DESIGN & SIMULATION OF THE ENGINE OUT VIRTUAL NOx AND SOOT SENSOR FOR DIESEL ENGINES

The automotive field has been into a transitioning phase since stringent emission norms have been imposed by various authorities all over the world. To comply with these regulations, automotive manufacturers are coming up with new technologies and components. On the other side, there are certain issues related to the performance of various sensors used to measure engine-out emissions. Novel concepts related to combustion like the Miller cycle and low-temperature high exhaust gas recirculation (EGR) fraction are highly tested on new vehicles. These new concepts heavily depend on in-cylinder parameters like chemical

composition, temperature, pressure, and flame characteristics. This leads to a complex and non-transparent engine control systems as diesel combustion itself is a complex phenomenon.

This research work aspires to establish a physics-based control-oriented diesel engine combustion model to estimate in-cylinder states like a mass burnt fraction (MFB), rate of heat release (ROHR), and cylinder pressure traces based on crank angle degree (CAD). Further, comprehensive prediction models are to be designed for NOx and soot based on these states. Additionally, the impact of exhaust gas recirculation (EGR) and turbo fractions on mass exchange in the cylinder during combustion also needs to be covered to make the model more realistic. A chemical kinetics model for diesel combustion based on reactants

and products involved in the combustion chemistry needs to be developed to determine the concentrations of products formed during the constant pressure and constant volume

adiabatic process. This analysis is focused on the prediction of nitric oxides (NOx) and soot based on the Extended Zeldovich mechanism and Hiroyasu Kadota approach respectively.

EGR significantly reduces the post flame NOx formation by introducing burnt and unburnt fractions from the exhaust gases. These gases reduce the oxygen concentration during

the combustion and ultimately reducing the flame temperatures. An appropriate control strategy is to be developed to control the EGR fractions to maintain the NOx levels within the legislative limits. Additionally, fuel consumption and turbo control are needed for the optimized NOx control and fuel economy without affecting the engine performance.

The proposed soot prediction model is established based on component level approach using MATLAB-Simulink as a programing tool. Some of the different subsystems involved in this model are – properties of injector nozzle, rate of heat release, in-cylinder temperature and pressure history, injection velocity of fuel jet and rate of fuel flow during the engine cycle.

The conventional approach of state space design has been adopted to derive mathematical models for both NOx and soot models. As NOx model has multiple input and outputs, it is further simplified and linearized to obtain SISO system. A state feedback controller is designed for NOx model and PI controller for soot model based on the system requirements.

Model validation is to be done by comparing the results to the high-fidelity GT-Power model controlled with the appropriate controller designed in the work. This GT-Power model has been developed for considering Cummins 6.7L diesel engine as a benchmark and results have been obtained for the same. The conceptual results prove the approach selected for modeling is correct as they agree with the theory behind it.

The obtained results from both uncontrolled predictive model and controlled model have been compared with GT-Suite reference data for both NOx and soot in steady state and

transient test cases. The controlled response shows good response with reference data with an accuracy around 2% and shows root mean square errors within acceptable limits below 0.5.

The developed model for both NOx and soot integrated with control system approach has significant advantage over the reference model developed in GT-Suite. The time taken

by GT-Suite model is significantly higher and thus are not suitable for real time application. On the contrary, the proposed approach uses a real time prediction of engine out emission and can readily be used in vehicles with minor modifications.