Benchmarking Distributed Optimization Formulations With a Multidisciplinary Feasible Approach For Aero-Structural Aircraft Design

Kurzfassung

In this paper, an aero-structural aircraft wing optimization capability is presented that overcomes the limitation of academic optimization methods of requiring the interaction of all disciplines including full gradients and at the same time increases the achievable improvement compared to traditional design methods. State-of-the-art distributed optimization methods including the bilevel optimization proposed by the Institut de recherche technologique Saint-Exupery (IRT) and a newly introduced distributed optimization formulation based on a Block Coordinate Descent, applying a manual decomposition of the problem, are compared to a Multidisciplinary Feasible architecture. Compared to other state-of-the-art distributed optimization methods including the bilevel formulation, our approach demonstrates a comparable reduction in the global objective function, while the computational time is reduced by about 58%. The more conventional embedded formulation yielded final designs of much lower aircraft performance, demonstrating the superiority of the applied distributed optimization methods to enable more efficient and robust strategy for aircraft wing design, while ensuring a separability of the disciplines as favored by existing process requirements in industry. The test case used for demonstration is a conceptual short-to-medium range airliner of industrial complexity, the DLR-F25 configuration, exhibiting a high-aspect ratio, composite wing.