TY - JOUR
T1 - Impact of Injection Valve Condition on Data-driven Prediction of Key Combustion Parameters Based on an Intelligent Diesel Fuel Injector for Large Engine Applications
AU - Laubichler, Christian
AU - Kiesling, Constantin
AU - Warter, Sven
AU - Kober, Martin
AU - Wimmer, Andreas
AU - Coppo, Marco
AU - Negri, Claudio
AU - Koegeler, Hans-Michael
AU - Kammerdiener, Thomas
PY - 2024/4/9
Y1 - 2024/4/9
N2 - The advent of digitalization opens up new avenues for advances in large internal combustion engine technology. Key engine components are becoming "intelligent" through advanced instrumentation and data analytics. By generating value-added data, they provide deeper insight into processes related to the components. An intelligent common rail diesel fuel injection valve for large engine applications in combination with machine learning allows reliable prediction of key combustion parameters such as maximum cylinder pressure, combustion phasing and indicated mean effective pressure. However, fault-related changes to the injection valve also have to be considered. Based on experiments on a medium-speed four-stroke single-cylinder research engine with a displacement of approximately 15.7 liter, this study investigates the extent to which the intelligent injection valve can improve the reliability of combustion parameter predictions in the presence of injection valve faults. Injector fault is considered through two artificially aged injection valve variants (clogged nozzle hole and clogged feeding hole on the orifice plate). A comprehensive database obtained using a design of experiments approach is used to build machine learning models for combustion parameter prediction with and without the value-added data from the intelligent fuel injection valve. The comparison of the results reveals the impact of the faulty injection valve variants on the prediction accuracy. It can be found that the intelligent injection valve can improve the data-driven prediction of key combustion parameters when injection valve faults occur.
AB - The advent of digitalization opens up new avenues for advances in large internal combustion engine technology. Key engine components are becoming "intelligent" through advanced instrumentation and data analytics. By generating value-added data, they provide deeper insight into processes related to the components. An intelligent common rail diesel fuel injection valve for large engine applications in combination with machine learning allows reliable prediction of key combustion parameters such as maximum cylinder pressure, combustion phasing and indicated mean effective pressure. However, fault-related changes to the injection valve also have to be considered. Based on experiments on a medium-speed four-stroke single-cylinder research engine with a displacement of approximately 15.7 liter, this study investigates the extent to which the intelligent injection valve can improve the reliability of combustion parameter predictions in the presence of injection valve faults. Injector fault is considered through two artificially aged injection valve variants (clogged nozzle hole and clogged feeding hole on the orifice plate). A comprehensive database obtained using a design of experiments approach is used to build machine learning models for combustion parameter prediction with and without the value-added data from the intelligent fuel injection valve. The comparison of the results reveals the impact of the faulty injection valve variants on the prediction accuracy. It can be found that the intelligent injection valve can improve the data-driven prediction of key combustion parameters when injection valve faults occur.
U2 - 10.4271/2024-01-2836
DO - 10.4271/2024-01-2836
M3 - Conference article
SN - 0148-7191
JO - SAE Technical Papers
JF - SAE Technical Papers
M1 - 2024-01-2836
T2 - 2024 WCX SAE World Congress Experience
Y2 - 16 April 2024 through 18 April 2024
ER -