Forward and inverse fuzzy based model predictive control of a turbocharged diesel engine with high-pressure exhaust gas recirculation system

The main purpose of this study is to reduce the number of iterations and reduce the optimization duration by taking advantage of the initialization of the optimization solver in an appropriate point. The model to be used for estimation in the controller is a fuzzy model and it is non-linear. Fuzzy modeling is a widely used method for modeling high non-linear systems. Since the engine is an example of high non-linear systems, the fuzzy model is employed in order to capture the dynamics of the engine in all operation range. The fuzzy model is based on first order regression method, which takes one sample delayed outputs as inputs in order to be a dynamical model. An inverse fuzzy model is employed to initialize the optimization solver in an appropriate point and it is based on second order regression method which takes one and two sample delayed outputs as inputs. The solver in a classical MPC uses the last manipulated variable values as the initial points in the current time step whereas the initialization is based on the inverse fuzzy model in this study. If the optimization is initialized with a proper point, then the optimization may be completed before restricted by the limits or find a better solution while the limits are active. The optimization algorithm that finds the optimal solution is the 'fmincon' function which is available in Matlab environment. The proposed MPC, standard MPC and PID controller performances in terms of the weak and superior aspects are compared. In addition, studies on fuzzy modeling are discussed. It is aimed to increase the knowledge and know-how in this area by applying an advanced control method to a difficult control problem.