screw compressors, computational fluid dynamics, heat transfer
In order to design screw compressors for optimal performance, it is crucial to understand the complex fluid flow processes within them. Computational fluid dynamics (CFD) is one approach for doing so. Considerable progress has been made over the last several years in both commercial and academic solution packages for this application; however, due to the complex moving geometries of the screw rotors and the tight clearances between the moving parts, a major challenge that remains is the generation of numerical grids that are increasingly efficient, accurate, robust, and easily created. In this study, an alternate methodology for this problem is presented. The grid is created automatically at every time step based on the instantaneous geometry using a Cartesian cut-cell based method which preserves exactly the changing control volume shapes. Automatic mesh refinement is employed to adaptively increase mesh resolution where the flow variables have large gradients in order to resolve the large-scale flow structures. To address the problem of efficiently modeling the flows in the small clearance gaps, an empirical model is applied so that the cells within the gaps can remain relatively coarse. This removes a major bottleneck from the computational cost and allows more mesh resolution to be applied in accurately capturing the physics of the port flows. The effect of the thermal expansion on the gap sizes is accounted for by considering the heat transfer from the fluid to the solid walls and then periodically solving the solid to steady state using cycle-averaged heat transfer coefficients; the clearances therefore vary throughout the length of the rotors. The model is validated against experimental measurements of the internal pressure, mass flow rate, temperature, and power for two operating conditions. A global grid convergence study demonstrates the spatial and temporal convergence of the numerical model, and establishes necessary computational costs for varying levels of accuracy. It is shown that for the tested configurations, numerically accurate results are achieved with a total turn-around time that is low enough for practical use in engineering applications.