基于广义回归神经网络的发动机排放预测  

Forecast of Engine Emissions Characteristic Based on Generalized Regression Neural Network

作　　者：俞明[1] 林冬燕[1] 孙国斌[1] 黄榕清[1] 

机构地区：[1]华南理工大学交通学院,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2003年第1期51-56,共6页Journal of South China University of Technology(Natural Science Edition)

摘　　要：神经网络是当前最主要的智能控制技术之一 ,它模拟人脑的结构及对信息的记忆和处理功能 ,具有擅长从输入输出数据中学习有用的知识的特性 .发动机性能预测是根据发动机结构参数和运转参数来估算推测发动机的各种性能指标 ,因此可以利用神经网络的学习性的特点 ,借助神经网络将各种影响汽油机燃烧过程的主要参数对汽油机的非线性影响以网络模型的形式表示出来 .文中讨论了如何抛开数学建模的方式 ,选用广义回归神经网络 ,进行发动机排放特性的预测 .应用MATLAB软件工具箱编程 。The Neural Network is one of the most important artifical intelligence technology at present. It simulates the structure of human brain and its function of dealing with or memorizing the information. The Neural Network is adept at learning from the input and output data and draw out an useful information. The predicting of the performance of engine is to predict the engine performance using the structure and operation parameters of the engine. So we can use the learning peculiarity of Neural Network to show how the structure and operation parameters of the engine affect the performance of engine, and describe them with a Network model. The authors of this article discusse how to choose the suitable network and uses it to construct the model of engine instead of physical model, then give an example of an engine model of emissions, and programs on MATLAB.

关 键 词：广义回归神经网络 发动机 排放预测 预测模型 智能控制 汽油机 排放控制 

分 类 号：TK411.5[动力工程及工程热物理—动力机械及工程] TP183[自动化与计算机技术—控制理论与控制工程]