A Softly Switched Multiple Model Predictive Control of A Turbocharged Diesel Engine

The paper presents a Model Predictive Control (MPC) design for a Diesel engine air path system equipped with Variable Geometry Turbine (VGT), Exhaust Gas Recirculation (EGR) and a Variable Geometry Compressor (VGC). Starting from a validated nonlinear engine model, multiple linearized models were obtained at different operating conditions and a switched linear MPC controller was designed to deal with the nonlinearity of the system. However, such a hard switching controller can generate unexpected outcomes due to the abrupt changes when crossing the switching boundaries. In order to mitigate the issues related to switching-type controllers, a softly switched mechanism is proposed here, which uses a convex combination of the objective functions during the switching process. A suboptimal solution is then adopted to reduce the computational complexity, allowing one to derive the explicit solution of the finite horizon optimization problem. Finally, simulation studies are conducted, and the results confirm the effectiveness of the proposed methodologies.