Development Of A Generalized RNG Turbulence Model And Its Application To Computations Of A Diesel Engine Operating In A Low Combustion Regime

Multi-dimensional numerical simulations of the combustion processes in internal combustion(IC) engines are rapidly being accepted as an important tool for modern engine product design and optimization. As one of the most important and fundamental parts, numerical simulation of turbulence plays a significantly important role in the simulations of the whole engine combustion processes. However, it is extremely complex and hard to model the engine turbulence motion accurately, due to its strong transient, strong compressible, strong roational and anisotropical characteristics. In recent years, despite continued advances in computing power, the computational cost is still very high using state-of-the-art Large Eddy Simulation (LES), and even the Direct Numerical Simulation (DNS) techniques, although they offer advantages for engine simulations such as higher-fidelity predictions, and the ability to resovle cycle-to-cycle variations in flow structure and combustion behavior compared to the traditional Reynodels Averaged Navier-Stokes (RANS) models. Simulations based on the higher computational efficient and better practicability RANS approaches are likely to remain the dominant method for simulating engine flows for the next decades. Despite several decades of use, however, the simulation accuracy of these RANS approaches to model the engine turbulence has not yet been satisfactorily resolved. The main reason is that these RANS models can not regenerate the physical characteristics of the turbulence completely, which leads to the inaccurate coupling of the gas-liquid phase. As the results, the results computed by the turbulence models do not agree well with the experimental data. Thus, research on improving the accuracy of RANS engineering turbulence models and theirs engineering applications still has significant practical values. This is also the key task of the present work.The performances of the standard k-ε model and renormalization group (RNG) k-ε model adopted widely in industry are assessed comprehensively using the numercial computation method, and the experimental data of the imcompressible air jet and compressible helium jet flows, the Laser Doppler Velocimeter (LDV) and Particle Image Velocimetry (PIV) measurements of the complex engine flows, and the DNS data. The results show both of the two turbulence models do not perform well in these cases. The predicted turbulence quantities, especially for the growth of the turbulent length scales are almost unable to match the experiments. One potential approach to resolving this problem is to employ the model coefficients that vary with the characteristic strate rate (or turbulent-to-mean time scale ratio) of the flow.For the argument mentioned above, a generalized RNG closure model based on the’dimensionality’of the flow strain rate was proposed to improve the predictions of the turbulence quantities. In this modeling approach the model main coefficients are all constructed as functions of the local flow strain rate under the rapid distortion limit assumption of turbulence flow and the isotropic turbulence decay analysis. In order to validate the ability of the proposed turbulenc model, this model is applied to several classic development test cases for turbulence models in present dissertation. As the results, the generalized model significanltly improved the flow predictions compared with the results of the standard k-ε model and the standard RNG turbulence model, and agreed better with the related experimental data.In the engine combustion case, the thermal field is strongly coupled with the flow dynamics and the flow turbulence plays an important role in the combustion processes, particulately, in a mixing controlled combustion regime. Accordingly, the performance of the generalized RNG turbulence model was further examined in a direct injection (DI) diesel engine undergoing a partial premixed compression ignition (PPCI) controlled low temperature combustion. The simulated mixture formation of fuel and air, combustion heat release, and engine emissions such as the unburned hydrocarbon (UHC), carbon monoxide (CO), nitrogen oxide (NOx) and Soot are compared with the results predicted by the standard RNG k-εmodel, and the available experimental data. The results showed that the generalized RNG turbulence model improved significantly the results of the diesel engine. The predictions were in better agreement with the experimental data, especially for the predictions of UHC formations and theirs space distributions.