This content is not included in your SAE MOBILUS subscription, or you are not logged in.
Intake O2 Concentration Estimation in a Turbocharged Diesel Engine through NOE
ISSN: 0148-7191, e-ISSN: 2688-3627
To be published on September 27, 2020 by SAE International in United States
Diesel engines with their embedded control systems are becoming more and more complex as the emission regulations tighten, especially concerning NOx pollutants. The combustion and emission formation processes in diesel engines are closely correlated to the intake manifold O2 concentration. Consequently, the performance of the main engine controllers can be improved significantly, if a model-based or sensor-based estimation of the intake O2 concentration is available in the ECU. The paper addresses the modeling of the intake manifold O2 concentration in a turbocharged diesel engine. Dynamic models, compared to generally employed steady state maps, capture the dynamic effects occurring over transients. It is right in the transient that the major deviations from the stationary maps are found. The dynamic model will positively affect the control system making it more effective. Furthermore, models can exploit information coming from sensors providing thus a more robust prediction performance. As a first result a Nonlinear Output Error model (NOE), with simulation focus, fed with four inputs is presented. The nonlinear function is a neural network. The inputs are engine BMEP, engine RPM and the position of EGR and VGT valves. This kind of model is characterized by an intrinsic feedback of the previous output samples. Two distinct datasets are used for training and validation of the NOE model. These sets of data are generated using GT-Power implementing a fine model of the engine previously validated on experimental measurements taken on the real engine. The NOE training procedure is performed in MATLAB environment through NNSYSID toolbox. In addition to the transient validation the chosen NOE model was tested against GT-Power outputs on step tests involving the two actuators of EGR and VGT. At last the network output is compared with an O2 steady state map over a transient in normal and faulty conditions. The performance of the model is satisfactory in both conditions. As a second experiment the potential benefits of installing an O2 sensor in the intake manifold is presented. In this situation a Nonlinear Auto-Regressive with eXogenous input (NARX) model is considered and compared to the previously investigated NOE. The results prove that exploiting through feedback of the past values of the output coming from O2 sensor significantly improves the model prediction capability.