Estimation of fuel properties in a common rail injection system by unscented kaiman filtering

The online estimation of fuel property parameters (density and bulk modulus) for use in automotive engine control is considered. The estimation is carried out through state augmentation (including the parameters in the state vector) and use of an Unscented Kalman Filter (UKF) that is based on a physical model of the common rail fuel injection system. As it is known that state augmentation usually leads to biased estimates and strongly depends on the filter tuning parameters, the method is first tested with data from a simulation model. Through this, it was found that it is possible to estimate the parameters with negligible bias and that the method is generally suitable. The method is then tested on experimental data from a fuel injection system test rig that was specifically constructed for this purpose. Several modifications of the physical model are considered to improve the accuracy of the estimation.