Aerodynamic Optimisation of a Centrifugal Compressor Return Channel and U-Turn With Genetic Algorithms

Abstract

This paper presents the numerical analysis and aerodynamic optimisation of a return channel system and U-turn for multi stage single shaft centrifugal compressor machinery. An optimisation program is used, based on an OpenSource 3D viscous Navier Stokes Solver and an OpenSource evolutionary optimisation algorithm. The coupling between the Navier Stokes flow solver and the evolutionary algorithm is direct, without any meta-model such as ANN (Artificial Neural Network) or a Kriging model. Two different U-turn geometries — based on two different flow inlet profiles — are optimised, analysed and finally coupled to a return channel blade geometry, which is optimised itself. The optimisation of the return channel blade has been set up in two different ways, once with parallel meridional hub and shroud walls and once with a non-parallel optimised meridional shroud contour line. Results show a reduction of total pressure loss by 3% for the optimised blade shapes compared to a standardised blade shape of constant change of tangential momentum. A significant reduction in pressure loss coefficient from 0.825 to 0.627 could be achieved by the optimisation of the meridional contour lines of the return channel. Furthermore, the effect of non-sufficient cross -sectional area at the return channel trailing edge on the overall performance is presented.