基于最小二乘支持向量机的超磁致伸缩执行器磁滞非线性模型  

Hysteresis Nonlinearity Model of Giant Magnetostrictive Actuator Based on Least Squares Support Vector Machine

作　　者：李跃松[1] 朱玉川[1,2] 吴洪涛[1] 牛世勇[3] 田一松[3] 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016 [2]浙江大学流体动力与机电系统国家重点实验室,浙江杭州310027 [3]西安飞行自动控制研究所,陕西西安710065

出　　处：《机床与液压》2012年第23期4-6,共3页Machine Tool & Hydraulics

基　　金：国家自然科学基金资助项目(51175243;50805080);航空科学基金资助项目(20110752006);流体动力与机电系统国家重点实验室2011年度开放基金资助项目(GZKF-201116);江苏省普通高校研究生科研创新计划资助项目(CXZZ11_0196)

摘　　要：为准确模拟超磁致伸缩执行器的磁滞非线性特性,基于最小二乘支持向量机与实验数据建立了超磁致伸缩执行器的模型,结果显示所建模型都能够充分逼近非线性实验数据点。为提高模型的求解速度,仅选用一半的实验数据建立了小样本模型,并与神经网络所建小样本模型对比,结果表明:基于最小二乘支持向量机所建模型的位移预测误差小于1.2μm,而基于神经网络所建模型的位移预测误差大于1.5μm。The model of giant magnetostrictive actuator based on least squares support vector machine and the experimental data was developed to describe the hysteresis nonlinearity of giant magnetostrictive actuator. The results show that least squares support vec- tor machine can fit experimental data accurately. In order to solve the model quickly, only a half of experimental data were used for modeling, and the model based neural network was also built to make comparison with the model based on least squares support vector machine. The results show that the displacement error of model based on least squares support vector machine is less than 1.2μm, the displacement error of model based on neural network is more than 1.5 μm.

关 键 词：超磁致伸缩 磁滞非线性 最小二乘支持向量机 神经网络 

分 类 号：TH113.1[机械工程—机械设计及理论]