钟祥麟¹，于全顺¹，高忠明¹，陈旭²

1.中国汽车技术研究中心有限公司，天津  300300；2.山东大学能源与动力工程学院，山东 济南  250061

摘要：为实现柴油机排放预测，利用试验设计（design of experiment，DoE）试验获取满足柴油机工作范围的稳态试验数据，利用反向传播（back propagation，BP）神经网络搭建柴油机气缸模型，使用GT-Power软件搭建柴油机的进排气系统模型，将两者耦合搭建柴油机整机模型，并通过试验验证模型在稳态及瞬态工况下的预测精度。结果表明：稳态工况下模型的NOx排放预测相对误差为4.1%，瞬态循环工况下模型的NO_x排放预测相对误差为1.2%；该模型可以较准确地预测柴油机的排放。

关键词：柴油机；BP神经网络；GT-Power；NO_x排放预测；瞬态循环

Abstract：In order to realize the emission prediction of diesel engine, the design of experiment (DoE) test is used to obtain the steady-state test data meeting the working range of diesel engine, the cylinder model of diesel engine is built through back propagation (BP) neural network, the intake and exhaust system model of diesel engine is built by GT-Power software, and the two are coupled to build the whole diesel engine model. The prediction accuracy of the model under steady-state and transient conditions is verified by experiments. The results show that the NO_x emission prediction error of the model is 4.1% under steady-state conditions and 1.2% under transient cycle conditions. The model can accurately predict the emission of diesel engine.

Keywords:diesel engine;BP neural network;NO_xemission prediction;transient cycle