Simultaneous two-component measurements of gas velocity and multi-dimensional numerical simulation are employed to characterize the evolution of the in-cylinder turbulent flow structure in a re-entrant bowl-in-piston engine under motored operation. The evolution of the mean flow field, turbulence energy, turbulent length scales, and the various terms contributing to the production of the turbulence energy are correlated and compared, with the objectives of clarifying the physical mechanisms and flow structures that dominate the turbulence production and of identifying the source of discrepancies between the measured and simulated turbulence fields. Additionally, the applicability of the linear turbulent stress modeling hypothesis employed in the k-ε model is assessed using the experimental mean flow gradients, turbulence energy, and length scales.
Production by the isotropic portion of the normal stresses (bulk compression) is the principal source of turbulence energy at low swirl ratios, and remains significant for all swirl ratios. At higher swirl ratios, production associated with radial and axial gradients in the mean swirl velocity generally dominate near TDC. Squish produces turbulence energy predominantly by modification of the swirl velocity distribution, not directly through production associated with gradients in the squish velocity. Consequently, increased swirl and control of the spatial distribution of the swirl velocity appears the most effective method of increasing combustion chamber turbulence. Bulk expansion, leading to negative production by the normal stresses, is a significant sink of turbulence energy beyond TDC and usually dominates over all other production terms. Additional sources of turbulence energy operative at specific locations and periods of the cycle are identified and discussed.
The measured and simulated mean flow fields show generally good agreement. Observed mean flow discrepancies are consistent with inaccuracy in the prediction of the total in-cylinder angular momentum (swirl ratio). The major source of differences observed between the measured and simulated turbulent kinetic energy fields is underestimation of k well before TDC and, consequently, underestimation of further turbulence production by the isotropic portion of the normal stresses. Differences in the production by the anisotropic (shear) stresses are also observed, and are often consistent with discrepancies found when the stress model is applied to the experimental data.
The linear turbulent stress model generally produces reasonable estimates of the shear stress at low swirl. For high swirl ratios, however, erroneous stresses can be predicted resulting in significant turbulence production that is not supported by the measurements. The normal stresses are found to be qualitatively well predicted only when turbulence production by shear is not significant.