. Currently, data structures for more detailed actuator models are under development.

E. Control allocation schemes


A very important input, especially for closed-loop flight dynamics models, are control allocation schemes. This
input specifies how the aircraft-specific controls are to be employed in order to realize higher level commands, such
as roll, pitch and yaw inputs, or even steering and braking inputs on the ground. To this end CPACS allows to define
so-called "Command Cases" and "Control Distributors". For example, a roll command can be realized by associating
possible low- and/or high-speed ailerons, or maybe flaperons in the wing structure and deflecting these in an opposite
manner. Steering and braking commands can be associated with available landing gears, etc.

IV. Flight Dynamics Model integration


Basis for the integration of model data are the nonlinear 6DOF Newton-Euler equations of motion. For performance
and mission analyses, it is mostly sufficient to consider the aircraft as a point-mass.

A. Equations of Motion
The Newton-Euler equations of motion, as described in [16] - are formulated for rigid aircraft (6DOF) in body frame.
The respective translational accelerations can be determined, using the principle of linear momentum:
 uÛ K  
  X A XT    − sin Θ   qK wK − rK vK 
  
1
         
=  vÛ K  = Tbs (α) Y A  + Y  + g  sin Φ cos Θ  − rK uK − pK wK 

  T 
Û Kb
V
       
(1)
  m  A  T     
wÛ K  
 Z  Z   cos Φ cos Θ  pK vK − qK uK 
 b   s  b    b
with the translational speeds V = [uK , vK , wK ]T , rotational speeds Ω = [pK , qK , rK ]T , Euler angles Φ, Θ, Ψ.
The transformation matrix (stability-frame→body-frame) Tbs , aerodynamic and propulsion forces can be generally
determined by:
cos α 0 − sin α  X A −CD  XT  T 
        
Tbs (α) =  0 0 , Y  = q̄S  CY  , Y  =  0 
 A  T
1
     
 sin α 0 cos α 
   A    T  
 −CL 
  b 0
Z  Z   
  s  
Dynamic pressure is defined by q̄ = ρ1 Va2 with air-density ρ and wing reference area S, aerodynamic coefficients
CD, CY , CL = f α, β, α,
Û β,
Û p, q, r, ηi, Re, Ma and thrust T.


The respective angular accelerations can be derived by the following expressions:


 pÛK  
  L A + LT   qK rK (Iz − Iy ) − pK qK Ixz  
  
     

Û K b =  qÛK  = I
  

A + MT  −
Ω −1
M rK pK (Ix − Iz ) + (p2K − rK2 )Ixz  (2)
  
b   
 N + N b
 pK qK (Iy − Ix ) + qK rK Ixz  
  A T
   
 rÛK  
   
 b  b
with the aerodynamic and thrust induced torque:

5
 LA   Cl   LT   0 
       
 M  = q̄Slre f Cm  ,  M  = lT z T 
 A  T
(3)
   
 A    T  
N   Cn  N 
 b lT y T 
 
 b   

b written for the symmetric aircraft I xy = Iyz = 0 :


and the inverted inertia tensor I−1


−1
 Ix 0 −Izx   I y Iz 0 Iy Izx 
 1 
b =  0
I−1 0  = 2
Ix Iz − Izx (4)
Iy  0 0 
   
 2
Iy (Ix Iz − Izx ) 
 
−Izx 0 I z b I I 0 I I
 y zx x y b
 

The 3DOF point-mass equations of motion neglect attitude dynamics. Translational degrees of freedom are then
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

more conveniently expressed in the form of flight path related states Vf , γ f , χ f , which are the inertial velocity, flight
path elevation and course angle respectively. These equations are derived in [16] and used for 3DOF simulations in [17].

B. Model integration
For actual model integration and for generation of runtime models, the model generation process uses Modelica,
as well as a dedicated Flight Dynamics Library [18, 19]. Modelica [20] itself is an object-oriented, equation based,
multi-domain physical modeling language for the modeling of complex physical systems. The advantage of using
Modelica in combination with the Dymola environment [21], is the possibility to generate various types of runtime
models from a single model implementation. For example, the generation of 3DOF or 6DOF simulation models is the
matter of a single switch. Also inverse models for Dynamic Inversion Control (to be discussed subsequently) can be
provided automatically.
The general implementation structure is depicted in figure 4. The central CPACS components include data as
described in the previous section. Aerodynamic and propulsion tables, as well as control allocation schemes are
included by means of external ANSI C-files that are automatically generated during data pre-processing. This allows
for the use for different application rules that may come with the model data. For example, data from Lifting Line or
Panel-based methods need to be combined in different ways in order to result in correctly computed aerodynamic forces
and moments.
The landing gears, actuation and engines block may include an arbitrary number of gears, actuators and engines
respectively. Their characteristics and number are obtained from the database during model generation.

Figure 4 DLR Modelica Flight Dynamics Library

6
V. Inverse Model-based Nonlinear Control
The most important novelty of the presented modeling process is the automatic design of control laws. During
flight operations, a large number of functions is used, varying from manual attitude control to autopilot modes, or
functions for controlling the aircraft on ground. In order to facilitate a wide spectrum of design analyses, the control
architecture includes most of these functions. For example, handling quality evaluation, manual control laws for attitude
as well as load factor command variables are available. For mission analyses, autopilot modes may be commanded
from a dedicated mission module. For flight loads analyses, via the CPACS input file, it is possible to specify dynamic
characteristics of these functions individually, see figure 5(a), using attitude control functions as an ideal pilot model for
simulation of design maneuvers. These characteristics may be imposed by means of automatic selection of control law
parameters, either analytically or by means of multi-objective optimization (see [22]).
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

(a) Aircraft dynamic criteria (MOPS) for control law synthesis in (b) Step response for multiple optimization runs
CPACS

Figure 5 Aircraft dynamic behavior

For easy adaptation of the control laws to continuously changing aircraft design configurations, inverse model-based
control techniques (e.g. Nonlinear Dynamic Inversion (NDI), as well as Incremental Nonlinear Dynamic Inversion
(INDI)) offer a great advantage, as they provide four separate aspects of control law design within four more or less
independent design degrees of freedom, compare with figure 6:
1) Control allocation: It is common practice to define generalized control commands for each individual degree of
freedom to be controlled. Based on controllability analyses, these commands comprise scaled deflections of
physical control effectors (for example: asymmetric deflection of left and right aileron for aircraft to achieve roll
control).
2) Decoupling and compensation for nonlinearities: Algebraic model equations relating differentiated command
variables (so-called “pseudo-controls”) and generalized commands are inverted, resulting in normalized and
decoupled pseudo controls responses. This part thus handles any vehicle-specific dynamics.
3) Command shaping: Desired behavior to command variable inputs is achieved by means of reference models.
These in turn generate references for the pseudo controls in the inverse model equations and the feedback
controller.
4) Disturbance rejection and robustness; to this end, a feedback controller is used.
The four degrees of freedom may in turn be designed with the help of appropriate methods and tools. E.g. for
control allocation, static and dynamic optimization schemes may be used. For feedback control, any classical or modern
synthesis approach can be applied. The inverse model equations can be automatically generated from the model
implementation in Modelica, see [19].
The underlying concept of Incremental Nonlinear Dynamic Inversion (INDI, see [5] and figure 7) has been extensively
evaluated for its application within the described Rapid-Prototyping process. Additionally, the same algorithms have
also been successfully flight tested with UAV (see van Ekeren [6]) and CS-25 (see Grondman [7]) class aircraft.

7
Figure 6 Common Structure of inversion-based control algorithms. Numbers indicate the design degrees of
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

freedom

Figure 7 INDI control law structure

For autoflight control laws, the Total Energy and Total Heading Control Systems (TECS [23], THCS [24]) are used.
As the derivation of these control law architectures is based on Point-Mass dynamics of the aircraft, they are easily
adapted to a new aircraft design configuration.

A. Integrated Reference Model


Utilizing general control law design rules, the overall control law architecture follows two basic principles: The
independent treatment of a predefined behavior and tracking functionality versus the suppression of external disturbances.
Therefore, the availability of a reference model ensures that controlled aircraft follow precisely the prescribed
reference characteristics within their physical limits (see figure 8). This feature can be used for a detailed investigation
of the overall dynamic properties of different aircraft designs, as similar system reactions are enforced. This is especially
useful, when comparing an aircraft design against operational requirements, certification rules and provides comparable
insight for succeeding disciplines.

Figure 8 Control law architecture with embedded reference model

Especially, as it allows comparable analysis with focus on different lines of analysis: E.g. commanded system
reaction vs. actuator activity vs. energy consumption, structural loads, and many more.

8
VI. flightSim - an Integrated Tool for Flight Dynamics and Control
The above described functionalities have been implemented over the recent years and are currently integrated and
available within the DLR-SR tool flightSim. This tool-set is dedicated to provide open- and closed-loop, as well
as mission simulation functionalities to other disciplines through the RCE-framework interfaces and the underlying
CPACS database. flightSim is implemented, using Modelica for the actual model implementation, Matlab (for parsing,
process-flow and trim routines), Simulink (for the generation of simulation models, control laws, code-generation),
ANSI-C databases (integrated into the model context - aerodynamics, propulsion and control functionalities) and
respective utilities (TIXI xml-parser). In addition, flightSim can be coupled with MOPS, the DLR "Multi-Objective
Parameter Synthesis" tool [22] for the tuning of outer-loop control modules. The respective optimization criteria define
the dynamic behavior of the aircraft in its entirety - and therefore fulfill the initial requirement for the active integration
of flight control law synthesis in the aircraft design loop, as proposed in figure 1(b).
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

In detail, flightSim covers the following functionalities - based on a valid and complete database for model generation in
CPACS:
• Open-loop flight dynamics model (3DOF and 6DOF)
• Closed-loop models based on INDI-control law functionalities (3DOF and 6DOF) - featuring:
– attitude-based control laws (RCAH, "rate command attitude hold")
– C*/nz-based control laws
– alpha-command based control laws
– TECS/THCS-based control - for vertical and lateral control / flightpath-control
– Autopilot functionality: Automatic Take-off, Cruise phase, Landing and functions for automatic and manual
on-ground control
• Mission simulation based on respective mission definitions in CPACS - utilizing the generated 3DOF model
coupled with the synthesized control law in order to perform a full mission
• Static trim analysis for various defined flight states (3DOF and 6DOF, open-loop)
• Generation of linearized models (6DOF, open-loop) - based on previous trim analysis
The implemented architecture allows for the flexible extension of flightSim with respect to new control- and
analysis-requirements. Taking the current state and the given flexibility into account, flightSim can be used for a large
spectrum of analysis tasks. The following section shall give an overview of potential use-cases, that can be implemented
on-top of it.

A. Applications of the Rapid Prototyping process in aircraft preliminary design


The availability of dynamic, open- and closed-loop aircraft models is a key fact for the investigation of the overall
dynamic behavior of a new aircraft model - and especially valuable in the very early stages of the design phase. Primary
design driver of each novel aircraft concept are airworthiness and performance requirements as defined by FAA FAR-25
/ EASA CS-25 [25] - Subpart B, or equivalent, as well as general performance-, dynamics- and/or endurance specific
design requirements.

The following items define a preliminary, non-exhaustive list, intended to show basic possibilities, utilizing a dynamic
modeling and simulation environment:
• Open-Loop analysis:
– Trim analysis of the respective aircraft model - deriving the achievable stationary flight envelope (e.g. CS
25.161)
– Preliminary determination of design speeds (e.g. Vst all, Vmc ) - as defined in CS 25.335
– Static and dynamic stability, linearized model analysis (e.g. CS 25.171 et seq.)
– Handling quality analysis, as utilized in HAREM [26]
• Closed-Loop analysis:
– Flight performance analysis for dedicated maneuvers: Take-off and landing run analysis, climb and decent
evaluation, take-off speeds for various load configurations, go-around evaluation (e.g. CS 25.105 et seq.)
– Comparison between diverse aircraft concepts with respect to dynamic behavior - utilizing the reference
model benchmark capabilities
– Evaluation of EFCS implications within the system (e.g. CS 25.143)

9
– Directional control of the aircraft in critical flight conditions (e.g. CS 25.147)
– Evaluation of required control surface derivatives with respect to specific design maneuvers (e.g. roll
maneuver, bank-to-bank, as for example defined in CS25 AMC 25.1420)
– Analysis of weather effects (wind, gust, turbulence) and load evaluation (e.g. CS 25.331)
– Flight characteristics of the partially degraded aircraft (e.g. One-Engine-Out, actuation system)
– Maneuver loads analysis of the controlled aircraft with respect to structural load analysis and sizing
– Deriving systems specifications - based on extensive maneuver analysis of the entire flight envelope (e.g.
for control surface actuator sizing and evaluation)
– Ground handling characteristics (CS25.231 et seq.)
The availability of respective analysis methods for the evaluation of the items stated above strongly support the
design process and decision making throughout the aircraft preliminary design phase and can be extended well into the
detailed design process and towards "virtual testing" - and in particular for "virtual certification".
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

VII. Example aircraft analysis


For demonstration purposes, the available DLR-F19 configuration (unmanned military configuration, see figure 9)
has been selected, as it is well defined (see [27]) and can therefore serve as reference case for the development and
testing of flightSim and associated analysis methods. All general parameters, geometry, base-line aerodynamics and
propulsion have been defined within a respective CPACS database.

Figure 9 DLR-F19 aircraft model - external shape

The following analysis / demonstrations are going to be performed:


• Mission simulation of the selected aircraft
• Combined control law demonstration
• Analysis of three control surface configurations with respect to roll behavior

A. Mission simulation example


A pre-defined mission can be simulated using the respective point-mass (3DOF) flight dynamics model in association
with the generated flight control laws - as generated in flightSim. The respective tracking functions for each flight
segment are generated by a dedicated algorithm. See figures 10 and 11 for an impression of the flown mission:

10
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

Figure 10 Mission simulation example (DLR-F19) - trajectory (Base image by Google Earth)

Altitude 104 Thrust


15000 6
Altitude / [m]

Thrust / [N]

10000 4

5000 2

0 0
0 100 5 103 1 104 0 100 5 103 1 104

Mach Fuel Mass


1 6000

0.8
Fuel Mass / [kg]

4000
Ma / [1]

0.6

0.4
2000
0.2

0 0
0 100 5 103 1 104 0 100 5 103 1 104

Lift coefficient Drag coefficient


0.6 0.03

0.025
0.4
CD / [1]
CL / [1]

0.02
0.2
0.015
0
0.01

-0.2 0.005
0 100 5 103 1 104 0 100 5 103 1 104
tmission / [s] tmission / [s]

Figure 11 Mission simulation example (DLR-F19) - results

11
B. Control law application example
The following simulation results, depicted in figure 12 show a sequence of commands, as usually issued for detailed
mission simulation or maneuver analysis. Being widely influenced by aerodynamic vortices, nonlinear behavior can be
assumed over the entire flight envelope. In order to show the performance in all three axis, dedicated maneuvers have
been selected, simulating a departure scenario:
1) Climb to a defined altitude - controlled by a specified flight path angle command of γ = 10°
2) Course change of ∆X = 90°, followed by
3) Increased speed of ∆VIAS = 50 m/s

Airspeed Altitude Flight Path Angle


7000 15
200
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

10
V IAS / [mps]

HWGS / [m]
6000

/ [deg]
180
5

5000
160 0

140 4000 -5
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250

Roll Pitch Yaw


40 20 0

30 -20
15
/ [deg]

/ [deg]

/ [deg]
20 -40
10
10 -60
5
0 -80

-10 0 -100
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250

Elevator deflection Aileron deflection 104 Thrust


0 10 4
Aileron left
-1 Aileron right
5 3
/ [deg]

/ [deg]

-2
T / [N]

0 2
-3
e

-5 1
-4

-5 -10 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
tman / [s] tman / [s] tman / [s]

Figure 12 Mission simulation example (DLR-F19) - departure simulation (solid red: commanded values, blue:
aircraft response, dashed red: mode changes)

It can be deduced, that the aircraft in the defined configuration smoothly follows the issued commands. The coupling
between all axis and the respective thrust setting is clearly visible.

C. Configuration comparison - Roll analysis


One of the core use-cases for flightSim in aircraft preliminary design is the evaluation of control surface definitions.
Therefore in this example, the effectiveness of all controls is altered: 50 % of the nominal case, the nominal case itself
and 150 % of the nominal case. One of the criteria mentioned in CS25 AMC 25.1420 to be analyzed is a coordinated
bank-to-bank roll. This maneuver has therefore been selected for analysis, as depicted in figure 13. Furthermore, this
example shall depict the use of the embedded reference model for the roll axis - as it guarantees a valid benchmark for
the analysis of differing configurations.

12
Roll comparisson Pitch comparisson Yaw comparisson
50 4.5 −60
Φref cmd Θmeas: 50% ηeff
Φmeas: 50% ηeff Θmeas: 100% ηeff −80
Φmeas: 100% ηeff Θmeas: 150% ηeff
Φmeas: 150% ηeff 4 −100

Ψ / [deg]
Φ / [deg]

Θ / [deg]
0 −120

3.5 −140
Ψmeas: 50% ηeff

−160 Ψmeas: 100% ηeff


Ψmeas: 150% ηeff
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

−50 3 −180
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50

Elevator deflection Aileron deflection Virtual Rudder deflection


4 25 10
Elev: 50% η
eff
20
2 Elev: 100% η
eff
5
Elev: 150% ηeff 15
0
10
0
ηe / [deg]

ηa / [deg]

ηa / [deg]
−2 Ail − L: 50% ηeff
5
Ail − R: 50% η
−4 eff
−5
0 Ail − L: 100% η
eff
−6 Ail − R: 100% η Rud: 50% ηeff
−5 eff
Ail − L: 150% η −10 Rud: 100% η
−8 eff eff
−10
Ail − R: 150% ηeff Rud: 150% ηeff
−10 −15 −15
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
t / [s] t / [s] t / [s]
man man man

Figure 13 Roll simulation example (DLR-F19) - comparison of varying control efficiency (50 %, 100 % and
150 %) - red solid line: Roll reference model

The plots describe aircraft reaction of the three analyzed control configurations with respect to a precisely commanded
roll motion. As can be seen, the roll command is followed smoothly and in a very similar fashion - due to the exact
reference model. The respective reaction in the associated yaw axis can be seen respectively - as the aircraft tries to
maintain roll and pitch attitudes. Similar, the reaction of the control surfaces show differences in deflections, due to
the applied control surface efficiencies. Further analysis can show, if employed control surface, control allocation and
actuator combinations are capable of performing respective maneuvers, as described in airworthiness requirements.

VIII. Conclusion
The presented paper describes the theory and implemented methods of a fully automated aircraft model integration
and preliminary control law synthesis process for use in aircraft over-all design. Flight control algorithms are
implemented at a specification level and control laws are provided, that are adapted to the current aircraft design
configuration by means of a rapid-prototyping process. The implemented methods for model generation and automatic
control law synthesis have been outlined, together with potential application possibilities for controlled aircraft setups.
The dynamic results from such analysis can sub-sequentially be used by all other involved disciplines - e.g. flight loads,
maneuver performance analysis and systems design. Finally, the implemented control architecture supports a continuous
development process of the flight control system, ranging from preliminary to detailed design phases.
The respective work is ongoing and the authors are expecting further progress with respect to the analysis of
up-coming aircraft configurations, specific analysis cases and associated control-law requirements to be applied in
aircraft preliminary design loops.

13
References
[1] Roskam, J., “Airplane Design: Parts I-VIII,” DARcorporation, Lawrence, KS, 2006.

[2] Torenbeek, E., Synthesis of subsonic airplane design: an introduction to the preliminary design of subsonic general aviation and
transport aircraft, with emphasis on layout, aerodynamic design, propulsion and performance, Springer Science & Business
Media, 2013.

[3] Anderson, M., and Mason, W., “An MDO Approach to Control-Configured-Vehicle Design,” 6th AIAA/NASA/ISSMO Symposium
on Multidisciplinary Analysis and Optimization, Bellevue, 1996. doi:10.2514/6.1996-4058.

[4] ENNS, D., BUGAJSKI, D., HENDRICK, R., and STEIN, G., “Dynamic inversion: An evolving methodology for flight control
design,” Vol. 59, 1994, pp. 71–91.

[5] Sieberling, S., Chu, Q. P., and Mulder, J. A., “Robust Flight Control Using Incremental Nonlinear Dynamic Inversion and
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

Angular Acceleration Prediction,” JOURNAL OF GUIDANCE, CONTROL, AND DYNAMICS, Vol. Vol. 33, No. 6, 2010.

[6] van Ekeren, W., Looye, G., Kuchar, R., Chu, Q., and van Kampen, E.-J., “Design, Implementation and Flight-Tests of
Incremental Nonlinear Flight Control Methods,” submitted to AIAA Guidance, Navigation, and Control Conference, 2018.

[7] Grondman, F., Looye, G., Kuchar, R., Chu, Q., and van Kampen, E.-J., “Design and Flight Testing of Incremental Nonlinear
Dynamic Inversion-based Control Laws,” submitted to AIAA Guidance, Navigation, and Control Conference, 2018.

[8] Ciampa, P. D., Nagel, B., Meng, P., Zhang, M., and Rizzi, A., “Modeling for physics based aircraft predesign in a collaborative
environment,” 4th CEAS Air & Space Conference, 2013. URL http://elib.dlr.de/95184/.

[9] Seider, D., “Open Source Framework RCE: Integration, Automation, Collaboration,” 4th Symposium on Collaboration in Aircraft
Design, Toulouse, France, 2014. URL http://elib.dlr.de/93323/1/20141127_TCAD_Open_Source_Framework_RCE-
Seider.pdf.

[10] Liersch, C. M., and Hepperle, M., “A distributed toolbox for multidisciplinary preliminary aircraft design,” CEAS Aeronautical
Journal, Vol. 2, No. 1-4, 2011, pp. 57–68.

[11] Horstmann, K.-H., “Ein Mehrfach-traglinienverfahren und seine Verwendung für Entwurf und Nachrechnung nichtplanarer
Flügelanordnungen,” Ph.D. thesis, Deutsche Forschungs-und Versuchsanstalt für Luft-und Raumfahrt (DFVLR), 1987.

[12] Maskew, B., “Program VSAERO theory Document: a computer program for calculating nonlinear aerodynamic characteristics
of arbitrary configurations,” 1987.

[13] Flow Solutions Ltd., “NEWPAN,” , 2017. URL http://www.flowsol.co.uk/products/newpan/.

[14] Vechtel, D., Hauber, B., and Looye, G., “Analysis of a multi-functional high-lift system driven by an active differential gear
box,” CEAS Aeronautical Journal, 2014. URL http://elib.dlr.de/88474/.

[15] Deidewig, F., “Ermittlung der schadstoffemissionen im unter-und Überschallflug,” Ph.D. thesis, 1998.

[16] Brockhaus, R., Flugregelung, Springer-Verlag, 2013.

[17] Looye, G. H. N., “TECS/THCS-based generic autopilot control laws for aircraft mission simulation,” Second CEAS Specialist
Conference on Guidance, Navigation and Control, 2013.

[18] Klöckner, A., Looye, G., Müller, R., Kuchar, R., Re, F., and Leitner, M., “Object-Oriented Aircraft Modeling with the DLR
FlightDynamics Library,” AIRTEC 2014, 2014. URL http://elib.dlr.de/91597/.

[19] Looye, G. H. N., “An Integrated Approach to Aircraft Modelling and Flight Control Law Design,” Ph.D. thesis, 2007.

[20] Hilding, E., Otter, M., and Mattsson, S. E., “Modelica Language Specification Version 3.3,” Tech. rep., Modelica Association,
2012. URL https://www.modelica.org/documents/ModelicaSpec33.pdf.

[21] Dassault Systèmes, “Dymola 2016,” , 2015. URL https://www.3ds.com/products-services/catia/products/


dymola.

[22] Joos, H.-D., “MOPS-Multi-Objective Parameter Synthesis, User’s Guide V6. 6,” DLR, 2016.

[23] Lambregts, A. A., “TECS Generalized Airplane Control System Design – An Update,” Submitted to the 2nd CEAS GNC
conference, Delft, The Netherlands, 2013.

14
[24] Lambregts, A. A., “THCS Generalized Airplane Control System Design,” Submitted to the 2nd CEAS GNC conference, Delft,
The Netherlands, 2013.

[25] EASA, C. S., “Certification Specifications and Acceptable Means of Compliance for Large Aeroplanes CS-25,” Tech. rep.,
Amendment 20, European Aviation Safety Agency, 2017. URL https://www.easa.europa.eu/system/files/dfu/CS-
25%20Amendment%2020.pdf.

[26] Ehlers, J., “Flying Qualities Analysis of CPACS Based Aircraft Models-HAREM V2. 0,” 2013.

[27] Schütte, A., Huber, K. C., and Boelens, O. J., “Static and dynamic numerical simulations of a generic UCAV configuration with
and without control devices,” 32nd AIAA Applied Aerodynamics Conference, 2014, pp. 1–26.
Downloaded by UNIVERSITY OF ADELAIDE -INTERNET on January 17, 2018 | http://arc.aiaa.org | DOI: 10.2514/6.2018-0386

15

You might also like