Research On Performance And Emission Of Electronic Gasoline Engine Based On Grey Relational Analysis And Artificial Neural Network

In recent years, the government makes the increasing policy efforts on the "energy conservation and pollution reduction" for the every industry sector. For the transport sector, the mount of vehicles is increasing rapidly with the economic development, whereas the energy consumption and the environment pollution from the huge mount of vehicles also become more and more serious problem. So it is very important task to improve the engine performance and decrease its emission. This paper aim at improving the electronic control engine performance and emission, which is based on the experiments. And the main works are as follows:(1) Based on the platform test for the CF4g18 gasoline engine, the experiment data is acquired. And the grey relational analysis is used to determine the closeness of relationships between the emission pollutants and the various kinds of factors. Basis of that, to do the modeling task with the artificial neural network method (ANN).(2) In the modeling of the engine power performance, BP neural network is used. as the evaluations for the networks, the maximum absolute relative errors for the validation set of the performance parameters are follows respectively: 9.58% for the torque model, 9.36% for the power model ,4.42% for the specific fuel consumption model, 5.76% for the fuel consumption, 3.74% for the excess air coefficient and 6.97% for the exhaust temperature model, Which indicates BP neural network can satisfy the demands of the engine performance parameters modeling.(3) In the modeling of the engine emission performance, Elman neural network is used. as the evaluations for the networks, the maximum absolute relative errors for the validation set of the emission pollutants are as follows respectively: 8.7% for the CO emission model, 4.8% for the CO₂ emission model, 10.9% for the HC emission model and 9.53% for the NO emission model, which indicates Elman neural network can satisfy the demands of the engine emission performance modeling.(4) For the tendency of decreasing emission for gasoline engine with three-way catalytic converter, according to operating features of the three-way catalytic converter, the CF4g18 engine performances including the power performances and emission performances is predicted as result with the neural network models of engine, by adjusting the excess air coefficient for the CF4g18 engine.