基于BP神经网络的汽油机NO_(x)、CO和HC排放预测模型及试验研究  

Prediction Model and Experimental Study of NO_(x),CO,and HC Emissions from Gasoline Engines Based on BP Neural Network

作　　者：车金涛 范卓颖 陈铭世 Che Jintao;Fan Zhuoying;Chen Mingshi(CATARC Automotive Test Center(Guangzhou)Co.,Ltd.,Guangzhou,China)

机构地区：[1]中汽研汽车检验中心(广州)有限公司,广东广州

出　　处：《科学技术创新》2025年第7期213-216,共4页Scientific and Technological Innovation

摘　　要：构建了基于BP神经网络的汽油机NO_(x)、CO和HC排放预测模型,并设计试验验证了该模型的排放预测效果。该模型以转速、负荷、EGR率三类数据作为输入量,以NO_(x)、CO和HC的排放作为输出量。首先使用训练样本数据进行模型训练,确定模型的神经元数量、选择传递函数并进行归一化处理,得到最终的BP神经网络模型。然后将测试样本数据输入到该模型后,得到NO_(x)、CO和HC的排放预测值。从试验结果来看,排放预测值与试验值的误差控制在10%以内,说明基于BP神经网络的汽油机NO_(x)、CO和HC排放预测模型的预测精度较高。A gasoline engine NO_(x),CO,and HC emission prediction model based on BP neural network was constructed,and experiments were designed to verify the emission prediction performance of the model.This model takes three types of data:speed,load,and EGR rate as inputs,and NO_(x),CO,and HC emissions as outputs.Firstly,the model is trained using training sample data to determine the number of neurons,select the transfer function,and normalize it to obtain the final BP neural network model.Then,the test sample data is input into the model to obtain predicted emissions of NO_(x),CO,and HC.From the experimental results,it can be seen that the error between the predicted emission values and the experimental values is controlled within 10%,indicating that the prediction accuracy of the gasoline engine NO_(x),CO,and HC emission prediction model based on BP neural network is relatively high.

关 键 词：BP神经网络 排放预测模型 传递函数 归一化处理 

分 类 号：U464.171[机械工程—车辆工程] TK411.5[交通运输工程—载运工具运用工程]