基于RBF神经网络的汽油机电控参数的标定及优化  

Calibration and Optimization for Electronically Controlled Parameters of Gasoline Engine Based on the RBF Neural Network

作　　者：侯献军[1] 巩学军[1] 杜常清[1] 颜伏伍[1] 彭辅明[1] 

机构地区：[1]武汉理工大学,武汉430070

出　　处：《中国机械工程》2009年第18期2264-2267,共4页China Mechanical Engineering

基　　金：国家863高技术研究发展计划资助项目(2006AA060307)

摘　　要：为了寻求电控单元与发动机的最佳匹配,通过分析汽油机电控系统控制参数,建立了汽油机稳态性能预测的径向基函数(RBF)人工神经网络模型。通过LJ276M汽油机台架标定试验获取样本数据,利用训练过的RBF神经网络预测汽油机在其他稳态工况点的电控参数并检验所建立神经网络模型的性能。喷油脉宽和点火提前角的网络输出最大误差小于1%,平均误差小于0.6%。研究结果表明,该预测模型具有较强的泛化能力,能够准确地预测发动机电控参数。In oder to seek the better match of ECU（electronic control unit） and engine,the gasoline engine steady--state performance prediction model of RBF artificial neural network was build through the author＇s analysis of the electronic control parameters of gasoline engine. Sampling data of LJ276M gasoline engine in calibration test bench was obtained, and then the electronic control parameters of other steady--state conditions point in use of the trained RBF artificial neural network were forecasted and the establishment of the neural network model performance was validated. The maximum network output error of fuel injection pulse width and ignition advance angle is less than 1%, with an average error of less than 0.6%. The result shows that it is feasible to predict the control parameters of gasoline in use of the RBF neural network. Besides, this model has very strong generalization ability to be able to predict the control parameters accurately.

关 键 词：RBF神经网络 喷油脉宽 点火提前角 标定 训练 检验 

分 类 号：TK417[动力工程及工程热物理—动力机械及工程]