Influence of approaches in CFD Solvers on Performance Prediction in Screw Compressors

Computational Fluid Dynamics (CFD) offers insight into screw compressor designs beyond the capabilities of other conventional methods. It allows evaluation of local flow patterns which influence performance but are difficult or impossible to investigate experimentally. Implementation of CFD in these machines is challenging due to the physics of the flow, the properties of the working fluids and the complexity of flow passages which change size and position. This is additionally challenged by a lack of methodologies available to generate the meshes required for the full three dimensional transient simulations. Commercially available CFD solvers need to fully interact with customized grid generators to enable resolution of grid deformation during a flow solution. However, the factors that influence flow predictions are not only related to grids but also to the approach which CFD solvers use to calculate distribution of parameters such as pressure, velocities, temperatures, etc. In this paper, two approaches most commonly used in commercial CFD software are compared and analysed. The first is a segregated cell-centre based solver and the second is a coupled vertex-centre based solver. Both are pressure based finite volume solvers. Customized grid generation software is used for meshing of moving rotors and flow domains around the rotors in an oil free air screw compressor with ‘N’ rotor profile of 3/5 lobe combination. The deforming rotor grid is maintained as identical in both solvers. The performance predictions obtained by calculations with these two CFD models are compared with measurements obtained on the test compressor in the City University London test rig. The comparison includes pressure in the compressor chamber, mass flow rate, indicated power and the volumetric efficiency. The study reveals differences between the results obtained by two different solvers and the experimental results. Analysis presented in this paper provides a good basis for further consideration of differencing schemes and other characteristics and settings for different CFD solvers in order to achieve accurate predictions of flows in positive displacement machines.