P. Beard

The study of soot oxidation and CO formation in internal combustion engine applications is the subject of numerous recent investigations. Their modeling is particularly important for Diesel operating conditions. Models have been developed recently at IFP to account for the complicated kinetic processes involved in CO / soot production and oxidation. This paper presents the equations for CO formation and oxidation based on a reduced 6 step chemistry model coupled with the PSK reduced soot production and oxidation mechanism. The species are accounted for in the conservation equations. Model development is done on the framework of the ECFM3Z engine combustion model. The global CO/soot model is first validated in a constant volume high pressure cell against LII measurements. Model parameters are adjusted and kept constant for the remaining of the simulations. Then, engine simulations are used to validate the model behavior in conventional and HCCI Diesel conditions. A parametric study based on experimental results includes: EGR (Exhaust Gas Recirculation), geometry and nozzle tip protrusion (NTP) variations at high and partial loads. Simulated pressure evolutions are compared to experimental measurements to ensure that the upstream mechanisms (injection, mixture formation and auto-ignition) are well reproduced. Then, the different post-oxidation models are tested. We compare CO and soot models alone as well as within the coupled approach against measurements. The trends and orders of magnitude are in good agreement for all cases. The limitations are due to the fast chemical times associated with the simplified mechanisms which can penalize the computation convergence. The coupled approach allows improvement of both CO and soot emissions modeling within reasonable CPU times.

A lumping model has been formulated to calculate the thermodynamic properties required for internal combustion engine multidimensional computations, including saturation pressure, latent heat of vaporization, liquid density, surface tension, viscosity, etc. This model consists firstly in reducing the analytical data to a single (i.e. pure) pseudo-component characterized by its molecular weight, critical pressure and temperature, and acentric factor. For a gasoline fuel, the required analytical data are those provided by gas chromatography. For a Diesel fuel, the required data are a true boiling point (TBP) distillation curve and the fuel density at a single temperature. This model provides a valuable tool for studying the effects of fuel physical properties upon the behavior of a vaporizing spray in a chamber, as well as upon direct injection gasoline and Diesel engines using the multidimensional (3D) KMB code. Computations have shown that the vaporization rate may increase up to 13% using a spring gasoline instead of a summer gasoline. Comparisons of engine calculations using the standard KIVA2 DF2 Diesel fuel and the new Diesel lump called DF1 have shown that the mixture preparation, combustion and emissions are affected by the thermodynamic properties used.

Three dimensional CFD tools are commonly used to simulate spray injection and combustion in DI Diesel engines. However typical computations are strongly mesh dependent. By now it is not possible to enhance grid resolution since it would violate the underlying assumptions for the Lagrangian liquid phase description. Besides, a full Eulerian approach with an adapted mesh is not practical at the moment mainly because of prohibitive computer requirements. Based on the Lagrangian-Eulerian approach, new approaches have been developed: the Coupled Lagrangian-Eulerian (CLE) model for the two-way coupling between the spray and the air flow and a new combustion model (CFM3Z) which allows a description of the fuel-oxidizer sub-grid mixing. The previously introduced CLE model consists in retaining vapor and momentum along parcel trajectories as long as the mesh is insufficient to resolve the steep gradients created by the spray. Vapor and momentum are gradually released following specific diffusion laws and criteria. The new combustion model (CFM3Z) is briefly described. In this approach, fuel and oxidizer, originally separated, are continuously mixed by the turbulent flow leading to the creation of a mixed zone. The local structure of the gas phase can then be divided into three zones: the first contains only fuel, the third only air (with EGR) and the second zone, where chemical reactions occur, contains both fuel and air mixed with combustion products. Simulations are performed using the KMB code, a modified version of KIVA-II. Comparisons of computed and experimental results of combustion in a DI Diesel engine are presented for several operating conditions. The influence of the engine speed and load, nozzle tip protrusion, piston bowl shape, EGR and pilot injection is investigated. Calculations show an improved predictivity of the engine performance compared to conventional models. The sensitivity of pollutant emissions to operating conditions is also better reproduced.

Soot models applied to Diesel combustion can be grouped into two classes, one based on the flamelet concept and the other based on the homogeneous reactor concept. The first assumes that the laminar diffusion flame structure of the reaction zone, in the mixture fraction space, is preserved while convected and strained by the turbulent flow. The second assumes that the properties of the reaction zone are locally homogeneous. Thus the aerodynamic and chemical reaction interactions are modeled with opposing assumptions: the first assumes fast chemistry, the second fast mixing. In this work, we first compare results obtained with a flamelet library approach to those with a homogeneous reactor approach. Recognizing that both types of models apply in different regions of Diesel combustion, we then propose a new approach for soot modeling in which they are coupled. The source terms for soot from each model are weighted and added with a criterion based on a comparison between characteristic times for turbulence and fuel oxidation. In all calculations a new spray model based on an extended Lagrangian-Eulerian approach and a new combustion model which allows a description of the fuel/oxidizer subgrid mixing are used. The results of the above three models are compared to experimental data in a high pressure combustion cell.

The Lagrangian-Eulerian approach is commonly used to simulate engine sprays. However typical spray computations are strongly mesh dependent. This is explained by an inadequate space resolution of the strong velocity and vapor concentration gradients. In Diesel sprays for instance, the Eulerian field is not properly computed close to the nozzle exit in the vicinity of the liquid phase. This causes an overestimated diffusion that leads to inaccuracies in the modeling of fuel-air mixing. By now it is not possible to enhance grid resolution since it would violate requested assumptions for the Lagrangian liquid phase description. Besides, a full Eulerian approach with an adapted mesh is not practical at the moment mainly because of prohibitive computer requirements. Keeping the Lagrangian-Eulerian approach, a new methodology is introduced: the full Lagrangian-Eulerian Coupling (CLE). It consists in retaining vapor and momentum along parcel trajectories as long as the mesh is unsufficient to resolve the steep gradients. Vapor and momentum are gradually released on the mesh following specified diffusion laws. This model is briefly described. CFD simulations are performed using the KMB code, a modified version of KIVA-II. Comparisons are presented of computed and measured penetrations of liquid and vapor in a high pressure, high temperature Diesel simulation cell. The new CLE model is shown to reproduce properly the evaporating spray structure and penetration while having a very limited mesh sensitivity. DI Diesel engine calculations using a rather crude mesh show a very good agreement of liquid and vapor penetrations. This insures an improved predictivity of the combustion in full load cases.

The flow-field of an automotive DI Diesel engine is characterized by experiments in a motored engine using Laser Doppler Velocimetry and by CFD simulations. Only one cylinder is active and a specific swirling intake duct is used. Various bowls with different shapes are investigated: fiat or W-shaped bowls, with or without re-entrant. The influence of engine speed is also studied. The mean velocity and turbulence evolutions are measured with back-scatter LDV experiments using an optical access in an extended piston. The simulations are performed using the KMB code, a modified version of KIVA-II. Along with the detailed flow-field description, integral quantities characterizing the flow are derived. The comparison between LDV data and CFD results is shown to be satisfactory. The effects of geometry and engine speed on spatial profiles and temporal evolution of mean and turbulent velocities are correctly reproduced. This investigation has allowed a better understanding of the piston bowl shape effects on the combustion chamber flow. Among these effects, the role of the re-entrant in the W-shaped bowls is shown to increase both swirl number and turbulence level around TDC.