Application study of neural networks in engine modeling

CHEN An-cheng1， MU Xi-hui2， DU Feng-po2， GUO Hao-liang1
（1. Department of Mechano-Electronic， Ordnance Engineering College， Shijiazhuang 050003， China；
2. Research Institute of Ordnance Technology， General Armament Department， Shijiazhuang 050003， China）

Abstract： In order to solve the accuracy problems of engine modeling and curve fitting，the BP neural network was introduced which applied to engine modeling and curve fitting. The neural networks method and the least square method were used respectively to simulate and practice the external characteristic and universal characteristics of engine. On the basis of Matlab/Simulink modeling simulation，the curve fitting precision and error were evaluated. The simulation results show that the neural networks method has higher curve fitting precision and simplifies the calculation，which means a lot to the fuel economy and dynamic property of the vehicle. Further more，it affords the theory base to find the best matching strategy between engine and the transmission system of vehicle.
Key words： engine modeling； neural networks； curve fitting