Model-based estimation and fault diagnosis in diesel engine applications.

摘要

Diesel engines are used in a variety of applications from small sedan diesel cars to heavy duty applications in huge generators and hydraulic pumps. Various types of studies have been carried out on engine control and fault diagnosis and design. Presented in this work are two parts. The first part is on biodiesel blend estimation and the second one is on cylinder fault diagnosis.;Compared to the conventional diesel fuel, biodiesel, as a renewable alternative fuel, produces lower exhaust emissions with the exception of nitrogen oxides (NOx). Reducing nitrogen oxides emitted from engines miming on biodiesel requires proper engine controller adaptations that are linked to the specifics of the fuel blend. Therefore, online estimation of fuel blend is a critical step in allowing diesel engines to maintain performance while simultaneously meeting emission requirements when operating on biodiesel blends. Presented in the first part of this work are three different model-based biodiesel blend estimation strategies using: (i) crankshaft torsionals, (ii) Nitrogen oxides emissions measurement from the exhaust stream, and (iii) oxygen content measurement of the exhaust stream using a wide-band oxygen sensor. Each approach is investigated in terms of the accuracy and robustness to sensor errors. A sensitivity analysis is conducted for each method to quantify robustness of the proposed fuel blend estimation methods.;Modern diesel engines are equipped with high pressure injection systems which makes them powerful and efficient. The control of these high pressure systems is difficult due to discrepancies in the injectors. The fuel quantity must be accurately controlled to prevent any difference in the torque production of any individual cylinder in one engine cycle. This cylinder to cylinder variation of torque production causes undesired crankshaft torsionals vibrations. The second part of this work proposes two cylinder imbalance fault detection methods based on crankshaft rotational speed filtering, frequency analysis and fuel injection analysis. The first method utilizes the low-frequency signals extracted from the instantaneous engine speed to detect the faulty cylinder. The second method utilizes fueling information analysis to determine multiple faults. Experimental results validate the developed concept that utilized engine simulations.