Accelerating aircraft design using automated process generation: An experimental architecture for aircraft design software

Accelerating aircraft design using automated process generation: An experimental architecture for aircraft design software

Ramakers, M.A.Y.

Vos, R. (mentor)
Hoogreef, M. (mentor)

Aerospace Engineering

AWEP

FPP

2015-10-20

The aircraft industry has seen many evolutionary changes in the past decades. Since the Boeing 707 however, the general shape and configuration of transport aircraft have remained similar and so has the aircraft design process. Since the aviation community is about to face a revolutionary breakthrough featuring new aircraft concepts like blended-wing body aircraft and Prandtl planes it is time to adapt the design process as well. The aircraft design process may have been extended by computational novelties and has been largely computerized, but the fundamental design process is still similar to what is used for conventional aircraft. This research investigates the effect of the design process on the design outcome and strives to find a method of automatically generating an aircraft design process given a set of computational modules, initial values and design goals. A software architecture based on a strict separation of components and process modelling approach is proposed. A framework based on this architecture is developed in which design parameters and computational modules are modelled as nodes in a graph. A subset of the conceptual aircraft design process is simplified and implemented. Twelve Algorithms are proposed to perform the automated ordering of the computational modules. The ordering is based on the module run time, estimated impact of the module on the design or the current state of the aircraft design, among others. The design process used in the Initiator aircraft design software is used as a benchmark. Two test cases are formulated, representing Class I and Class II design problems. Two additional Class II test cases are formulated to rule out the impact of the scheduling overhead by simulating the module runtime of fully-featured modules. Each test case is then solved by the proposed algorithms. The result is a system capable of generating a feasible (but sub-optimal) design process out of a set of stand-alone computational modules. This is achieved with no prior knowledge on the design process. From the test cases it can be concluded that the design outcome is not affected by the design process order that is employed, if the used set of modules is not changed. It is also concluded that the classical, fixed design process outperforms the design processes generated by any of the algorithms for Class I and Class II design problems by 20-40%. From the designed algorithms, the algorithm that orders modules based on the expected change of re-evaluating the module versus the expected run time of the module, performs best for the complex (Class II) design case. The impact of the scheduling overhead is shown to be negligible. Since the system is capable of generating feasible, but sub-optimal design processes, it is recommended that it be used as a tool to generate new design processes for unconventional aircraft configurations. Finally it is recommended that the process modelling based and separation of components based software structure that is presented is adopted for a next version of the Initiator.

aircraft design
process modelling
automated
software

http://resolver.tudelft.nl/uuid:1cdca7a1-a440-4dce-9fce-859fd9701839

Student theses

master thesis

(c) 2015 Ramakers, M.A.Y.