基于模型群预测法对汽油机稳态原排的预测  

Prediction of gasoline engine steady state exhaust based on model group prediction method

作　　者：陈涛 秦静[1,2] 赵华[1] 苏庆鹏[1,3] 吕永 钟凯 王膺博 裴毅强 CHEN Tao;QIN Jing;ZHAO Hua;SU Qing-peng;LYU Yong;ZHONG Kai;WANG Ying-bo;PEI Yi-qiang(State Key Laboratory of Engines,Tianjin University,Tianjin 300072,China;Internal Combustion Engine Research Institute,Tianjin University,Tianjin 300072,China;GAC Automotive Research&Development Center,Guangzhou 511434,China)

机构地区：[1]天津大学内燃机燃烧学国家重点实验室,天津300072 [2]天津大学内燃机研究所,天津300072 [3]广州汽车集团股份有限公司汽车工程研究院,广州511434

出　　处：《吉林大学学报（工学版）》2021年第5期1565-1574,共10页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金项目(51676136).

摘　　要：针对通过数值模拟减少汽油机台架试验环节的工作量、提高试验效率的问题,采用模型群预测法(优化后的人工神经网络方法)对汽油机台架试验过程中的NOx、CO、HC等稳态原排进行建模及预测分析。结果表明:与传统的单个模型预测方法相比较,模型群预测法具有较高的可靠性,能较好地提升预测结果的准确度。采用隔点取点法适当减少神经网络建模的训练数据集,仍能保持较好的预测能力,在项目开发过程中只需进行30%的测试量,将试验结果用于神经网络模型训练,可较好地预测剩余工况排放。通过对其他机型的验证分析,模型群预测法在内燃机稳态原排的预测过程中具有较好的普适性。This paper aims to reduce the workload of gasoline engine bench test and improve the experimental efficiency by numerical simulation technology.The model group prediction method(optimized artificial neural network method)was used to model and predict the steady-state emissions of NOx,CO and HC in the process of gasoline engine bench test.The results showed that the model group prediction method has high reliability and may improve the accuracy of prediction results compared with the traditional single model prediction method.The training data set of neural network modeling may be appropriately reduced by using the method of separating points,and the better prediction ability may still be maintained.In the process of project development,only 30%of the test quantities were needed,and the test results may be used for the training of neural network model,which may better predict the emission of the remaining working conditions.Through the validation analysis of other models,the model group prediction method had a good universality in the prediction process of engine steady-state exhaust.

关 键 词：汽油 神经网络 预测 发动机排放 台架试验 

分 类 号：TK411[动力工程及工程热物理—动力机械及工程]