Seri bağlı aşırı doldurma sistemine sahip dizel motorlar için akıllı manifold basıncı kontrolcüsü tasarımı

As a subsystem of automotive control field, engine control has evolved as one of the key research areas in control applications. Due to having highly nonlinear behaviour, relatively fast system dynamics and variety of constraints (emissions, Hardware limitations, performance requirements), engine controls softwares includes vast amounts calibrateable map and curves. Current methods in engine controls are mainly employing conventional PID-PI schemes with additional map based feedforward compensation with anti windup controllers. With the increasing complexity of the engine control systems resulting from stringent emissions and customer performance requirements, control problem is becoming multi-dimensional and multi-variable. Moreover, as the system complexity increases, hardware limitations and constraints are increased proportionally. Today, state of the art diesel engine airpath control is basically consists of two controller loops. First loop is controlling the amount of total air for EGR operation and the second loop is controlling the Boost pressure for performance requirement and efficient combustion. Main actuators for MAF control are intake throttle and low/high pressure EGR valve where for Boost pressure control Variable Nozzle Turbocharger(VNT) and Wastegate valves. The term two-stage turbocharging described by system which involves 2 turbocharger units connected in series. This layout has several advantages. • Higher boost pressure levels resulting in high mean effective pressures. • Increased charging efficiency • Higher engine out temperatures • Better low-end torque Below are the drawbacks of two-stage turbocharged system. • Increased turbolag as two rotors needs to be accelerated • Higher thermal inertia • Bigger hardware package Control and regulation of two stage system increases the effectiveness of turbocharger operation and results in wider flow range. The Boost level can be controlled via regulation of exhaust energy utilized by two stage turbines, using actuators such as a Turbine Bypass Valve (TBV), wastegate (WG) or VNT. Two stage turbocharger system sometimes called as series sequential turbocharger system. As being one of the key elements of diesel engine airpath as well as multivariable control capability makes the two-stage turbocharged diesel engine control a challenging control application. Standard industrial boost pressure controllers for two-stage turbocharged diesel engine control employs gain-scheduled PI-PID controllers with boost pressure feedback. There are more than 500 parameters to be tuned which requires high level of engine dynamometer and vehicle testing. In addition, over-boost and under boost capability of the system should be considered seperately by calibrating rate limitation or by means changing boost pressure setpoints. In this thesis the main focus is to design an intelligent boost pressure controller which would reduce the time required to calibrate controllers as well as incorporating system pressure limits explicitly. Model predictive control stands as a candidate controller algorithm in terms of constrained optimal control formulation enables to handle input, output and state constraints via a online computation of optimal input trajectories. Big drawback of this method is the computational effort required. Explicit MPC method is based on the replacement of the online optimization with offline calculations with apriori knowloedge about system states, input and outputs. Quadratic cost function in optimal control formulation can be analyzed as a multi-parametric quadratic programme. Modeling of the two-stage turbocharger system is a key milestone to achieve satisfactory control design. During this study, two-stage turbocharged system has been modeled in terms of nonlinear and linear models. Nonlinear model is used to simulate controller performance offline. Nonlinear model has been chosen as 1D model which incorporates 1-dimensional fluid flow across engine ducts. 1D engine model has been modeled in WAVE and then converted into MATLAB-SIMULINK environment by means of WAVE RT module. 1D engine offers several advantages over mean value engine models in terms of reduced parametrization, therefore data requirements. Linear model has been used to model system linear response to be used as a prediction model in MPC formulation. As nonlinear model includes complex fluid dynamics compared to mean value engine models, linearization would not be feasible. Therefore parameter identification methods has been used to derive linear models. It would not be feasible to define one local model for entire engine operating region, therefore engine operating region has been divided into 7 main sub-regions which have individual local linear models. The input-output data has been recorded during real-time engine dynamometer testing and post-processed in MATLAB-SIMULINK. The PEM and N4SID methods offerred best fits and been used a black-box models. Controller design has been performed seperately for each of the local linear regions. Controller design has been performed by introducing Kalman filter, Prediction-Control horizon tuning as well as constraint and weighting definitions. Explicit MPC formulation has been calculated by using MATLAB Hybrid Toolbox offline. Two controllers has been designed. SISO controller employs TBV only control action in which WG is always set as zero. Whereas MIMO controller utilizes both TBV and WG. Controller objective are defines based on the experience in industrial boost pressure controller sign off criteria. Offline simulations are performed in MATLAB-SIMULINK environment with the WAVE RT nonlinear model. 2 simulation sets are defined with mixed transient cycles and simulations at fixed engine speed with small-medium and large boost pressure setpoint applied. Simulation results shows that overall response of the system satisfied controller objective in terms of both overshoot and rise time. In addition MIMO and SISO controller shows close performance where in MIMO controller optimal control action is mostly calculated as fully closed. As controllers are validated with offline simulations by using high fidelity nonlinear engine model, final validation has been performed by a series of in-vehicle and engine dynamometer testing. 2 different methods have been used for controller implementation. In vehicle testing has both been performed random testing on road for subjective evaluation as well as testing on chassis dynamometer. Test results shows that controller setpoint tracking performance is satisfactory. During the testing, compressor and turbine operation is observed to be inside hardware limits. Chassis dynamometer testing includes the replication of customer cycles and NEDC to compare conventional PID based controller and MPC performance. MPC performance shows close at some points even better against PIDs. Statistical analysis of boost deviation is also smaller in MPC operation compared to PIDs. Engine dynamometer testing consists of step response tests a fixed engine speed where load is changing. MPC performance satisfies the controller objectives in terms of rise time and overshoot. During this study MPC proves to be succesfull candidate to replace conventional PID in terms of multivariable control capability and constraint handling. On the other hand it should be noted that MPC performance is strongly affected by prediction model accuracy. Further study, shall contain the inclusion of mean value 0D engine model with compressor speed estimation as well as effect of EGR operation.