应用于机械手的磁致伸缩触觉传感器阵列与物体识别  

Object Recognition of Magnetostrictive Tactile Sensor Array Applied to Manipulator

作　　者：黄珊 王博文[1,2] 赵智忠[1,2] 王亮[1,2] 翁玲[1,2] Huang Shan;Wang Bowen;Zhao Zhizhong;Wang Liang;Weng Ling(State Key Laboratory of Reliability and Intelligentization of Electrical Equipment Hebei University of Technology,Tianjin,300130,China;Key Laboratory of Electromagnetic Field and Electrical Apparatus Reliability of Hebei Province Hebei University of Technology,Tianjin,300130,China)

机构地区：[1]河北工业大学省部共建电工装备可靠性与智能化国家重点实验室,天津300130 [2]河北工业大学河北省电磁场与电器可靠性重点实验室,天津300130

出　　处：《电工技术学报》2021年第7期1416-1424,共9页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(51777053,51801053);河北省自然科学基金(重点项目E2017202035,青年项目E2019202315)资助项目。

摘　　要：利用磁致伸缩材料铁镓合金(Galfenol)设计制作了触觉传感单元,并将其集成为阵列,安装在二指机械手上。基于电磁原理、逆磁致伸缩效应和欧拉伯努利动力学原理,建立触觉传感单元力测量模型。设计的传感单元测力范围为0~3N,当压力小于1.5N时,灵敏度为151mV/N,在1.5~3N区间,灵敏度约为109mV/N,测量较低的静态力时具有更高的灵敏度,在2~4Hz范围内对动态力具有动态响应快、灵敏度高的特点。根据机械手的测量范围,选取形状较为规则的六种饮料瓶样品进行了抓取,搭建实验系统进行数据采集与存储,测试了传感器阵列的输出特性,集成的传感器阵列能很好地辨识所选样品,对于装满水样品有更高的精度。选择合适的特征值,应用支持向量机算法,通过选择不同的特征值数量和训练集与分类集的比例,对采集的数据进行分类识别。经计算可知,选择三个特征值且训练集与分类集比例为9:1时识别率最高,为87.5%。The tactile sensor unit was designed and manufactured using the magnetostrictive material Galfenol,which was integrated into an array and installed on the two-finger manipulator.The tactile sensor unit force measurement model was established based on the electromagnetic principle,the inverse magnetostrictive effect and the Euler Bernoulli dynamics principle.The measuring range of the sensor unit is 0~3N.When the pressure is less than 1.5N,the sensitivity is 150mV/N.In the range of 1.5~3N,the sensitivity is about 103mV/N.It has higher sensitivity when measures lower static force,and has the characteristics of fast dynamic response and high sensitivity to dynamic force in the range of 2~4Hz.According to the measurement range of the manipulator,six kinds of beverage bottles with regular shapes were selected.The experimental system was built for collecting and saving data and testing the output characteristics of the sensor.The integrated sensor array can identify the selected sample well,and has higher accuracy for the sample filled with water.By selecting appropriate eigenvalues and applying support vector machine algorithm,the collected datas are classified by selecting different eigenvalues and the proportion of training sets and classification sets.According to the calculation,when three eigenvalues are selected and the ratio between training sets and classification sets is 9:1,the recognition rate is the highest,which is 87.5%.

关 键 词：铁镓合金 触觉传感器阵列 支持向量机 物体识别 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]