Data-Based In-Cylinder Pressure Model including Cyclic Variations of an RCCI Engine

Abstract

For advanced pre-mixed combustion concepts, Cylinder Pressure-Based Control is a key concept for robust operation. It also opens the possibility for on-line heat release shaping. For cost and time efficient development of these controllers, fast control-oriented combustion models that predict average in-cylinder pressure traces have been proposed. However, they are not able to capture cyclic variations. In this study, a data-based modelling procedure is proposed to predict the in-cylinder pressure trace and cyclic variation during the combustion cycle. The inputs to the model are the in-cylinder conditions at intake valve closing and the fuelling settings. The proposed model is based on experimental data, Principal Component Analysis and Gaussian Process Regression. This new data-driven approach is applied to model the combustion behaviour of a Reactivity Controlled Compression Ignition engine running on Diesel and E85. The resulting model has a root-square-mean-error of average behaviour and cyclic variance of 0.8° and 0.2°2 in CA50, 0.1bar and 0.03 bar2 in Gross Indicated Mean Effective Pressure, and 0.1% and 0.001 %2 in the Gross Indicated Efficiency, respectively.