A Distributed Multi-Disciplinary Optimisation of a Blended Wing Body UAV Using a Multi-Agent Task Environment
J.P.T.J. Berends, Michel J.L. van Tooren, D.N.V. Belo · 2006
The Multi-Disciplinary Design and Optimisation process of products can be supported by automation of analysis and optimisation steps. A Design and Engineering Engine (DEE) is a useful concept to structure this automation. Within the DEE, a product is parametrically defined using a Knowledge Based Engineering approach. The analysis of a particular instantiation of the parametric model is performed by discipline analysis tools. By defining design variables and fixing parameters, an optimisation on these design variables can be performed against an objective function. The focus of the benchmark in this article is on developing a multi-objective distributed optimisation capability for the DEE using an optimisation problem of a blended wing body (BWB) unmanned aerial vehicle (UAV) aircraft, as a study object. The multi-objective nature of the optimisation asks for a high number of evaluations of the objective and (in)equality constraint functions. The term ‘distributed’ indicate that the calculation of these objective and constraint functions of the optimisation does not necessarily takes place within a single computer and a single optimisation program, instead these calculations are physically distributed over multiple processes and computers. A prototype capable of supporting such distributed MDO analysis, using the concept of a DEE, is the Multi-Agent Task Environment. Using the unique data-pull features of the TeamMate agent system, the user denominated as Operator can select which analysis of the product model should take place and which not by selecting the various constraints and objective functions to be included in the optimisation. If the Operator includes a certain constraint function, all data needed to evaluate the constraint will be automatically generated. The Operator can, in this case, 'shape' the analysis process by deliberately selecting (or de-selecting) certain analysis to be carried out. Results from two separate optimisation cases are reported and discussed.