A clustered and surrogate-based MDA use case for MDO scenarios in AGILE project
Lefebre, T.; Bartoli, N.; Dubreuil, S.; Panzeri, M.; Lombardi, R.; Lammen, W.F.; Mengmeng, Z.; Gent, I. van; Ciampa, P.D.
19th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Atlanta, Georgia2018
In this paper methodological investigations regarding an innovative Multidisciplinary Design and Optimization (MDO) approach for conceptual aircraft design are presented. These research activities are part of the ongoing EU-funded research project AGILE. The next generation of aircraft MDO processes is developed in AGILE, which targets significant reductions in aircraft development cost and time to market, leading to cheaper and greener aircraft solutions. The paper introduces the AGILE project structure and recalls the achievements of the first year of activities where a reference distributed MDO system has been formulated, deployed and applied to the design and optimization of a reference conventional aircraft configuration. Then, investigations conducted in the second year are presented, all aiming at making the complex optimization workflows easier to handle, characterized by a high degree of discipline interdependencies, multi-level processes and multi-partner collaborative engineering activities. The paper focuses on an innovative approach in which knowledge-based engineering and collaborative engineering techniques are used to handle a complex aircraft design workflow. Surrogate models replacing clusters of analysis disciplines have been developed and applied to make workflow execution more efficient. The paper details the different steps of the developed approach to set up and operate this test case, involving a team of aircraft design and surrogate modelling specialists, and taking advantage of the AGILE MDO framework. To validate the approach, different executable workflows were generated automatically and used to efficiently compare different MDO formulations. The use of surrogate models for clusters of design competences have been proved to be efficient approach not only to decrease the computational time but also to benchmark different MDO formulations on a complex optimization problem.