Abstract
Ozone depletion and global warming are major drivers for the development of cooling systems. Accordingly, new governmental regulations for the equipment are continuously introduced. The redesign of corresponding centrifugal compressors for new refrigerants is therefore necessary and remains a very challenging task. The vicinity to the two-phase region requires accurate real gas models, and the complexity of the flow patterns a robust numerical integration of the underlying governing equations. This is particularly true for the considered low flow coefficient two-stage centrifugal compressor. Classical design and optimization methods are limited in their flexibility because of the geometry parametrization, thus affecting the development of the next generation of turbo compressors. This article describes the development and application of a fully coupled, pressure-based computational fluid dynamics (CFD) framework, incorporating a highly flexible discrete adjoint method used for redesign and optimization purposes. The maturity of the underlying CFD and of the optimization algorithm makes it possible to account for real gas and turbulence effects, as well as multiple mixing-plane stage interfaces. Accordingly, it is possible to fully exploit the major advantage of this gradient-based optimization strategy, the independency between computational effort and the number of degrees-of-freedom in the geometrical description. For a two-stage centrifugal compressor, the return channel is optimized to achieve an overall reduction in entropy generation. Although based on a single-point optimization, the resulting performance characteristic shows a consistent improvement over the whole operating range with maximum values of up to 2%. The resulting geometry is manufactured and experimentally tested on the test rig. The computationally obtained efficiency increase can be confirmed by the experimental data.
