Analysis of 3 Dimensional Turbine Flow by Using Mode Decomposition Techniques

Abstract

Today one of the most popular ways of lowering the fuel consumption and emissions of the Internal Combustion Engine (ICE) is by downsizing the engine. Downsizing means that the swept volumes of the cylinders are decreased; this lowers the frictional and thermal losses. By combining the downsizing with a well matched turbocharger system the performance is preserved while the advantages are retained. Since more and more of the development work is being performed by simulations there is an increasing need for more accurate methods. These methods are more complex and require more resources than the simpler, faster and more robust models used today. In this study Large Eddy Simulations (LES) of the unsteady flow in a radial turbine designed for a gasoline ICE has been performed and analyzed. The flow inside the turbine is highly 3 dimensional, pulsating and characterized by secondary flow motions and high curvatures. All these are reasons for which the method of choice should be LES. LES is able to resolve a large range of scales and capture the flow dynamics. The considered case concerns a non-pulsating flow condition but with engine like mass flow and temperature. Post-processing tools based on Proper Orthogonal Decomposition (POD) and Dynamic Mode Decomposition (DMD) are used to analyse the large amount of LES based flow data. The POD method is used to investigate the energy content of the dominant, large structures present in the flow. The DMD method on the other hand is used to reveal the flow structures responsible for specific frequencies found in the flow field. Preliminary data show a fair agreement between experimental data and LES results in terms of predicting the turbine performance parameters.