基于BP神经网络预测正庚烷-乙醇混合燃料自燃温度  

Prediction on Auto-ignition Temperature of N-heptane and Ethanol Blended Fuel Based on BP Neural Network

作　　者：陈若龙 韩永强[1] 李润钊 张一鸣[1] 安东 孙博 

机构地区：[1]吉林大学汽车仿真与控制国家重点实验室,吉林长春130022 [2]长春一汽四环发动机制造有限公司,吉林长春130013

出　　处：《车用发动机》2018年第1期10-15,共6页Vehicle Engine

基　　金：国家自然科学基金(51576089);研究生创新基金(2016026);吉林大学研究生创新基金资助项目(2017125)

摘　　要：正庚烷-乙醇混合燃料的自燃温度对研究反应控制压燃(RCCI)具有重要的参考意义。采用BP神经网络预测正庚烷-乙醇混合燃料自燃温度,该神经网络模型以正庚烷掺混比、当量比和进气压力为输入,自燃温度为输出,单层隐含层有16个节点时迭代过程均方误差和训练状态梯度均最小。研究结果表明:对神经网络模型训练、验证、测试的线性系数和全局线性系数R分别为0.997 78,0.997 9,0.994 92和0.997 33,预测精度较高;验证了该神经网络模型对正庚烷掺混比、当量比和进气压力变化的泛化能力,预测值与试验值的误差均在允许范围内,因此本模型得到的预测值与试验值具有良好的一致性。The auto-ignition temperature of n-heptane and ethanol blended fuel has an important reference significance for studying the reaction controlled compression ignition. The auto-ignition temperature of n-heptane and ethanol blended fuel was predicted by BP neural network,which used n-heptane mixing ratio9 equivalence ratio and inlet pressure as input and auto-igni-tion temperature as output. When there were 16 nodes in a single hidden layer 9 the mean square error in the iterative process and the training state gradient were the minimum. The results show that the prediction accuracy is higher when the training 9 verification,test and global linear coefficients R of neural network model are 0.997 789 0.997 9 9 0.994 92 and 0.997 33 respec-tively. The generalization ability of neural network model for the variation of n-heptane mixing ratio 9 equivalence ratio and inlet pressure is verified and the error between the predicted and experimental values is within the allowable range. Therefore,the predicted values obtained by this model are in good agreement with the experimental values.

关 键 词：BP神经网络 正庚烷 乙醇 混合燃料 自燃温度 预测 

分 类 号：TK464[动力工程及工程热物理—动力机械及工程]