以发动机转速、进气量、循环油量、发动机出水温度、中冷后进气温度、进气湿度、排气背压、柴油温度作为输入，NOx排放质量流量为输出，优化隐层节点和迭代次数，并经过样本训练，构建了NOx排放预测模型。结合台架试验数据，验证了模型的泛化能力，其预测值与试验值间误差小于1.5%。在此基础上，利用模型进一步分析了试验因素的重要度和试验控制性。结果表明：发动机转速、循环油耗、中冷后进气温度、排气背压对柴油机NOx排放的影响相对较高；进气湿度控制范围过宽，对NOx排放测试结果影响高于其他因素。
冀树德
,
高华伟
,
邬旭宏
,
刘志刚
,
张伟
,
郝冀雁
,
陈东峰
,
李全
,
梁玉明
. 基于多因素神经网络模型的柴油机NOx排放预测及试验研究[J]. 车用发动机, 2018
, 0(2)
: 41
-45
.
DOI: 10.3969/j.issn.1001-2222.2018.02.007
Taking the engine speed, intake air mass, cycle injection mass, cooling water temperature, intercooled air temperature, intake humidity, exhaust back pressure, diesel temperature and the NOx mass flow as the input and output parameters respectively, the NOx emission prediction model was built through optimizing hidden layer node number and iteration times and training test samples. The generalization ability of model was verified by testing the bench test data. It was found that the error between test and prediction value was less than 1.5%. The importance and control characteristics of test factor were further analyzed with the model. The results show that the engine speed, cycle injection mass, intercooling air temperature and exhaust back pressure have the greater influences on the NOx emission. In addition, the influence of intake humidity on NOx emission is higher than any other factors because of its wide range.
