Project Details
Projekt
Shape optimizations using AI agents to accelerate numerical flow field simulation and to control the island model for massive parallelization
Applicant
Professor Dr.-Ing. Stefan Riedelbauch
Subject Area
Hydraulic and Turbo Engines and Piston Engines
Fluid Mechanics
Fluid Mechanics
Term
since 2022
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 501932169
The overall goal of the project is to expand the automated design process for complex fluid mechanical components. In order to accelerate the design process with the general design system based on Computational Fluid Dynamics (CFD) simulations developed at the applying Institute, it is coupled with the island model for parallelizing the optimization. The island model divides the process into serial optimization runs on several interacting islands. Based on the results of the first funding period, for example the support of sensitivity analysis, automatic meshing and data-based flow field evaluation, the application of data-driven artificial intelligence methods will be further developed for the entire optimization process. During the first funding period, a data set for an axial hydraulic turbine was made publicly available for the development of trained agents, which will also be used as a reference for further agent development in the second funding period. Based on flow fields and / or geometry parameters, the fitness is estimated and a ranking is then determined. It serves as the basis for a data-based selection as well as migration operator for application in evolutionary optimization. These agents are continuously retrained during an optimization run. The aim is, for example, the prevention of time-consuming CFD simulations of expected very poor geometries or the identification of non-converging solutions. Both lead to a substantial reduction in design time. The promising agents from the first funding period with regard to mesh topology will be extended to three-dimensional flow areas in a further step, thus significantly increasing geometry flexibility. This will eliminate the need for time-consuming manual interaction. Describing the geometry of entire turbomachines requires a large number of parameters. Agents are developed that identify the essential parameters or generate parameter combinations. The reduced model developed from these agents is used to automatically design a high-quality machine that fulfills the specified properties in the best possible way. The developed agents in combination with the optimization method will be evaluated at the end of the project on a further, more complex “real world” example of a turbomachine.
DFG Programme
Priority Programmes
