韩荣港,梁兴雨,吕旭,等.基于神经网络的柴油机润滑油稀释率预测方法研究[J].内燃机工程,2023,44(5):74-81.
基于神经网络的柴油机润滑油稀释率预测方法研究
Study on Prediction Method of Lubricating Oil Dilution Rate of Diesel Engine Based on Neural Network
DOI：10.13949/j.cnki.nrjgc.2023.05.010
关键词:柴油机  远后喷策略  润滑油稀释  果蝇优化算法  广义回归神经网络
Key Words:diesel engine  late post-injection strategy  oil dilution  fruit fly optimization algorithm（FOA）  generalized regression neural network（GRNN）
基金项目:内燃机可靠性国家重点实验室开放性课题项目（skler-202004）
作者单位E-mail
韩荣港* 天津大学 内燃机燃烧学国家重点实验室，天津 300072 hrg_0910@tju.edu.cn 
梁兴雨* 天津大学 内燃机燃烧学国家重点实验室，天津 300072 lxy@tju.edu.cn 
吕旭 天津大学 内燃机燃烧学国家重点实验室，天津 300072 lvxu96@tju.edu.cn 
王昆 天津大学 内燃机燃烧学国家重点实验室，天津 300072 kwang5@tju.edu.cn 
刘军 潍柴动力股份有限公司，潍坊 261041 liujun02@weichai.com 
王意宝 潍柴动力股份有限公司，潍坊 261041 wangyib@weichai.com 
摘要点击次数: 663
全文下载次数: 476
摘要:为了实现柴油机润滑油稀释程度的准确、快速检测，基于试验建立了润滑油稀释率与理化参数指标的数据集。利用果蝇优化算法（fruit fly optimization algorithm， FOA）搜寻最优解更新广义回归神经网络（generalized regression neural network， GRNN）的光滑因子，提出了一种多理化指标参数融合的润滑油稀释程度预测方法。仿真结果表明：该模型的拟合优度为99.9%，均方根误差为0.106。通过将4种模型进行对比，证明了FOA–GRNN模型在预测精度、收敛速度及稳定性上的优越性。在实际柴油机远后喷试验中，将该预测方法与气相色谱（gas chromatograph， GC）法进行对比，二者的绝对误差在0.5%之内。该预测方法在保证检测精度的同时大大缩短了检测时间，为柴油机实现按质换油提供了理论和技术指导。
Abstract:To realize accurate and rapid detection of diesel engine lubricating oil dilution degree， the data set of the lubricating oil dilution rates and physical and chemical properties associated with the lubricant was established through experiments. The fruit fly optimization algorithm（FOA） was used to search the optimal solution to updating the smoothing factor of the generalized regression neural network（GRNN）， and a prediction method of lubricating oil dilution rate based on FOA–GRNN model was then proposed. The simulation results show that the goodness of fit of the model could reach 99.9%， and the root mean square error was 0.106. Compared with other network models， FOA–GRNN model is proved to be superior in prediction accuracy， convergence speed and stability. The proposed prediction method was verified by gas chromatograph（GC） method in the actual diesel engine late post-injection experiment， and the error between the modeling results and measurements results was within 0.5%. The prediction method ensures the detection accuracy while shortening the detection time. And the method provides theoretical and technical guidance for the oil change of diesel engines.
查看全文  HTML   查看/发表评论