摘要
无级变速器起步离合器的占空比变化率和油温的变化,影响了离合器的线性机理模型反映执行机构实际性能的准确性,引起了车辆起步控制品质的下降。针对这一问题,基于台架试验数据,分别建立了基于支持向量机和径向基神经网络的离合器数学模型,在不同的油温和占空比变化率下,对线性机理建模、支持向量机和神经网络三种建模方法进行了对比分析。结果表明,离合器支持向量机模型和神经网络模型的控制精度明显优于其线性机理模型,同时支持向量机模型比神经网络模型具有更强的泛化能力。
Both the duty cycle change rate and then oil temperature changge of continuously variable transmission（ CVT） starting clutch influence the accuracy that linear mechanism model reflects real property of the actuator,which cause the decline of vehicle starting control quality. To solve this problem,the clutch mathematic models of Support Vector Machine（ SVM） and Radial Basis Function（ RBF） neural network were established based on the bench experimental data. Under the different temperature and duty cycle change rate,contrastive study of three modeling methods,namely linear mechanism,Support Vector Machine and RBF Neural Network,were done. The results show that the control precision of RBF Neural Network and SVM are obviously superior to linear mechanism model. At the same time,SVM model has more strong generalization than RBF neural network.
出处
《机械设计与研究》
CSCD
北大核心
2015年第2期77-80,88,共5页
Machine Design And Research
基金
国家自然科学基金资助项目(51175156)
国家国际科技合作专项资助(2014DFA70170)
关键词
离合器
试验建模
支持向量机
RBF神经网络
clutch
experimental modeling
Support Vector Machine
RBF neural network
