摘要
采用高斯拟合法对实测离子电流进行拟合,并从中提取出离子电流峰值及峰值位置等9个特征参数,应用BP神经网络算法根据离子电流特征参数计算得出缸内压力参数。研究结果表明:应用高斯拟合法得到的拟合电流曲线与实测离子电流曲线吻合良好;利用高斯拟合法简化了离子电流信号特征参数的提取过程,且可以进行离子电流信号的去噪,方便对后续离子电流信号的处理与压力计算;应用BP神经网络可以根据离子电流的9个特征参数计算得出压力的峰值及峰值位置,压力峰值的平均绝对误差小于0.1MPa,平均相对误差小于2%,位置的平均绝对误差小于2°CA,平均相对误差小于等于5%。
Gaussian curve fitting method was applied to fit the ion current, and nine characteristic parameters such as ion current peaks and their positions, etc. were extracted from the ion current signals. According to these parameters,the max. in-cylinder pressure was calculated based on the BP neural network algorithm. Results show that measured and fitted ion current curves agree well with each other. The Gaussian curve fitting method can greatly simplify the ion current signal characteristic parameter extraction process and carry out ion current signal de-noising, thus greatly convenient for the latter signal processing and pressure computation. Based on analysis of these 9 ion current characteristic parameters,in-cylinder pressure peak and its position can be effectively calculated by using BP network. For the pressure peak,the mean absolute error is less than 0. 1 MPa and the mean relative error less than 2%;for the pressure peak position, the mean absolute error is less than 2℃ A and the mean relative error not more than 5%.
出处
《内燃机工程》
EI
CAS
CSCD
北大核心
2013年第5期13-17,共5页
Chinese Internal Combustion Engine Engineering
基金
国家自然科学基金项目(51176150)
陕西省教育厅专项科研计划项目(2010JK689)
西安科技大学培育基金项目(2009018)
关键词
内燃机
离子电流
高斯拟合
BP神经网络
缸内压力
IC engine
ion current
Gussian curve fitting
BP neural network
in-cylinder pressure
