Composing MDAO symphonies: graph-based generation and manipulation of large multidisciplinary systems

摘要

This paper proposes a novel methodology and its software implementation, called KAD-MOS (Knowledge- and graph-based Agile Design for Multidisciplinary Optimization System), which aims at increasing the agility of aircraft design teams that perform collaborative multidisciplinary design analysis and optimization (MDAO) by means of graph manipulation techniques. By agility, the ease and flexibility to assemble, adapt and adjust MDAO computational systems is intended here, as necessary to better fit the iterative nature of the aircraft design process. KADMOS has been developed on the notion that a formal specification of an MDAO system is required before its actual implementation, especially to be able to compose large and complex systems in multidisciplinary design teams. This specification system is under development as part of the EU project AGILE where a new generation of aircraft MDAO systems is investigated to support collaboration of heterogeneous teams of experts. KADMOS improves the agility of the design team in three ways: 1) reducing the set-up time required to compose large and complex MDAO models, 2) enabling the systematic inspection and debugging of this model, and 3) manipulating the model for automated creation and reconfiguration of optimization strategies, including the accompanying executable workflow. This is achieved by means of a graph-based analysis system that combines different existing advantageous techniques for performing MDAO, such as the use of a single shared data schema containing a parametric representation of the aircraft, knowledge-based technologies, and simulation workflow (SWF) software packages. Two MDAO case studies will be presented in the paper. The first case study is based on a simple analytical problem, generally used in literature for MDAO benchmarking studies. The second case study concerns a detailed wing aerostructure design using a collection of wing design tools. While the simple and compact analytical problem is used in this paper to demonstrate the functionalities of the tool, the wing design case demonstrates the capability of KADMOS to support quick formulation, (re)configuration, and execution of MDAO workflows using distributed and heterogeneous sets of analysis tools.