一种基于逻辑回归的柴油机工况分类模型  

A Classification Model of Diesel Engine Operating Conditions Based on Logistics Regression

作　　者：吉黎明[1] 熊兴旺 杨子荣 JI Liming;XIONG Xingwang;YANG Zirong(China National Accreditation Service for Conformity Assessment,Beijing 100062,China;CATARC New Energy Vehicle Test Center(Tianjin)Co.,Ltd.)

机构地区：[1]中国合格评定国家认可中心,北京100062 [2]中汽研新能源汽车检验中心(天津)有限公司

出　　处：《小型内燃机与车辆技术》2023年第2期6-9,20,共5页Small Internal Combustion Engine and Vehicle Technique

摘　　要：利用一台国六柴油机进行热态WHTC循环试验,对循环中所有工况点,依据排气温度值和原始排气NOx体积浓度值,进行二值化状态标识(0或1)。以发动机转速、进气空气流量、燃油流量,功率和转矩作为输入变量,工况点二值化后的状态标识变量作为模型输出,构造逻辑回归分类模型。将循环所有工况点按照一定比例随机划分为训练工况数据点集与测试工况数据点集,训练逻辑回归分类模型。模型预测结果表明,测试工况数据点集对应的模型准确率和模型对标识为1的工况的召回率均大于83%。Hot-started WHTC cycle tests were carried out on a ChinaⅥstage diesel engine.According to the exhaust temperature and NOx volume concentration of the exhaust gas without purification,all the operating conditions were marked with the binary state identification(0 or 1).A logistic regression classification model is built.The engine speed,intake air flow,fuel flow,power and torque are selected as the input variables,and the binary state identification of operating point is selected as the output of the model.According to a certain proportion,all operating conditions of the cycle are randomly divided into the training data sets and the testing data sets.The logistic regression classification model is trained.The prediction results of the classification model show that the accuracy of the model corresponding to the testing data sets and the recall rate of the model for the operating conditions identified as 1 are all larger than 83%.

关 键 词：工况 逻辑回归 阈值 准确率 召回率 

分 类 号：TK421.5[动力工程及工程热物理—动力机械及工程]