NEURAL NETWORK-BASED MODELING AND OPTIMIZATION FOR EFFECTIVE VEHICLE EMISSION TESTING AND ENGINE CALIBRATION.

In automotive manufacturing, vehicle emission testing and engine calibration are the key to achieving emission standards with satisfactory fuel economy. Because of the complexity of physical and chemical phenomena occurring during engine combustion and catalytic conversion and the lack of real-time measurements of key process and performance parameters, engine calibration and emission testing are still experiment-assisted trial-and-error practices, which are always expensive and inefficient. In this article, a neural network (NN)-based modeling approach is introduced to characterize engine and catalytic converter operations. A model-based optimization method is also introduced to identify optimal engine calibration parameters so that emission reduction and fuel efficiency improvement can be achieved simultaneously. This development facilitates a comprehensive performance analysis of engine and catalytic converters with much less effort required for experiments. [ABSTRACT FROM AUTHOR]