基于粒子群算法与BP网络的机床主轴热误差建模

发布时间：2018-09-01 06:41

【摘要】：为了避免基于反向传播(BP)神经网络的主轴热误差模型精度低、通用性不强和收敛性较差等缺点,利用模糊聚类理论与相关分析法对温度变量进行优化,并且选取热敏感点以挖掘温度变量与热误差间的相关性,降低温度变量间的耦合性.利用粒子群优化算法(PSO)将预测输出与期望输出间误差平方和的倒数作为个体适应度函数,将个体头部分与身体部分的表现码分别映射为网络的隐含层节点数、权值和阈值,实现了对BP网络的拓扑结构的有效优化,通过跟踪个体极值和全局极值实现了粒子群个体速度与位置的更新.分别建立了基于BP和PSO-BP网络的热误差模型,以精密坐标镗床主轴为研究对象,采用五点法对高速主轴热误差进行测量.结果表明,PSO-BP模型可实现不同工况下主轴空间位姿状态的高精度预测,验证了测量及建模方法的有效性.
[Abstract]:In order to avoid the shortcomings such as low accuracy, poor generality and poor convergence of the spindle thermal error model based on back-propagation (BP) neural network, the fuzzy clustering theory and correlation analysis are used to optimize the temperature variables. In order to reduce the coupling between the temperature variables and the temperature variables, the correlation between the temperature variables and the thermal errors is excavated by selecting the heat sensitive points. Using particle swarm optimization (PSO) algorithm, the reciprocal of square sum of error between predictive output and expected output is taken as individual fitness function, and the performance codes of individual head part and body part are mapped to the number of hidden layer nodes, weights and thresholds of the network, respectively. The topology of BP network is optimized effectively, and the individual velocity and position of PSO are updated by tracking individual extremum and global extremum. The thermal error models based on BP and PSO-BP networks are established respectively. Taking the spindle of precision coordinate boring machine as the research object, a five-point method is adopted to measure the thermal error of high-speed spindle. The results show that the PSO-BP model can be used to predict the spatial position and attitude of the spindle under different working conditions, and the validity of the measurement and modeling method is verified.
【作者单位】：西安交通大学机械制造系统工程国家重点实验室;
【基金】：国家科技重大专项资助项目(2014ZX04001051-07) 国家自然科学基金创新群体项目(51421004) 中国博士后科学基金特别资助项目(2014T70910)
【分类号】：TG502

【相似文献】