Publications

Multidisciplinary design optimization (MDO) plays a vital role in designing aircraft. Thus, the modularization of various disciplines is necessary to efficiently couple them in MDO or multidisciplinary design analysis (MDA). In this paper, we demonstrate how recent library extensions \citeBackhaus_2021 lead to combining both the HPC ecosystem FlowSimulator, developed by DLR and partners, and the MDA/MDO framework, OpenMDAO, for a gradient-based MDO scenario of a generic static aeroelastic problem for a transport aircraft. In the MDO problem, the software architecture enables the optimal integration of high-fidelity workflows with central data management for high performance computing (HPC). To this end, we demonstrate how the FlowSimulator Data Manager (FSDM) enables the associated HPC simulation components to exchange massively parallel data in memory. Here, the algorithms provided by OpenMDAO—linear and nonlinear solution methods, and forward and backward processing of sensitivity derivatives—operate on the infrastructure of the FlowSimulator HPC ecosystem. Based on the defined scenario, different solution strategies can be applied. In this work we define a nested aeroelastic coupling setup within a trim equilibrium constraint definition. In particular, we highlight how an efficient modularization of the individual disciplines supports an effective gradient assembly and thus makes them available to the individual solution hierarchies and strategies. With this approach, we show how it is possible to transform a baseline constrained optimization problem into an unconstrained optimization definition, while still ensuring the trim equilibrium through an implicit solution based on the Schur complement method.