织机车间直捻机筒纱抓取机械臂逆运动学求解方法研究  

Study on the Inverse Kinematic Solving Method of Grasping Mechanical Arm in Loom Workshops

作　　者：于俊康 江维 李红军 陈伟 陈振 YU Junkang;JIANG Wei;LI Hongjun;CHEN Wei;CHEN Zhen(Wuhan Textile UniversitySchool of Mechanical Engineering and Automation,Wuhan 430200,China;Wuhan Textile UniversityHubei Provincial Key Laboratory of Digital Textile Equipment,Wuhan 430200,China)

机构地区：[1]武汉纺织大学机械工程与自动化学院,武汉430200 [2]武汉纺织大学数字化纺织装备湖北省重点实验室,武汉430200

出　　处：《纺织工程学报》2024年第1期21-32,共12页JOURNAL OF ADVANCED TEXTILE ENGINEERING

基　　金：数字化纺织装备湖北省重点实验室开放课题资助项目(DTL2023013)。

摘　　要：纺织车间中通过协作机械臂代替工人实现对直捻机筒纱的自动更换,可降低工人劳动强度,提高筒纱卷绕的生产效率。在对机器人进行运动学建模与分析时,逆运动学求解是机器人运动学中关键部分。传统法求机器人逆解存在求解速度慢、求解过程复杂、结果稳定性差等问题,采用传统BP(BackPropa-gationgNeuralNetwork)神经网络求解又容易陷入局部极小值,针对上述问题,提出一种基于PSO优化算法(ParticleSwarmOptimizationAlgorithm)的BP神经网络机械臂逆运动学求解方法,通过PSO算法对BP神经网络的权值和阈值进行多次迭代优化,避免了局部最小值的问题,提高了神经网络的全局搜索能力。采用D-H法建立机器人运动学模型,根据机器人正运动学方程由关节角度解得末端位姿,将解得结果作为数据集,通过学习算法经多次迭代确定神经网络的模型参数,并对神经网络进行性能检验。实验结果表明:PSO-BP神经网络相比于传统的BP神经网络收敛速度快,该模型在搬运机器人逆运动学求解中精度高,满足纺织车间直捻机筒纱抓取作业的需要。In the textile workshop,the automatic replacement of cylindrical yarn packages on the twisting machine is achieved by using a collaborative robotic arm instead of human workers.It can reduce the labor intensity of workers and improve the production efficiency of winding yarn packages.In the process of modeling and analyzing the robot's motion,the solution to inverse kinematics is a critical aspect of robot kinematics.Traditional methods for solving the inverse kinematics of robots suffer from issues such as slow computation speed,complex solving processes,and poor stability of results.Utilizing the traditional Back Propagation Neural Network(BPNN)for solving also tends to get stuck in local minima.To address these challenges,this paper proposes a method for solving the inverse kinematics of a robotic arm based on the Particle Swarm Optimization Algorithm(PSO)optimized Backpropagation(BP)neural network.Through multiple iterations using the PSO algorithm,the weights and thresholds of the BP neural network are optimized, preventing it from getting stuck in local minimaand enhancing its global search capabilities. The robot's kinematic model is established using the Denavit-Hartenberg (D-H) method, and the end-effector pose is obtained by solving joint angles through the robot's forwardkinematic equations. The results obtained serve as a dataset, and the model parameters of the neural networkare determined through multiple iterations using a learning algorithm, followed by performance testing.Experimental results indicate that the PSO-BP neural network converges faster compared to the traditional BPneural network. The model exhibits high precision in solving the inverse kinematics of the material handling robot,meeting the requirements for yarn-grabbing operations in the textile workshop.

关 键 词：纺织车间 直捻机 筒纱卷绕 机械臂逆运动学 PSO-BP神经网络 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]