Model-based control of a turbocharged diesel engine with high- and low-pressure exhaust gas recirculation

Modern Diesel engines fulfil challenging requirements for emission limits, fuel consumption and ride comfort by numerous modular combinable components and mechatronical actuators. These components are utilised for precondition and aftertreatment of air, fuel and exhaust gas, which is involved in the combustion process. In this dissertation a methodology for a model-based function development with semi-physical engine models for control of air path quantities of an exemplary Diesel engine with high-pressure (HP-EGR) and low-pressure exhaust gas recirculation (LP-EGR) is developed. In this framework for function development black-box models for stationary and dynamical emission formation are utilised to optimise reference values for the air path control and to rate the developed control scheme with regard to the cumulated driving cycle emissions of the new European driving cycle (NEDC). A combination of HP-EGR and LP-EGR represents a novel approach to significantly lower the particulate and NOx emissions of Diesel engines. A semi-physical mean value engine model with lumped parameters is the base to analyse the system properties of the complex air path. In doing so, the additional LP-EGR shows only minor influences to the quantities charge air pressure and HP-EGR, while there are significant influences of these quantities on the LP-EGR mass flow rate. Furthermore, the LP-EGR is characterised by significant gas propagation times in the intake and exhaust system. These delays are modelled by a gas composition model, which is incorporated into the control scheme. NOx and particulate emissions as well as engine torque are stationary modelled by local polynomial models with input quantities of the combustion process. These quantities are air mass flow rate, charge air pressure, intake temperature and crank angle of 50% mass fraction burned. A bilinear interpolation between engine speed and injection quantity transforms local polynomial models into global models. Models for the dynamical emission formation are given by considering the combustion as a batch process. Consequently all dynamics are included in the quantities of the cylinder charge at intake valve closing and the emission measurement dynamics. Thus, a combination of a dynamical gas composition model, stationary emission models and models for the emission measurement dynamics yield the dynamical course of the engine emissions. The investigated system properties and the emission models deliver the control variables charge air pressure, air content and intake temperature for the engine with VGT-turbocharger, HP- and LP-EGR. A stationary optimisation with regard to emissions and engine torque delivers reference values for the air path control and further shows the potential of the LP-EGR to lower the emissions. Due to the multi-variable characteristics of the air path with different dynamics, there are increased dynamical emissions at engine transients. These dynamical emissions are lowered by dynamical optimised reference values for the air path control. Generally, the air path is a strongly nonlinear process and the multitude of engine variants and engine operation modes result in a trade-off between achievable control quality, control robustness and number of control parameter sets. A semi-physical feedforward control, which is based upon parameterised model relationships of the mean value engine model delivers a good response to setpoint changes. Thus, the disturbance rejection can be achieved by relatively simple controllers. This results in an significantly lower application effort of control parameters and allows by its modular structure to exchange engine components without the drawback to completely re-parameterise the control parameters. A reference value transformation with modelled states of the gas composition model compensates long gas propagation times in the intake and exhaust system and delivers an optimal air content in the cylinder charge. All control concepts are validated with measurements at the engine test bench. Finally, the derived control concepts for the LP-EGR are compared to the classical HP-EGR control with regard to the cumulated driving cycle emissions. In this investigation the proportion of stationary and dynamical emissions is clearly quantified. In a nutshell this dissertation is an important contribution for model-based optimisation and function development for the air path control of Diesel engines. The given combination of models for dynamical emission formation, dynamically optimised reference values for the air path control and semi-physical control design are a holistic framework to master the complexity and variance of future Diesel and gasoline engines.