Simultaneous measurements of the radial and the tangential components of velocity are obtained in a high-speed, direct-injection diesel engine typical of automotive applications. Results are presented for engine operation with fuel injection, but without combustion, for three different swirl ratios and four injection pressures. With the mean and fluctuating velocities, the
r-θ plane shear stress
and the mean flow gradients are obtained. Longitudinal and transverse length scales are also estimated via Taylor's hypothesis. The flow is shown to be sufficiently homogeneous and stationary to obtain meaningful length scale estimates. Concurrently, the flow and injection processes are simulated with KIVA-3V employing a RNG
k-ε turbulence model.
The measured turbulent kinetic energy
k,
r-θ plane mean strain rates ( 〈
Srθ〉, 〈
Srr〉, and 〈
Sθθ〉 ), deviatoric turbulent stresses
, and the
r-θ plane turbulence production terms are compared directly to the simulated results. The model predicts the qualitative trends in
k well, but under-predicts the magnitude of the late-cycle turbulence at the higher swirl ratios. The mean strain rates, turbulent stresses, and turbulence production terms generally agree qualitatively. Both the experimental and the simulated results indicate that redistribution of the mean flow angular momentum by the fuel injection event is an important source of late-cycle turbulence. This redistribution enhances
r-θ plane turbulence production at low swirl ratios through formation of unstable mean flow distributions with negative radial gradients in mean flow angular momentum. Additionally,
r-z plane vortical flow structures are formed by a competition between the inward displacement of high angular momentum fluid by the fuel jets and the centrifugal forces acting to force the high momentum fluid back to the bowl periphery. Model results indicate that these flow structures can also be important sources of turbulence at higher swirl ratios.
The measured length scales, mean strain rates, and turbulent kinetic energy are used to assess directly the validity of the isotropic eddy viscosity hypothesis. This modeling hypothesis is found to provide good estimates of
and
, despite the high levels of flow swirl. However, the data indicate that the modeled
is underestimated when unstable mean flow distributions exist. Further, under some conditions, no model based solely on local flow properties is likely to predict the measured
. Poor agreement is found between the measured and modeled
.
The under-prediction of
k at high swirl, coupled with the generally good agreement in 〈
Srθ〉,
, and
Prθ, suggest that the under-prediction of
k is predominantly due to an over-estimation of
ε after the injection event. This suggestion is further supported by a comparison of the temporal evolution of the measured and modeled length scales.