基于动态递归神经网络的HCCI发动机燃烧相位辨识模型

被引:5
作者
谢辉 [1 ]
孙艳辉 [2 ]
夏超英 [2 ]
机构
[1] 天津大学内燃机燃烧学国家重点实验室
[2] 天津大学自动化学院
关键词
HCCI汽油机; 燃烧相位观测; 动态递归神经网络;
D O I
10.16236/j.cnki.nrjxb.2007.04.006
中图分类号
TK411 [理论];
学科分类号
摘要
为了实现HCCI汽油机闭环反馈控制,提出了一种利用动态递归神经网络从气缸压力信号在线辨识燃烧相位CA50(燃烧50%累积放热量的曲轴转角)的方法。该方法采集上止点附近40°CA范围的气缸压力信号,经过归一化和主元素法降维处理后,得到一个由9个特征数构成的时间序列。一个Elman动态递归神经网络以该序列为输入,计算出燃烧相位CA50。以基于全可变气门机构的汽油HCCI发动机为对象,选取了台架试验中4个典型的HCCI动态变负荷过程数据,其中一个作为训练样本,另外3个作为测试样本。测试结果表明:该方法对HCCI动态过程的燃烧相位CA50预测误差小于0.25°CA;与BP网络和RBF网络相比,具有更低的误差和更强的泛化能力;与直接热力学计算方法相比,具有突出的抗干扰性和容错能力。
引用
收藏
页码:352 / 357
页数:6
相关论文
共 8 条
[1]  
A Mean-Value Model for Control of Homogeneous Charge Compres-sion Ignition(HCCI)Engines. Rausen D J,Stefanopoulou A G,Kang J M,et al. American Control Con-ference . 2004
[2]  
Study on Ignition Timing Control for Diesel Engines U-sing In-Cylinder Pressure Sensor. Mamoru Hasegawa,Yunchi Shimasaki,Satoshi Yamaguchi,et al. SAE paper 2006-01-0180 . 2006
[3]  
Motion Pictures of EngineFlames Correlated with Pressure Cards. Withrow L,Rassweiler G M. SAE Paper970037 . 1997
[4]  
A Real-Time Platform for Closed-Loop Controland Crank Angle Based Measurement. Klas Telborn. . 1997
[5]  
Design and performance of an intelligent predictive controller for a six-degree-of-freedom robot using the Elman network. Koker,R. Information Sciences . 2006
[6]  
Speed Estimation of an In-duction Motor Using Elman Neural Network. Toqeer Raja S,Bayindir NSuha. Neurocom-puting . 2006
[7]  
Neural Network-Based Diesel Engine Emissions Pre-diction Using In-Cylinder Combustion Pressure. Michael L Traver,Richard J Atkinson,Christopher M At-kinson. SAE Pa-per1999-01-1532 . 1999
[8]  
Experimental Investigations on Con-trolled Auto-Ignition Combustion in a Four-Stroke GasolineEngine. Aaron John Oakley. . 2001