Investigation of the effects of engine design and operating parameters on performance and emissions of compression ignition engines by statistical methods

Significant problems can be experienced with performance and emission formation in the combustion process of compression-ignition (CI) engines, in which natural gas is used as the main fuel. One of the main reasons of the problem is the effects of parameters of regarding piston geometry and combustion process. These parameters determine important factors such as the mixing ratio of the air and natural gas, the formation of a homogeneous mixture, the zone inside the piston where the combustion starts, the ignition time and the combustion velocity. The effects of these engine design and operating parameters can be explained by surrogate models based on real models belong to data from experimental and computational fluid dynamics (CFD) analysis. The advantages of these methods are to obtain a sufficient number of estimation from a small number of data in which more parameters applied, reduce the actual experiment cost and shorten the computing time for CFD numerical analysis. In this study, the effects of piston geometry and combustion process parameters on performance and emission formation in methane fueled CI engines are investigated. In order to evaluate these effects, linear regression and Kriging statistical methods were used for the CFD data generated from the efficient sampling points by the DoE. The study consists of forth main parts. In the first part, experimental work of diesel engine belonging to three different piston geometries made in İTÜ Automotive Laboratory. Pressure and emission results obtained from experiments were evaluated for verification of numerical method used in the study. Numerical method confirmed by experimental results and the CFD-based "AVL Fire" combustion analysis program was used to generate test data. The second part is the combination of the DoE with a CFD based combustion analysis in order to obtain real data. In order that real data are composed of efficient sampling points, latin hypercube sampling was applied to the six independent parameters with being boundaries determined. Real CFD data were obtained by modeling 35 sampling points in the AVL Combustion Analysis program. The third part is the establishment of substituted models by applying the linear regression analysis and Kriging methods to the real CFD data. At this point, the correctness of statistical models established were tested and error rates were determined. When the estimation error rates were compared, it is seen that Kriging method had an less estimation error. At the final part, the results of the Kriging method was evaluated with less error rate. Linear regression model was established to test the advantage of creating faster solutions with a simpler structure. Despite the fact that it has a simple polynomial function for setting up easier, error rates are found to be high. Due to low error distribution, the Kriging method was selected for estimating the effects of engine design and operating parameters in this study. Since the Kriging model takes form according to the data, it has been revealed that its representation and estimation ability is much stronger than the linear regression model. Due to low error rates of the Kriging models, the Kriging method was applied in this study. A new and simple approach was applied for evaluating the Kriging method. Kriging method results were evaluated by improving a new approach. The purpose of this new evaluation method is to obtain correct estimation information from the Kriging maps. This evaluation method is explained in the 75 Kriging map charts obtained. In order to determine the effects of engine design and operating parameters improving performance and emissions, a common good region was formed from the Kriging models. From the Kriging maps separated by regions, the effects of dependent parameters on the reduction and increase of the independent parameters for each region are explained. A general matrix table was constructed by determining the effects of the independent parameters on the dependent parameters of decreasing and increasing states. This method, which is a different approach in evaluating the results of the Kriging method, was first applied to methane-fueled compression ignition engines in this study. With this method of evaluation, more general estimates were obtained using less data. This case study is the investigation of the effect of engine design and operating parameters on performance and emission formation in methane-fuelled compression ignited engines with two different statistical methods. Two different statistical methods were tried to solve the complex parameter design problem. The purpose of the trying of the linear regression model was to produce fast solutions due to its simple structure. But it generated more errors, as it did not have enough flexibility. The purpose of establishing the Kriging model is that it has a high representation and estimation ability, since it is a model that takes shape according to the data. One of the most important results of this study is that the Kriging method has a much stronger estimation capacity than the linear regression method. In this study, Kriging Soot and CO models could not be established and estimation information could not be obtained. Additionally, it was not reached significant information that can explain the effect of the equivalence ratio parameter on the best results of Kriging. Kriging maps commonly used in geostatic were analysed to explain the effects of design and operating parameters of methane fueled CI engines. In that manner, a new approach was used for simplifying the analysis of the number of 105 Kriging contour maps. The Kriging method predicted correctly that the narrowing of the bowl geometry and the late injection of fuel into the centre of the bowl, improved engine performance and emissions. The effect of piston geometry parameters that maximizes performance and emission formation at specified intervals; the effect of decreasing tendency of the outer bowl diameter is 100%. The effects of decreasing tendency of center depth and inner radius of bowl are 60% and 50%, respectively. The effect of combustion process parameters that maximizes the performance and emission formation at the specified intervals; the effect of delay at the fuel injection time is 62%. The improving effect of the fuel injection angle parameter is the tendency to spray towards the bowl center, with an effect of 30%. Significant information that could explain the effect of the fuel surplus ratio at the specified intervals was not reached. As a result, it was determined that the estimation information obtained from this study is compatible with similar studies in the literature. It is thought that the experimental verification of the estimation results of the Kriging method, which is compatible with the literature, will contribute positively to similar studies to be made in the future. The most important feature of this study is to transform a complex 6-d parameter structure to simpler surrogate models by Kriging method. It is also assessed that the accurate estimations of this study will contribute to studies on methane-fuelled CI engine concerning to improving performance and emissions. The most important feature of this study is to transform a complex 6-d parameter structure to simpler surrogate models by Kriging method. It is also assessed that the accurate estimations of this study will contribute to studies on methane-fuelled CI engine concerning to improving performance and emissions.