Search Results

A model of premixed turbulent combustion is modified for multi-dimensional computations of SI engines. This approach is based on the use of turbulent flame speed in order to suggest a closed balance equation for the mean combustion progress variable. The model includes a single unknown input parameter to be tuned. This model is tested against two sets of experimental data obtained by Bradley et al [17, 18 and 19] and Karpov and Severin [15] in fan-stirred bombs. The model quantitatively predicts the development of the turbulent flame speed, the effects of the initial pressure, temperature, and mixture composition on the turbulent flame speed, and the effects of r.m.s. turbulent velocity and burning mixture composition on the rate of the pressure rise. These results were computed with the same value of the aforementioned unknown input parameter of the model.

An expanding cylindrical laminar flame kernel affected by random external strain rates and diffusivity is numerically simulated in order to gain insight into the influence of small-scale turbulence on the combustion variability in engines. In the simulations, the kernel is strained, as a whole, by external velocity gradients randomly generated with either Gaussian or log-normal probability density functions. The influence of small-scale turbulent heat and mass transfer is modeled by turbulent diffusivity, the randomness of which is controlled by the fluctuations in the viscous dissipation averaged over the kernel volume. The computed results show that small-scale phenomena can substantially affect the quenching characteristics of a small flame kernel and the kernel growth history rj(t); the scatter of the computed curves of rf(t) being mainly controlled by the scatter of the duration of the initial stage of kernel development.

A Turbulent Flame Speed Closure Model is modified and implemented into the FIRE code for use in 3D computations of combustion in an SI-engine. The modifications are done to account for mixture inhomogeneity, and mixture compression through the dependency of local equivalence ratio, pressure and temperature on the chemical time scale and a global reaction time scale. The model is also subjected to further evaluation against experimental data, covering different mixture and turbulence conditions. The combustion process in a 4-valve pentroof combustion chamber is simulated and heat release rates and spatial flame distribution are evaluated against experimental data. The computations show good agreement with the experiments. The model has proven to be a robust and time effective simulation tool with good predictive ability.

Fan-stirred bombs, which are widely used worldwide, offer an unique opportunity to investigate basic features of S.I. engine combustion under well-defined experimental conditions. Extensive data bases on turbulent flame speeds have been generated by various groups utilizing such bombs. However, the use of these data bases is impeded by the fact that the measured flame speeds characterize an inherently transient process, i.e.. the speeds are time-dependent even if the pressure and the unburned gas temperature in the bomb are very close to the initial values; whereas the combustion theory and various models deal commonly with an asymptotically fully developed turbulent flame speed. The goal of this work is to test a method for evaluating the latter quantity by processing the published data on the flame radius growth, measured in expanding, statistically spherical, premixed flames.

Over the past years, the so-called Flame Speed Closure (FSC) model was shown to be a very promising tool for multi-dimensional simulations of premixed turbulent combustion in internal combustion and gas turbine engines. The laboratory tests and industrial applications of the model have been mainly limited to moderately turbulent flames. In the paper, three alternative versions of the FSC model, which yield different results at weak turbulence but similar results at moderate one, are discussed and numerically tested against recent experimental data reported by the Leeds [27,34] and Rouen [28] groups for expanding, statistically spherical, premixed, weakly turbulent flames. The computed and measured data on the mean combustion progress variable profiles, mean flame brush thickness development, and observed flame speeds are compared in order to assess and rank the submodels discussed.

In the first part of the paper, a generalization of the turbulent diffusivity concept is considered and a generalized diffusion coefficient is introduced to account for the development of turbulent diffusivity, pressure-driven countergradient transport, and effects of chemical reactions on turbulent scalar flux. The behavior of the generalized diffusivity is numerically studied in the 1-D statistically planar case and the contributions of the aforementioned processes to the diffusivity are assessed. In the second part of the paper, the generalized diffusivity is incorporated into the Flame Speed Closure (FSC) model of premixed turbulent combustion and the extended FSC model is applied to simulate recent experiments performed using the Leeds fan-stirred bomb. The extended FSC model well predicts the speed, thickness and structure of statistically spherical, premixed, turbulent flames that expand in the bomb after spark ignition.

Published experimental data obtained in well-defined simple cases are discussed in order to qualitatively test various models of premixed turbulent combustion, utilized in multi-dimensional numerical simulations of SI engines. An analysis of such data indicates that there exist several unresolved issues important for flame propagation in SI engines. Two of them, pressure dependence of turbulent flame speed St and turbulent flame development, are discussed in the paper. First, existing experimental data indicate an increase in St by pressure despite the marked decrease in the laminar burning velocity SL by P. Although this well established trend appears to be of substantial importance for SI engine applications, many combustion models utilize SL as the sole mixture characteristic and, hence, predict similar dependencies both of St and of SL on P, contrary to the aforementioned experimental results.