Abstract: To improve the accuracy of robot welding weld shape prediction,a fuzzy neural network weld shape prediction method based on the fusion of Intuitionistic Fuzzy C-means clustering and Adaptive inertial weight Particle Swarm Optimization (IFCM-APSO) was proposed.T  he foot width and welding height of T-shaped welds was taken as the evaluation criteria in this method,and four variables that affected welding quality:welding rate,laser power,wire feed rate,and shielding gas flow rate was selected as input parameters.The central value and width of the membership function in the network were optimized to ensure that the input and output parameters of robot welding had a good fit.Finally,simulations and experiments showed that the   proposed  method could better non-linearly fit the input and output parameters of robot welding,improve its global search ability and convergence speed,and solve the problem that traditional fuzzy neural network was easy to fall into the local minima during training.

Key words: intuitionistic fuzzy C-means clustering, inertial weighted particle swarm algorithm, adaptive fuzzy neural network, robot welding, weld shape prediction

摘要： 为提高机器人焊接焊缝外形预测的准确性,提出一种基于直觉模糊C均值聚类和自适应惯性权重粒子群算法(IFCM-APSO)相融合的模糊神经网络焊缝外形预测方法。该方法以T型焊缝的焊脚宽度和焊高高度作为评价标准,选择影响焊接质量的焊接速率、激光功率、送丝速率和保护气体流量这4种变量作为输入参数,对自适应模糊神经网络中隶属函数的中心值和宽度进行优化,以保证机器人焊接的输入和输出参数具有较好的拟合性。最后,经过仿真和实验表明,所提出焊缝外形预测方法能较好地对机器人焊接的输入和输出参数进行非线性拟合,提高了其全局搜索能力和收敛速度,解决了传统模糊神经网络训练过程中容易陷入局部极小点的问题。

关键词: 直觉模糊C均值聚类, 惯性权重粒子群算法, 自适应模糊神经网络, 机器人焊接, 焊缝外形预测