车轮多分力传感器静态解耦方法  

Research on Static Decoupling Methods for Self-developed Multi-component Wheel Force Transducer

作　　者：张小龙[1] 冯能莲[1] 张为公[1] 马德贵[1] 

机构地区：[1]清华大学汽车安全与节能国家重点实验室,北京工业大学环境与能源工程学院,东南大学仪器科学与工程学院,安徽农业大学工学院,博士后讲师(安徽农业大学),100084北京市,教授通讯作者,100022北京市,教授博士生导师,210096南京市,讲师,230036合肥市

出　　处：《农业机械学报》2008年第4期18-23,共6页Transactions of the Chinese Society for Agricultural Machinery

基　　金：安徽农业大学稳定和引进人才科研资助项目(项目编号:2007-01-01);安徽农业大学校长青年基金项目(项目编号:2007qnr17);江苏省交通科学研究计划项目(项目编号:05C02)

摘　　要：从标定方法和解耦算法两方面对车轮力传感器静态耦合特性进行了研究。给出了详细的标定过程和样本获取方法,基于实际标定样本应用3种回归模型对多分力传感器维间耦合进行量化对比分析。结果表明:标定主通道线性特性显著;自行研制轮力传感器静态耦合率与国外产品相当;最小二乘支持向量回归机回归精度高、泛化能力强和算法稳定;对标准正交最小二乘径向基神经网络算法改进回归效果显著。Both calibration method and static decoupling algorithm were employed to analyze wheel force transducer （WFT）. Firstly, the calibration procedure and extraction of sample data were presented in detail based on the self-developed hydraulic bench. Then, three decoupling methods were utilized respectively to quantify the coupling effects. The main findings are as the follows： the linearity of each main calibration channel is notable; the calculated rate of static coupling of the self-developed WFT is equal to the same-type foreign product; the least square support vector regression （LS-SVR） algorithm owns the characteristics of high regression precision, outstanding generalization and excellent algorithm stability; the method to modify the algorithm of the standard OLS-RBF NN （orthogonal least square radial-basis-function neural network） improves the regression performance significantly.

关 键 词：车轮力传感器 维间耦合 静态解耦 最小二乘支持向量机 

分 类 号：TP212.6[自动化与计算机技术—检测技术与自动化装置] U463.34[自动化与计算机技术—控制科学与工程]