Nonlinear Model Predictive Control of a Diesel Engine Airpath

In this paper we show the Nonlinear Model Predictive Control (NMPC) of an airpath of a diesel engine based on a Linear Parameter Varying (LPV) model. We used databased LPV modelling with real data from a dynamical engine test bench in order to obtain a nonlinear model of high quality. Because of the nonlinearity of the model the quadratic program (QP) of the NMPC needs to be set up afresh at each sampling instant, which is the main difference to standard linear MPC. For solving the QP efficiently, we employ the recently developed online active set as implemented in the software package qpOASES. We tested our controller in simulation on the LPV model identified on a mean value model and results show that the NMPC has a better tracking performance in terms of boost pressure and fresh air mass flow compared to the standard linear MPC approach also under the in uence of a model plant mismatch.