Hans-Dieter Joos

Design of Robust Dynamic Inversion Control Laws using Multi-Objective Optimization

The design of robust attitude control laws for a civil aircraft with nonlinear dynamic inversion and multiobjective optimization is discussed. Dynamic inversion is a methodology that achieves linearization and decoupled command responses of the closedloop system via inverse model equations in the feedback loop. Using a linear outer loop controller the desired dynamic behavior is imposed. For tuning the free controller parameters, multi-objective optimization is used. The required robustness is achieved via a multi-model approach, as well as local robustness measures (e.g. gain and phase margins) as optimization criteria. As a new approach, not only the linear controller gains are optimized, but also physical parameters in the inverse model that are considered uncertain in the design model. The resulting control laws are used as inner loops of an autoland system, which was assessed for JAR-AWO requirements and successfully ∞ight tested.

Design of Autoland Controller Functions with Multiobjective Optimization

The application of multiobjective optimization to the design of longitudinal automatic-landing control laws for a civil aircraft is discussed. The control laws consist of a stability and command augmentation, a speed/flight path tracking, a glide-slope guidance, and a flare function. Mulitobjective optimization is used to synthesize the free parameters in these controller functions. Performance criteria are thereby computed from linear as well as nonlinear analysis. Robustness to uncertain and varying parameters is addressed via linear robustness criteria and via statisticle criteria computed from online Monte Carlo analysis. For each controller function, an optimization problem setup is defined. Starting with the inner loops, the synthesis is sequentially expanded with each of these setups eventually leading to simultaneous optimization of all controller functions. In this way, dynamic interactions between controller components are accounted for, and inner loops can be compromised such that these can be used in combination with different outer loop functions. This reduces controller complexity while providing good overall control system performance.

Application of an Optimization-based Design Process for Robust Autoland Control Laws

The design of a robust autoland con- troller using an optimization based design process is described. A modular controller architecture is de- veloped flrst. Inner loops are based on Dynamic In- version, speed and longitudinal path tracking loops are based on the Total Energy Control System. Functions for lateral path tracking, ILS guidance, ∞are and runway alignment are based on classical PID structures. The free parameters in the con- troller functions are tuned using multi-objective op- timization. Performance criteria are directly derived from the imposed design requirements and com- puted from linear or nonlinear analysis (e.g. eigen- values, simulations). Robustness is addressed via a multi-model approach, via optimization criteria (e.g. gain/phase margins), and, as a new contribution, via statistical criteria computed from on-line Monte- Carlo analysis. The design process and selected ar- chitecture have been applied to a wide-body as well as a small passenger aircraft to demonstrate con- troller robustness and e-ciency of the process. In both cases, JAR-AWO speciflcations had to be met. ⁄ Research Engineer, PhD candidate y Research Engineer, PhD z Research Engineer, DLR-Braunschweig, Institute for ∞ight system dynamics Copyright c ∞2001 by the German Aerospace Center DLR (Deutsches Zentrum fur Luft- und Raumfahrt e.V.). Pub- lished by the American Institute of Aeronautics and Astro- nautics, Inc., with permission. The latter design was successfully ∞ight tested.

A fault diagnosis based reconfigurable longitudinal control system for managing loss of air data sensors for a civil aircraft

Abstract An integrated fault diagnosis based fault tolerant longitudinal control system architecture is proposed for civil aircraft which can accommodate partial or total losses of angle of attack and/or calibrated airspeed sensors. A triplex sensor redundancy is assumed for the normal operation of the aircraft using a gain scheduled longitudinal normal control law. The fault isolation functionality is provided by a bank of 6 fault detection filters, which individually monitor each of the 6 sensors using robust low order LPV residual generators. In the case of losses of up to 5 sensors, a fault estimation technique based on LPV estimators can be employed to reconstruct the missing sensor information necessary for gain scheduling. In the worst case of a total failure of all 6 sensors, a robust constant longitudinal control law is employed which ensures a basic longitudinal control performance. The proposed control architecture fulfils the basic requirements formulated in the Benchmark Problem in the RECONFIGURE project.

ANDECS: a computation environment for robot-dynamics design automation

Features of an available computation environment for robot-dynamics design are described. These are environments for object-oriented mechatronics/robot dynamics modeling, for executing dynamics simulation experiments, for multiobjective parameter and trajectory optimization, and for multivariate result visualization. The environments have a common look and feel and are coherently integrated using an engineering database system.<<ETX>>

A graphical user interface for flight control development

The FSA (first shot approach) Demonstrator of DLR and DASA/Airbus is a first prototype for an integrated multidisciplinary environment for flight control development. The FSA Demonstrator represents a state-of-the-art computational tool which facilitates the study of trade-off between competing specifications and performance metrics. It integrates an object-oriented modeling environment, a data and tool management system, general purpose system analysis, simulation and synthesis tools and an automatic multi-goal attainment optimization program. The FSA Demonstrator comes with a dedicated graphical user interface for problem setup and push-button program operation which allows easy setup and operation even by non-specialists. The main pay-offs of the FSA are a significant reduction in the design cycle and an improved performance of handling qualities.

Enhanced detection and isolation of angle of attack sensor faults

An enhanced detection and isolation method to monitor the state-of-practice triplex redundant angle of attack measurements on modern civil transport aircraft is presented. The developed fault detection and diagnosis architecture relies on advanced model and signal based techniques monitoring each of the three sensors individually. This allows the correct isolation of erroneous sensors also in case of multiple sensor faults. The gathered isolation information is used in an advanced sensor fusion scheme, allowing the propagation of an adequate angle of attack value to the flight control computer in case of failure. The fault detection and diagnosis system is validated using a high �delity benchmark model of a large commercial transport aircraft using different wind excitations together with challenging pilot and auto-pilot scenarios.

Application of Optimization-Based Worst Case Analysis to Control Law Assessment in Aerospace

The flight control law design assessment problem can be formulated as a robustness analysis problem, where a set of suitably defined assessment criteria must be checked to lie within certain limits for all admissible variations of vehicle parameters, external inputs and all flight conditions. Optimization based worst case analysis can be used to find those parameters/inputs/flight conditions for which the criteria are violated or poorly satisfied. The potential of this approach is its general applicability to any kind of simulation models and scenarios including complex non-linearity in control laws. But in order to confidently assert that no violation of assessment criteria exists, a global optimization problem has to be solved. Especially in case of many assessment criteria, global worst case search can lead to a huge computational effort. However, solving worst case problems as a multi-objective problem can help to reduce the number of computations since all or some of the assessment criteria can be considered simultaneously. Optimization-based approaches can also be used to detect parameter sensitivities on the assessment criteria and can help to find safe parameter regions.

Flight control law clearance using optimization-based worst-case search

Abstract The flight control law clearance problem can be formulated as a robustness analysis problem, where a set of suitably defined clearance criteria must be checked to lie within certain limits for all admissible variations of aircraft parameters, pilot inputs and all flight conditions. The idea of optimisation based worst case clearance is to use available and efficient optimisation methods to find those parameters/inputs/flight conditions for which the criteria are violated or poorly satisfied. The potentials of this approach are its general usability for both frequency-domain and time-domain analysis, and for both linear and non-linear models and control laws including protections. Moreover, there is no limitation on the number of parametric uncertainties that can be investigated and the method does not itself add conservatism to the clearance problem as many other alternatives do, mainly those based on approximations or simplifications. The most challenging problem with the optimisation based approach is not to find hidden weaknesses if they exist, but to confidently assert that no weakness exists for all parameters/inputs/flight conditions. To assert this a global optimum solution has to be found usually resulting in a huge amount of computational work. For highly nonlinear clearance problems regarding manoeuvrability in the peripheral flight envelope of a civil transportation aircraft the suitability of the optimisation based approach is demonstrated. Different global optimisation algorithms and strategies are investigated with respect to its reliability in finding worst cases, computational efficiency and its potential for distributed computation in order to reduce time to clearance.

Multiobjective design assessment and control law synthesis tuning for flight control development

Flight control law design is a multivariable control problem where various strict requirements from multiple disciplines have to be satisfied. This paper describes multiobjective control synthesis tuning and design assessment, with application to flight control design. The main feature of this methodology is that the various kinds of design objectives can be taken into account in their most natural form and design alternatives can be assessed most visibly with respect to given requirements. Multiobjective synthesis tuning by min-max parameter optimisation allows interactive compromising in the set of what can be best-possibly achieved with a chosen control law structure.