基于IFWA-ELM的六维力传感器解耦算法  

Decoupling Algorithm of Six-Axis Force Sensor Based on IFWA--ELM

作　　者：董翔 许子健 曹会彬[2,3] 孙玉香 高理富[2,3] DONG Xiang;XU Zijian;CAO Huibin;SUN Yuxiang;GAO Lifu(School of Electrical Engineering and Automation,Anhui University,Hefei Anhui 230601,China;Hefei Institutes of Physical Science,Chinese Academy of Sciences,Hefei Anhui 230031,China;School of Information Science and Technology,University of Science and Technology of China,Hefei Anhui 230027,China)

机构地区：[1]安徽大学电气工程与自动化学院,安徽合肥230601 [2]中国科学院合肥物质科学研究院,安徽合肥230031 [3]中国科学技术大学信息科学技术学院,安徽合肥230027

出　　处：《传感技术学报》2023年第12期1943-1951,共9页Chinese Journal of Sensors and Actuators

基　　金：安徽省重点研发计划项目(2022a05020035);中国科学院战略性先导科技专项项目(XDA22040303);安徽省科技重大专项项目(202103a05020022);国家自然科学基金重大研究计划重点支持项目(92067205)。

摘　　要：针对六维力传感器的维间耦合严重影响测量精度的问题,提出了一种基于改进烟花算法优化极限学习机(IFWA-ELM)的解耦算法。首先,对烟花算法的爆炸半径、变异算子和选择策略进行改进,形成改进烟花算法(IFWA)。其次,采用改进烟花算法寻找极限学习机的最佳网络参数,解决极限学习机随机生成初始权值和阈值导致网络不稳定、隐含层神经元数量对网络性能影响较大的问题。为了验证算法的解耦性能,以应用于4500 m深海机械臂的六维力传感器作为研究对象,采用最小二乘法(LS)、BP神经网络(BPNN)、极限学习机(ELM)和IFWA-ELM算法进行解耦实验。实验结果表明:IFWA-ELM算法具有较好的非线性解耦能力,解耦后Ⅰ类误差控制在0.27%以内,Ⅱ类误差控制在0.13%以内,有效提高了六维力传感器的测量精度。Aiming at the problem that the dimensional coupling of six-axis force sensor seriously affects the measurement accuracy,a decoupling algorithm based on the improved fireworks algorithm to optimize the extreme learning machine(IFWA-ELM)is proposed.Firstly,the explosion radius,mutation operator and selection strategy of fireworks algorithm are improved to form the improved fireworks algorithm(IFWA).Secondly,the improved fireworks algorithm is used to find the best network parameters of extreme learning machine,which solves the problems that the initial weights and thresholds randomly generated by extreme learning machine lead to network instability,and the number of hidden layer neurons has great influence on network performance.In order to verify the decoupling performance of the algorithm,the six-dimensional force sensor applied to the 4500 m deep-sea manipulator is taken as the research object,and least square method(LS),BP neural network(BPNN),extreme learning machine(ELM)and IFWA-ELM algorithm are adopted for decoupling experiments.The experimental results show that IFWA-ELM algorithm has good nonlinear decoupling ability.After decoupling,the class I error is controlled within 0.27%,and the classⅡerror is controlled within 0.13%,which effectively improves the measurement accuracy of six-axis force sensor.

关 键 词：六维力传感器 静态解耦 极限学习机 烟花算法 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]