In this paper, the computed results to the Air-Fuel Ratio (A/F) using three chemical equations are compared with the test results. With these four methods and the engine combustion product components, the A/F is deducted, and the results are different. Using neural networks (NN) mapping ability, the relation is found between A/F and combustion products components. Sample is fabricated and learned using those known information as input and output of NN. A/F can be mapping with combustion products components for engine by generalization of interpolation and extrapolation; prime information feature is abstracted based compared and verified with multi means. A/F can be well mapping in all work conditions with combustion products components. Each means restrict is solved and deduced precision is heightened. Through analyzing exhaust composition of engine with certain working condition, it shows that the maximal mapping error is 0.58%, which can satisfy measuring requirement of accurate control of present automobile to A/F. In addition, mapping of neural network to A/F has the characteristic of intelligence and flexibility. After learning and training of sample, using the extrapolation capability of network, the more accurate A/F can be mapping by mean of the engine emission information.
Application of the Neural Network Mapping Air-Fuel Ratio with Combustion Products Components
Sae Technical Papers
2008 SAE International Powertrains, Fuels and Lubricants Congress ; 2008
2008-06-23
Conference paper
English
Application of the Neural Network Mapping Air-Fuel Ratio with Combustion Products Components
British Library Conference Proceedings | 2008
|Application of the neural network mapping air-fuel ratio with combustion products components
Automotive engineering | 2008
|Study on the Relation Between Fuel-Air Ratio and Combustion Products
SAE Technical Papers | 2002
|Study on the relation between fuel-air ratio and combustion products
Automotive engineering | 2002
|Air/fuel ratio control for direct injection combustion engines using neural networks
Automotive engineering | 1997
|