Objectives and content:
In the current development process for auto land systems of civil airliners, high non-recurring costs are involved, because a time-consuming trial-and-error approach is used to design the control laws. This trial-and-error approach should guarantee that the closed-loop system is robust against the large number of parameter variations and model uncertainties that influence the design. Especially when control laws have to be redesigned in a late program stage, this kind of approach becomes very expensive, because for each iteration extensive tests must be performed to validate the controlled system. These tests consist of large batches of stochastic (Monte Carlo) simulations to verify compliance with the design requirements, statistical analysis tests to assess the robustness of the design, and ground-based simulations and flight tests to validate the design under realistic circumstances. Each update also affects the work carried out in related disciplines, such as system loads and structural dynamics analysis, which can cause substantial additional costs. For this reason, it is desirable to achieve a mature design in an early phase of the development program, when iterations are less expensive. This requires a new design procedure, based on advanced robust control techniques, in which model uncertainties and parameter variations are taken into account explicitly. The main goal of the REAL project is to develop this procedure, which is a major scientific and industrial challenge, because until now advanced robust control techniques have hardly ever been applied to solve practical autoland control problems.
Between October 1994 and May 1997 a European project has been performed, in which thirteen modern and classical control techniques were applied successfully on two fairly realistic flight control problems. An industrial evaluation of the designs has pointed out that most of the modern techniques have serious potential to solve robustness issues in a more systematic and efficient way than is possible with classical design techniques. The REAL partnership, consisting of NLR (project coordination), A‚ rospatiale (evaluation), Daimler-Benz Aerospace Airbus (economical benefits study), DLR (flight tests), Delft University of Technology (benchmark problem definition) and ONERA-CERT (design), will base its work on the control problem definitions, standard nomenclature, software models and tools, designs, and evaluation results from the European project mentioned above. First, a realistic autoland benchmark problem will be defined. Subsequently, three teams from France, Germany and the Netherlands will each develop a design procedure to solve this benchmark problem. To focus the project on cost-effectiveness, an cost benefit study will be carried out to determine in which design steps the highest cost reduction can be expected. In order to test the design procedures, each team will make two designs, based on civil aircraft models provided by A‚ rospatialeand DLR respectively. The first set of designs will be evaluated with an industrial autopilot evaluation tool, while the second set will be flight tested. Finally, the REAL partnership will combine the best elements of each design procedure to agree upon a European robust and efficient autopilot design approach, which will substantially reduce the nonrecurring costs involved in the design and development of autoland control systems. BE97-4113
Funding SchemeCSC - Cost-sharing contracts
2600 GB Delft