Development and software in the loop validation of a Model-based water injection combustion controller for a GDI TC engine

Turbocharged (TC) engines work at high Indicated Mean Effective Pressure (IMEP), resulting in high in-cylinder pressures and temperatures, improving thermal efficiency, but at the same time increasing the possibility of abnormal combustion events like knock and pre-ignition. To mitigate knocking conditions, engine control systems typically apply spark retard and/or mixture enrichment, which decrease indicated work and increase specific fuel consumption. Many recent studies have advocated Water Injection (WI) as an approach to replace or supplement existing knock mitigation techniques. Water reduces temperatures in the end gas zone due to its high latent heat of vaporization. Furthermore, water vapor acts as diluent in the combustion process. In this paper, the development of a novel closed-loop, model-based WI controller is discussed and critically analyzed. The innovative contribution of this paper is to propose a control strategy based on an analytical combustion model that describes the relationship between the combustion phase and the Spark Advance (SA), considering also the effects of the injected water mass. Such model is calibrated with experimental data acquired during dedicated experimental tests on a GDI TC engine, equipped with a prototype Port Water Injection (PWI) system. At first the WI setup is described, and the main experimental data are presented and processed for model identification. Two algorithm versions are then explained in detail and implemented in Simulink environment, with a Real-Time (RT) oriented approach. In the last part of this work, the WI control strategy is tested in a Software in the Loop (SiL) system, coupled with a one-dimensional Fast Running Engine Model (FRM). The controller is tested on several engine points in steady state and transient conditions and the Root Mean Squared Error (RMSE) is calculated for the control targets. In this way, the performance of the model-based controller is verified, and the two versions of the algorithm are quantitatively compared.