Centrifugal Compressors Return Channel Optimization by Means of Advanced 3D CFD

Abstract

In the Oil&Gas industry, the quest for turbomachinery high efficiency is continuously pushing technology towards new improvements. Rotor (impeller) optimization is a rather established practice in the centrifugal compressors design both in the scientific literature and in the industrial experience, but the same focus is not always applied to statoric components optimization. This is due to the smaller impact of plenums and return channels on the overall compressor performance in comparison to the impeller. However, further improvements in the design of centrifugal compressors stage have to deal with the loss minimization of statoric parts, also considering the advanced level of aerodynamic detail reached by Original Equipment Manufacturers (OEMs). In the present paper, the optimization of the return channel is performed by means of 3D Computational Fluid Dynamics (CFD) with the objective of loss coefficient minimization. The CFD simulations are run with a non-commercial proprietary software (Tacoma) considering steady flow with Reynolds Averaged Navier Stokes (RANS) approach and turbulence k-ω model with strain correction; full validation of the employed method was performed in the past against experimental campaigns and is available in the literature. In order to reproduce as much as possible full stage operating conditions, simulations should include impeller, diffuser and return channel, but the computational cost of an optimization with many iterations required the reduction of the domain. In order to preserve simulation accuracy at the same time, flow profiles at impeller exit have been imposed at the considered computational domain inlet (i.e. diffuser inlet), whilst cavity effects and secondary flows have been considered adding source terms taken from the full stage simulation of the baseline geometry. Return channel blades have been parameterized in terms of angle distributions with Bézier curves at hub and shroud, for a total of 18 Bézier poles. Each different design has been simulated with its speedline consisting of 7 operating conditions, whereas progressive optimization based on response surfaces have been considered starting from an initial Design of Experiment (DOE). Over 100 designs have been simulated and the most efficient allows a 20% loss coefficient reduction on a real case stage design at design point.