Large-Eddy Simulation and Analysis of Turbulent Flows in a Motored Spark-Ignition Engine

Abstract

Advanced research in Spark-ignition (SI) engines has been focused on dilute-combustion concepts. For example, exhaust-gas recirculation is used to lower both fuel consumption and pollutant emissions while maintaining or enhancing engine performance, durability and reliability. These advancements achieve higher engine efficiency but may deteriorate combustion stability. One symptom of instability is a large cycle-to-cycle variation (CCV) in the in-cylinder flow and combustion metrics. Large-eddy simulation (LES) is a computational fluid dynamics (CFD) method that may be used to quantify CCV through numerical prediction of the turbulent flow and combustion processes in the engine over many engine cycles.

In this study, we focus on evaluating the capability of LES to predict the in-cylinder flows and gas exchange processes in a motored SI engine installed with a transparent combustion chamber (TCC), comparing with recently published data. Numerical simulations are performed using the commercial CFD software, ANSYS Forte, employing a classical Smagorinsky sub-grid-scale (SGS) model for the LES approach. Two important aspects of the model, namely the coefficient of sub-grid viscosity used in the Smagorinsky model, and the numerical scheme for discretizing the convection term in the momentum transport equation, are evaluated.

Simulations are performed for 20 consecutive engine cycles after the simulation setup is validated by the predicted in-cylinder pressure, trapped mass, and temperature data. LES-predicted phase-averaged-mean and root-mean-square (RMS) velocity fields are compared with high-speed particle image velocimetry (PIV) data. The comparison and analysis are performed at two crank angles, representing intake and compression strokes, and on two different planes for measurement in the engine combustion chamber. A proper orthogonal decomposition (POD) technique is applied to quantify CCV in both the LES results and the PIV data, to provide a quantitative assessment of the predictions from LES. The flow field statistics predicted by the LES-Smagorinsky model match well with experimental results. Based on these simulation results, optimal practices for the use of Smagorinsky model with respect to the numerical schemes are summarized.