Multidisciplinary aircraft design optimisation using an improved blackboard framework

Abstract

The commercial aircraft design process is controlled by chief engineers that meet at regular intervals to make key decisions. This has remained largely unchanged since the early days of aircraft design and has prompted researchers and industry practitioners to explore various communication architectures under the topic of Multidisciplinary Design Optimisation. Although many have been widely studied, they are rarely used in industrial design primarily because they fail to integrate well within the existing organisational structure of aircraft companies.
A legacy blackboard framework for Multidisciplinary Design Optimisation has been the subject of this study. Blackboard frameworks promote concurrent engineering practices using a database, some form of system level controller and a flexible arrangement of the knowledge sources that make up the design process. The specific framework considered in this thesis uses an automatic rule base to iteratively change the bounds on the shared design variables until they converge to what is deemed to be a single optimal design. The thesis covers the development and testing of a novel rule set, which has been given the name “Multidisciplinary Pattern Search” by the author, to reflect that its logic combines ideas from several well established heuristic optimisers.
Two aircraft design test cases demonstrate the merit of the Multidisciplinary Pattern Search, as well as the work done on the database and visualisation modules. The results indicate that the revised blackboard performs better than the distributed Collaborative Optimisation approach, albeit sometimes worse than the monolithic Simultaneous Analysis and Design method that tends to be very organisationally disruptive to implement. An additional 25% reduction in the convergence rate was achieved simply by reusing the available data in the database.
Finally, a team based application investigated the ease of use of the revised blackboard method. The feedback highlighted that the process was intuitive and largely easy to use, but further work is needed on a better human process interface.

More information

Identifiers

ORCID for Andy Keane: orcid.org/0000-0001-7993-1569

Catalogue record

Export record