Citation:
David Manuel Judt, Craig Lawson, Development of an automated aircraft subsystem architecture generation and analysis tool, Engineering Computations, Vol. 33 Iss: 5, pp.1327 - 1352
Abstract:
Purpose – The purpose of this paper is to present a new computational framework to address future
preliminary design needs for aircraft subsystems. The ability to investigate multiple candidate
technologies forming subsystem architectures is enabled with the provision of automated architecture
generation, analysis and optimization. Main focus lies with a demonstration of the frameworks
workings, as well as the optimizers performance with a typical form of application problem.
Design/methodology/approach – The core aspects involve a functional decomposition, coupled
with a synergistic mission performance analysis on the aircraft, architecture and component levels.
This may be followed by a complete enumeration of architectures, combined with a user defined
technology filtering and concept ranking procedure. In addition, a hybrid heuristic optimizer, based on
ant systems optimization and a genetic algorithm, is employed to produce optimal architectures in both
component composition and design parameters. The optimizer is tested on a generic architecture
design problem combined with modified Griewank and parabolic functions for the continuous space.
Findings – Insights from the generalized application problem show consistent rediscovery of the
optimal architectures with the optimizer, as compared to a full problem enumeration. In addition
multi-objective optimization reveals a Pareto front with differences in component composition as well
as continuous parameters.
Research limitations/implications – This paper demonstrates the frameworks application on a
generalized test problem only. Further publication will consider real engineering design problems.
Originality/value – The paper addresses the need for future conceptual design methods of complex
systems to consider a mixed concept space of both discrete and continuous nature via automated methods.