Ion Current and Torque Modeling for Combustion Engine Control

Engine control development constantly requires more information about the combustion process, to meet present and future environmental regulations for vehicles. The demand for combustion monitoring and control from engine developers drives the research in several parallel fields of technology to find the best and most cost effective sensor solution for each case. The ionization sensor and the crankshaft torque sensor are two sensor candidates which are investigated in this work. Their hardware is relatively cheap and robust but the development of methods for extracting important information content is subject to research. This thesis presents models that connects combustion properties of a spark ignited engine to the measured ionization and torque signals. The two sensor technologies are treated separately, but both with the same overall goal of solving the more interesting inverse problem, which is to derive combustion information from the measurements. The ionization current model consists of several sub-models which enables the estimation of cylinder pressure, combustion temperature and nitric oxide formation, based on the measured ionization current. One strength of the model is that after calibration, it has only two free parameters, burn angle and initial kernel temperature. Experimental validation of the model is done on the pressure estimation part, which shows good accuracy. The thesis develops a torque signal model that includes a strategy to linearize the crankshaft dynamics. A linear inverse algorithm is presented that separates the individual cylinder torque contributions from the measured signal. A new method, called the \emph{torque ratio concept}, is presented to provide means to estimate combustion properties from individual cylinder torque. The method contains a parameterized combustion model that captures burn duration and a measure for combustion phasing. The torque signal model and the torque ratio concept together provide a scheme for estimating individual cylinder combustion phasing, based on torque measurements. The algorithms were implemented on a real-time platform for engine control. Closed-loop control of combustion phasing is demonstrated under the influence of unknown disturbances, e.g. varying fraction of the residual gas