The department focuses on research and development of software frameworks and the integration of simulation components for multidisciplinary analyses and optimizations of aircraft. The software developments aim at consistent and highly-accurate numerical descriptions that can be embedded in variable-fidelity strategies. Software methods and algorithms are being developed and implemented to allow for modular integrations enabling scalable high-fidelity analyses on modern high-performance computers (HPC).
Among the main research activities are:
Software frameworks for deterministic multi-disciplinary optimization (MDO) are developed and evaluated to modularly integrate HPC analysis tools. The focus is on gradient-based optimization strategies to efficiently deal with a large number of optimization parameters and constraints. The user is assisted in the setup of complex optimization scenarios with emphasis on automatic processing of sensitivity derivatives for multidisciplinary problems in conjunction with different numerical simulation plug-ins.
By means of algorithmic differentiation (AD) techniques, consistent and efficient approaches to computing sensitivity derivatives are developed and provided in both forward and reverse mode, in particular for iterative and transient simulation plug-ins such as CFD or CSM methods. The ability to efficiently compute exact sensitivity derivatives is a key ingredient enabling large-scale gradient-based simulation and optimization.
Simulation templates for complex, unsteady fluid-structure coupled simulation scenarios are developed based on the HPC environment FlowSimulator and its simulation plug-ins. The software solutions are designed in a flexible and modular way. Robustness, parallel and algorithmic efficiency are key requirements. Simulation templates are being developed in cooperation with other DLR institutes and project partners and will be provided for different scenarios such as load analysis or aircraft flight maneuvering.
Central simulation modules required for large-scale product analysis and optimization are being developed, process-integrated and provided to other DLR Institutes and project partners. Among the simulation modules are finite element-based structure mechanics tools, massively parallel non-linear and linearized CFD methods, e.g. for the efficient analysis of time-periodic problems, and robust panel methods for the rapid prediction of hundreds of thousands of aerodynamic load cases, many of which are transient.
Parametric CAD solutions for simulation-based product analyses and design optimizations are to be developed and provided. A unified and consistent geometry approach is required in conjunction with models of multiple physics and levels of fidelity. CAD integration into simulation and optimization processes is a key issue. The developments aim at a systematic simulation-link-to-CAD and CAD-in-the-loop capabilities. Gradient-based shape optimization scenarios require for the differentiated CAD models.
Digital twins of complex systems like aircraft need to manage heterogeneous data in large scale. The individual digital-twin realizations and implementations will be case-specific. However, key software competences and building blocks centrally developed and provided in a modular way allow for hierarchically setting up and operating predictive digital twins. A vital software ingredient is a flexible and adaptable data management that can handle various real-world or simulation data from different sources in a functional way. Provenance information must be recorded and provided to keep track of processes and data, e.g. in the context of virtual certification.