Online Fault Detection of a Heavy Duty Diesel Engine with Model-Based Methods

A rapid reduction in CO2 emissions from agricultural machinery can only succeed if diesel engines are operated with fuels that have a low CO2 impact. In agriculture, first generation biogenic fuels such as natural rapeseed oil are particularly suitable. This fuels can be produced by the farmers themselves and used on their own farms. The short transportation distances and simple production process can save up to 91% of CO2 emissions. However, these fuels have not yet become established. Apart from economic reasons, the reliability required in agricultural machinery has often not been met. Especially at cold temperatures, the operation with natural rapeseed oil in the past led to problems in the fuel system of the machines. One way to increase machine reliability and customer acceptance is to have an internal combustion engine that allows conventional diesel to be blended with biogenic fuels. In order for the engine to adjust to the respective fuel mixture, fuel detection is necessary. For this purpose, models based on the oxygen sensors in the exhaust tract (Oxygen-Mixture Model) and the current consumption of the fuel pump (Fuel Pump Current Model) are created in the first part of this thesis. These models can be used to detect fuel mixtures between diesel and natural rapeseed oil as well as mixtures between diesel and rapeseed methyl ester. In addition, a new low-pressure fuel system is being developed which is designed to operate with the three fuels and significantly improves rapeseed oil operation at cold temperatures. In addition, extensive adjustments are made to the control and regulation functions in the engine control unit, so that in operation with fuel mixtures a similar behavior is achieved as in pure diesel operation. In the second part of this thesis, faults of the diesel engine in pure diesel operation are detected. This serves to further increase the reliability of agricultural machines. In principle, the models are also suitable for natural rapeseed oil and rapeseed methyl ester. However, since some important basic information is only available for diesel fuel, all tests are performed with fossil diesel. In total, three models are being developed with which the injected fuel mass can be calculated. With these models not only the loss of performance due to insufficient injection quantities can be detected, but also increased performance due to illegal chip tuning. Since chip tuning can lead to early damage to machines, a reliable chip tuning detection with regard to possible warranty claims is financially very important for the manufacturer. One of the three developed injection quantity models is the Rail Pressure Based Fuel Estimation Model where the injection quantity is calculated based on the rail pressure or density changes of the fuel in the rail. Another injection quantity model is the Suction Control Valve Model. Here the injected fuel mass is calculated based on the metering unit of the high pressure pump. The third model is the Oxygen-Fuel model. It is completely independent of faults in the fuel system as it calculates the injected fuel mass based on residual oxygen content in the exhaust gas and the intake air mass flow. Another problem that affects the reliability of the diesel engine are injector deposits. In order to detect these deposits on the machine, an Injector Deposit Detection Model is presented. With this model it is possible to classify injectors and derive repair strategies. The detection of injector deposits is realized by multiple sampling of the rail pressure signal. Altogether, this research work shows that the operation of biogenic fuels is also possible with modern diesel engines. However, it becomes clear that a high parameterize effort for the engine control unit would be necessary to get an engine (which is suitable for different fuel mixtures) ready for series production. Furthermore, the results of the fault detection models show potential to further increase the reliability of diesel engines