Assessment of the numerical and experimental methodology to predict EGR cylinder-to-cylinder dispersion and pollutant emissions

Abstract

EGR cylinder-to-cylinder dispersion poses an important issue for piston engines, since it increases NOx and particulate matter (PM) emissions. In this work, the EGR distribution on a 6-cylinder intake manifold is analyzed by means of experiments, 0D/1D engine modeling and 3D CFD simulations at three different working points. Using a comprehensive set of measurements, statistical regressions for NOx and PM emissions are developed and employed to quantify the sensitivity of numerical configuration to EGR dispersion and subsequent increase of pollutants. CFD mesh and time-step size independence studies are conducted, taking into account their interrelation through the Courant number. The obtained numerical configuration is validated against experimental measurements, considering different unsteady RANS turbulence submodels ( $k-\epsilon$ and $\mathrm{SSTk}-\omega$ ) as well as the inviscid case. The agreement of the different approaches is quite sensitive to the operating conditions, obtaining root mean square errors for the average cylinder-to-cylinder EGR distribution between 1% and 17% and for the transient $C{O}_2$ traces between 8% and 29%. However, for the worst-case scenario, the error in NOx and PM emissions prediction is below 2%. The regressions are employed to find a greater EGR distribution impact on pollutants when EGR rate or dispersion are increased. Flow investigation reveals the underlying reasons for the discrepancies and similarities between the predictions of the different turbulence submodels. A statistical analysis shows the significant errors that average $C{O}_2$ probes make when assessing EGR cylinder-to-cylinder distribution, which is explain by the flow heterogeneity at some operating conditions.