Diesel Engine Injection Rate Shape Optimization Using Genetic Algorithms And Multidimensional Modeling

Author : Jost Weber
Publisher : Cuvillier Verlag
ISBN 13 : 373692724X
Total Pages : 266 pages
Book Rating : 4.7/5 (369 download)

DOWNLOAD NOW!

Download or read book Optimization Methods for the Mixture Formation and Combustion Process in Diesel Engines written by Jost Weber and published by Cuvillier Verlag. This book was released on 2008-09-02 with total page 266 pages. Available in PDF, EPUB and Kindle. Book excerpt: The optimization of the combustion and mixture formation process in Diesel engines by CFD simulations requires a reliable model approach as a pre-requisite in order to predict combustion and emissions. A general and commonly used model for the liquid spray is the discrete droplet model. Sub-models for droplet breakup, collision and coalescence, and evaporation are available in the CFD code. With regard to combustion, the flamelet model approach is interactively coupled with the CFD code, known as RIF model. It benefits from a one-dimensional description of the thin reaction zone in the flame. By this approach, a detailed reaction mechanism for the model fuel can be used. Sub-mechanisms for NOx formation and a soot model are included. The reaction mechanism has been modified in this work to account for a correct ignition delay and heat-release at low-temperature conditions e.g. in the PCCI combustion. The modeling of the mixture formation in a spray contains uncertainties in the model constants and initial conditions. Spray data is required to calibrate the spray model. At least, the spray penetration has to be measured under engine like conditions as performed in a spray chamber. The spray penetration is interpreted as a criterion for the mass and momentum exchange between the spray and the surrounding gas on a macroscopic level. Finding a good agreement for the spray penetration between simulation and experiment defines an optimization problem. That agreement is expressed in an Euclidean norm as a merit function. The objective is to minimize this merit function. The search for an appropriate set of spray model parameters and initial conditions is denoted here as calibration of the spray model. Six parameters have been identified, spanning a six dimensional parameter space. A manual search is not feasible anymore but the implemented Genetic Algorithm is suitable to find a global optimum where a good agreement between measured and simulated spray penetration is obtained. If the same spray parameters are applied to a virtual engine case, a similar good agreement is achieved although the mesh resolution is much finer and the mesh topology is different than for the spray chamber simulation. From this result, spray data for engine simulations should be provided and be used for sake of calibration before the engine simulation is conducted. Additionally data is obtained by PDA measurements at discrete points in the spray. That measurement technique is, however, limited to less dense areas. Nevertheless, it shows that also local data is in agreement with the simulation data. Agreement with spray penetration is thus a relatively good choice and accounts also for the physics on a local or microscopic level. That hypothesis is well supported by the data from the ethanol spray calibration. The excellent agreement with regard to the global spray penetration is reflected by the 2D comparison of liquid and vapor fuel concentrations and temperature, respectively. Furthermore, a similar good agreement in spray penetration is obtained if the breakup and collision model is not used. In that case, the spray penetration is only controlled by the evaporation process. The Genetic Algorithm finds a point in the parameter space with an initial SMR that is of the order of size of the outcome of the secondary droplet breakup. However in engine simulations, spray data is not always available. In that case the spray parameters have to be adjusted. That adjustment is carried out following a methodology that is presented in this work. Mainly, SOI and EGR variations have to be used to calibrated the spray and combustion model. That approach has been investigated for three different engine data sets for conventional and PCCI combustion mode. On the Cummins QSX engine, a conventional combustion has been studied. Spray parameters are subject of adjustment. On the Duramax 6600 Diesel engine, a conventional and PCCI combustion mode are investigated. For the PCCI combustion mode, the reaction mechanism is modified in order to account for a correct ignition delay in the low temperature combustion regime. The comparison between engine data and results from the simulation indicates a good agreement for the combustion and engineout emissions. On the Duramax full load case, most uncertainties are addressed to the spray-wall interaction. Uncertainties from physical not well based models will always occur in the engine simulation. Therefore, calibration of these models is a mean to quantify its influence and minimize the discrepancies.