Development and validation of a phenomenological diesel engine combustion model

Abstract

Reduction of engine development time and cost is of primary interest for combustion engine manufacturers. This reduction is complicated by the fact that it has to be realized simultaneously with the fulfilling of ever more stringent legislative demands regarding emissions of NOx and Particulate Matter (PM) and demand from the (transportation) market to reduce fuel consumption. Meeting these demands has resulted in the addition of new technologies with their own degrees of freedom. Furthermore new advanced combustion concepts, such as low-temperature high-EGR combustion are being developed. These new concepts put great demands on in-cylinder charge condition (composition, temperature, pressure) and combustion control. This has led to complex and non-transparent engine control systems which increase development time and costs. The main contribution of this work is the development of a physically-based control oriented in-cylinder CI engine combustion model which predicts the interaction between a) the fuel injection rate and rate of heat release and b) the fuel injection process and the emission of NO (main component of NOx) and soot (main component of PM) emission for both conventional and advanced, high-EGR, CI combustion. Although the model is developed to be generally applicable, the validation process has been limited to only include Heavy Duty DI diesel engines. The NO, soot and heat release rate model combine (existing) zero and one-dimensional phenomenological models. These models describe the essential combustion physics and kinetics following the latest insights on the primarily mixing-controlled diesel spray-combustion process. This physical basis distinguishes the combustion model from the majority of combustion models for control applications found in literature that have a more empirical nature. It increases the predictive capabilities of the model and makes it more generic. As a result the amount of measurement data required for model identification will be lower and will reduce development time and costs. The soot and heat release rate model are based on existing state-of-the art models from literature. They are adapted to obtain a more accurate physical basis and to allow the prediction of advanced, high-EGR, combustion. For complete model identification and validation, this thesis presents a complete chain of measurements, comprising dedicated measurement set-ups. This chain starts with the characterization of the fuel injection equipment performing mass flow rate and momentum flow rate measurements, which allows reconstructing accurately the fuel injection rate corresponding to a performed engine test. Thereafter, the fuel spray behaviour is characterized, mainly by tuning existing spray-correlations on the basis of dedicated measurement data from an optically accessible high pressure chamber (The Eindhoven High Pressure Cell). Main emission model identification and validation is performed on the basis of measured heat release rates (i.e. as derived from measured in-cylinder pressure curves) from a research-type single-cylinder Heavy Duty diesel engine. The resulting identified and validated model is directly applied to a multi-cylinder Heavy Duty diesel engine. Results show that the heat release rate and emission formation models of NO and soot show satisfactory qualitative agreement with measurements for changes in fuelling (injection timing, pressure and quantity), EGR rate and engine speed, for both conventional and high-EGR combustion with conventional timing. These results validate the use of the model as predictive tool in the control development process. For NO, quantitative accuracy is in-line with or better than comparable state-of-the-art models with a more profound empirical nature. However, the obtained accuracy is currently not sufficient to use the model as virtual NO sensor in control applications. For accurate NO prediction a high accuracy is required on the prediction of the NO formation temperature evolution. The most important phenomena that influence this evolution have been examined and implemented by physically-based models that are supported by detailed computations on laminar flamelets and homogeneous reactor models. Simulation results show that the primary NO reducing phenomena are dissociation and flame straining by turbulence. Fuel evaporative cooling and hot soot particle radiative cooling are of secondary importance. The reduction in NO formation caused by hot gas radiative cooling only has a marginal influence. Entrainment of fresh, cold, oxidizer into the hot combustion products, in which the NO formation occurs, has to be accounted for to allow accurate NO formation. Equivalent reductions cannot be obtained by aforementioned NO reducing phenomena. The prediction of NO and soot emissions from a multi-cylinder engine is significantly more accurate when the emission formation of each individual cylinders are evaluated instead of using data from only one cylinder. This requires additional sensors, which is not desirable for cost reduction. Finally, the emission formation prediction accuracy when using predicted heat release rate profiles instead of ‘measured’ curves is examined. This shows that especially a very high accuracy on the instantaneous heat release rate is required while this is of much lesser importance for accurate NO formation prediction.