Simulation and optimization of centrifugal compressors for pure hydrogen and hydrogen-methane mixtures

Abstract

Hybrid optimization methods for aerodynamic design combine gradient-free and gradient-based approaches to benefit from their advantages: wide design domain exploration from the first, and invariance of computational costs with the number of design variables (DV's) from the latter. Gradient-Enhanced Response-Surface Model (GERSM) coupled to optimizers have a similar potential as RSM-assisted gradient-free optimization, but, because sensitivities data are also used for the surrogate model calibration, the number of CFD solutions required for metamodel fitting is much smaller. Despite these characteristics, there is a gap on GERSM optimization of radial turbomachinery. Hence, the main goal of this Research Project is to conduct a GERSM optimization of a radial impeller and vaned diffuser operating with H2+CH4. The required data will be calculated with the open-source SU2 suite, which can solve multi-domain turbulent compressible and non-ideal CFD problems, and is also capable to efficiently calculate sensitivities with the adjoint-method. However, the code still cannot cope with optimization problems using thermodynamic models described by Look-up Tables (LuT), such as those of a mixture of real gases. Thus, the addressment of this gap in SU2 capabilities is also a goal of this Project. To summarize, the objectives of the Research Project for Internship Abroad are: 1) to become familiar with SU2; 2) to verify SU2 results against ANSYS CFX for a H2+CH4 radial compressor stage (rotor + vaned diffuser); 3) to implement the lines of code required for optimization with real gases and mixtures described by LuT's; 3) to optimize a baseline compressor stage with a gradient-based method; 4) to optimize a stage with a GERSM optimization. Finally, the results acquired will be utilized to compare costs, capabilities, and limitations of different optimization approaches. (AU)