Abstract
The continuously tightening emission legislations all over the world require more and more complex engine control and exhaust gas after-treatment systems. This confronts current engine development with the need for ever more complex engine systems with a continuously increasing amount of parameters that have to be calibrated and controlled in continuously decreasing time frames. In order to keep the development costs within reasonable limits, the use of model-based approaches is currently widely spreadening. Due to their high potential for calibration, the role of statistical models is also gaining importance. The combination of static statistical models trained by measurement data gained by the classical DoE process has shown great success in recent years and is now state of the art. However, the ability of static models to predict the exhaust gas emissions during the start and the warm-up phase, which are currently coming into the focus of exhaust gas regulations, is still very limited. In this work, we present an approach, which achieves good accuracy for the prediction of the exhaust gases in these crucial operating states, by combining dynamical statistical models with a dynamical DoE for the test bed measurements.
