一种大摆角五轴联动混联机床位置正解的神经网络分析法  

The Forward Kinematic Solution for Actuator of Five-axis Linkage Novel Hybrid Kinematics Machine With Large Pendulum Angle

作　　者：薄瑞峰[1] 鲁岩[1] 冯鹏升 李瑞琴[1] 

机构地区：[1]中北大学机械与动力工程学院,山西太原030051

出　　处：《机械设计与研究》2017年第1期22-26,共5页Machine Design And Research

基　　金：国家自然科学基金资助项目(51275486);山西省回国留学人员科研资助项目(2014050);山西省国际科技合作资助项目(2015081016)

摘　　要：根据大摆角五轴联动混联机床中并联模块的结构特点,提出两种求解位置正解的方法:一是以并联模块中各个驱动支链之间的距离为约束条件,提出一种解析方法求解正运动学解;二是以位置反解结果作为训练样本,提出一种利用多层前向神经网络求解机构位置正解的方法,通过构造神经网络并采用levenbergmarquardt算法训练,实现了机床并联模块从关节变量空间到工作空间的非线性映射,从而求得其运动学正解。结果表明:与解析法相比,该方法计算精度高,计算过程简洁,耗时少,可用于该机床工作空间的求解和控制。Based on the structure of parallel module which belongs to five-axis linkage novel hybrid kinematics machine （HKM） with large pendulum angle, two effective methods for the forward kinematic solution were presented. The first method is an analytical method, which takes the distance between of the driving chain of parallel modules as the constraint condition to get the forward kinematic solution. The second is a multi-layer forward neural network based method, in which the inverse solution result is regarded as training samples to train the ANN using LM （levenberg- marquardt） algorithm. A nonlinear mapping of the parallel module from the joint variables space to work variables space is obtained. The trained ANN can be adopted to acquire the result of the forward kinematic solution. The results show ANN based method computational precision with less complicated formula and calculations. Moreover, the calculation process is comparatively simple and less time consuming. It provides a promising way to solve the forward kinematic solution to control or analyze the workspaee of the HKM.

关 键 词：五轴联动混联机床 并联模块 位置正解 解析方法 BP神经网络 

分 类 号：TH124[机械工程—机械设计及理论]