Measurements and Modeling of Reynolds Stress and Turbulence Production in a Swirl-Supported, Direct-Injection Diesel Engine

Abstract

Measured and numerically predicted components of the mean rate-of-strain tensor _Sij_ and the Reynolds stress _u_ru___ are examined and compared to elucidate the source and scrutinize the modeling of late-cycle turbulence production in swirl-supported, direct-injection diesel engines. The experiments are performed with combustion in the engine inhibited, to eliminate the complicating influence of heat release on turbulence generation and to reduce the problem to one more closely approximating constant-density turbulence. Both the measurements and the calculations indicate that the primary influence of the mean flow swirl on turbulence production is confined to two separate periods: (1) shortly after the end of injection and (2) in the late-cycle period, when large positive levels of _u_r,u___ are observed. Formation of the positive Reynolds stress coincides with the development of a negative radial gradient in mean angular momentum, indicating an unstable mean flow field. At this time, the measured velocity fluctuations show a large increase, approximately doubling in magnitude compared to fluctuations measured without fuel injection. Predicted velocity fluctuations, obtained via k-_ turbulence modeling, show a similar late-cycle increase, although the magnitude of the increase is not quantitatively captured. To evaluate its applicability during the period in which the unstable, negative radial gradient in angular momentum is present, the isotropic eddy viscosity hypothesis is examined. The Reynolds stress estimated from the measured _Sr__ using the eddy viscosity hypothesis is found to mimic the measured stress with reasonable accuracy, and the measured and calculated r-_ plane turbulence production terms are shown to have excellent qualitative and quantitative agreement. The quantitative agreement, however, appears largely providential, as the measured and predicted values of _Sr__ differ by a factor of 2. This discrepancy is compensated for by the underpredicted turbulent kinetic energy and seemingly high values of the dissipation.