Quick Look

Feasibility Investigation of the On-Board Fuel Property Classifier for Internal Combustion Engine Control Systems, 03-9478

Printer Friendly Version

Principal Investigators
Junmin Wang
Gary D. Neely
Thomas W. Ryan III

Inclusive Dates:  06/01/04 - 10/01/04

Background - As engine emission regulations become increasingly stringent, more and more sophisticated calibration- or model-based control strategies and algorithms are being applied to control the engine combustion process and the after-treatment systems to achieve the emission regulations and desired performance simultaneously.

As both the energy and the emission sources in internal combustion engines, fuel properties are vital. The fuel affects the in-cylinder combustion and the after-treatment systems. As the emission regulations become significantly more stringent, the fuel property variations become more and more important to the engine emissions for model-based and calibration-based engine control systems.

However, the properties of the fuel could vary greatly as a result of the crude source, refining processes, distribution, and storage methods as well as the additives. In addition, alternative and renewable fuels (such as ethanol derived from corn) and different fuel blends are being actively developed for engine and vehicle applications because of the tightening emission regulations and increasing global petroleum supply concern. The properties of these fuels can vary significantly as well. For different fuels, the engine in-cylinder combustion and characteristics of the after-treatment system will be different and the engine control systems should treat them with respect to the corresponding properties of the fuel the engine is running to obtain the optimal performance and emissions. It is hard to imagine that the engine control systems can perform well without knowing what type of fuel it is running.

Moreover, as the engine and vehicle makers incorporate after-treatment systems, such as the diesel particulate filter (DPF), lean NOx traps (LNT), and diesel oxidation catalyst (DOC), these after-treatment systems also need to be managed by the engine control systems to achieve the best performance. To optimize control of the after-treatment system, minimize the fuel penalty caused by the regeneration for these devices, and realize reliable fault diagnostic approaches for these devices, accurate engine-out emission estimation models are necessary. Those models, however, are dependent on the fuel properties. Thus, the fuel property information is a necessary input to the engine control system in order to achieve the optimal management and fault diagnostics of the after-treatment system as well as minimize the fuel penalty.

Approach - The on-board fuel property classification approach aims at classifying the properties of the fuel that the engine is running by using the engine input-output response characteristics that could be possibly measured from the standard engine sensors at certain operating conditions. Figure 1 shows the structure of the fuel property classifier. The engine operating conditions, the control inputs to the engine, and the corresponding engine outputs and responses measured from the standard engine and vehicle sensors are inputted to a neural network-based fuel property classifier. The classified fuel property information is then provided to the engine control system to select the optimal parameters/maps/strategies for engine models, in-cylinder combustion, and after-treatment system management and fault diagnostics correspondingly. Figure 2 shows the structure of the fuel property-adaptive engine control system. The fuel property classifier employs the signals measured from the engine standard sensors. No dedicated sensor is required for the classifier.

Accomplishments - The on-board neural network-based fuel property classifiers have been developed and verified for three different fuels and different engine operating conditions at two classification scenarios on a light-duty diesel engine. The results show that the classifiers can classify clearly and accurately the fuels that the engine is running by using the engine input-output response characteristics measured from the standard sensors equipped on the engine.

Some suitable engine operating conditions for the fuel property classifier to be active have been experimentally determined and compared from the three tested conditions. It turns out that the light load condition is the most suitable condition among the three for the classifier to be active. As a result, the classifier active range covers a wide range of the vehicle driving time because the light load condition is one of the most common conditions for the normal vehicle driving.

The fuel property classifier can greatly enhance the model- or calibration-based engine control system by making the engine control system to be fuel property-adaptive in the sense of optimal control for the engine in-cylinder combustion and the management of the after-treatment systems with respect to the properties of the fuel that the engine is running.

Figure 1. The structure of the fuel property classifier using engine operating condition, engine control inputs, and engine outputs. Figure 2. The fuel property-adaptive engine control system based on the fuel property information provided by the classifier for both the in-cylinder combustion and the exhaust treatment systems.

2004 Program Home