利用PVDF压电传感器实现接触滑动的快速检测  

Rapid detection of contact sliding using PVDF piezoelectric sensors

作　　者：吴海彬[1,2] 黄力文 WU Haibin;HUANG Liwen(School of Mechanical Engineering and Automation,Fuzhou University,Fuzhou 350116,China;The Key Laboratory of Special Intelligent Equipment Safety Measurement and Control,Fuzhou 350007,China)

机构地区：[1]福州大学机械工程及自动化学院,福州350116 [2]福建省特种智能装备安全与测控重点实验室,福州350007

出　　处：《振动与冲击》2024年第24期135-144,共10页Journal of Vibration and Shock

基　　金：科技部重点研发计划课题(2018YFB1308603)。

摘　　要：工业机器人在进行工件抓取过程中,往往存在夹持力过大使工件破损、夹持力过小导致工件滑落的矛盾。为此,提出一种接触滑动的快速检测方法,采用聚偏二氟乙烯(polyvinylidene fluoride, PVDF)压电传感器作为滑觉感知元件。首先,利用阿基米德优化算法(Archimedes optimization algorithm, AOA)优化变分模态分解(variational mode decomposition, VMD)对传感器信号进行分解与重构,降低噪声干扰;然后,提取信号的时频域特征,构建信号特征集;最后,使用蜣螂优化算法(dung beetle optimization, DBO)优化选取长短期记忆网络(long short-term memory networks, LSTM)参数,将DBO优化选取后的参数和信号特征集用于构建滑动检测识别模型。将所提滑动检测方法应用于电动夹爪抓取试验,结果表明,该方法实现了接触状态的精准快速识别,准确率达到100%,识别时间在20 ms以内,根据识别结果可实时调整电动夹爪夹持力大小。In the process of grasping workpieces,industrial robots often face the contradiction of excessive gripping force causing damage to the workpiece,and too little gripping force leading to slippage.To address this issue,a rapid sliding detection method was proposed using polyvinylidene fluoride(PVDF)piezoelectric sensors as tactile perception elements.First,the sensor signal was decomposed and reconstructed using the variational mode decomposition(VMD)optimized by the Archimedes optimization algorithm(AOA)to reduce noise interference.Next,the time-frequency domain features of the signal were extracted to construct the signal feature set.Finally,the dung beetle optimization(DBO)algorithm was used to optimize the selection of parameters for long short-term memory networks(LSTM).The optimized parameters obtained from DBO along with the signal feature set were applied to construct the sliding detection recognition model.The proposed sliding detection method was applied to an experiment involving electric gripper grasping.Results demonstrate precise and rapid recognition of contact status,achieving 100%accuracy with recognition times under 20 ms.Based on the recognition results,the gripping force of the electric gripper can be adjusted in realtime.

关 键 词：滑动检测 聚偏二氟乙烯(PVDF) 阿基米德优化算法(AOA) 变分模态分解(VMD) 长短期记忆网络(LSTM) 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置] TN911.7[自动化与计算机技术—控制科学与工程]