发动机支持向量机建模及精度影响因素

来源期刊:中南大学学报(自然科学版)2010年第4期

论文作者:陈然 孙冬野 秦大同 罗勇 胡丰宾

文章页码:1391 - 1397

关键词:支持向量机;发动机模型;辨识精度;参数选择

Key words:support vector machine; engine model; identification precision; parameters selection

摘    要:针对发动机具有非线性、时变性的特点以及采用常规神经网络辨识时的过学习等问题,提出基于支持向量机(SVM)的发动机模型辨识方法。该方法以大量实测数据为基础,采用结构风险最小化准则(SRM),保证网络具有很强的推广特性。以MATLAB为平台,依据实测试验数据,研究核函数、损失函数及惩罚参数对系统辨识精度的影响,确定各参数对模型精度影响的程度。在充分考虑各参数之间交互作用的前提下,利用循环嵌套查找方法,获得使支持向量机网络辨识精度达到最优时的各参数值,并以此建立发动机转矩及油耗模型。研究结果表明:基于支持向量机的发动机模型具有较强的泛化能力,为实现发动机与传动系统共同工作的最佳匹配控制奠定了基础。

Abstract: Focus on the nonlinearity and time-varying properties of engine and the existence of local minima, over-learning of conventional neural networks engine, a novel engine identification model was proposed based on support vector machine (SVM). Using the MATLAB,many precision influencing factors of engine torque were analyzed, and the precision influencing sequence of these parameters was confirmed. Considering the interaction among the parameters, the optimal parameters of SVM engine model was found out using loop nest program method. Testing results show that the engine identification model based on engine test data and structure risk minimization (SRM) principle has very strong generalization ability. The SVM engine reveals high precision and strong generalization ability, which provides a foundation for engine matching power train.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 主办 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(湘)字第028号   湘ICP备09001153号