机床热误差测温点优化及建模方法研究  

Research on Optimization and Modeling Method of Temperature Measurement Point for Thermal Error of Machine Tool

作　　者：李焕昭 张爱梅[1] 裴雪巍 LI Huan—zhao;ZHANG Ai—mei;PEI Xue-wei(School of Mechanical Engineering,Zhengzhou University,He'nan Zhengzhou 450001,China)

机构地区：[1]郑州大学机械与动力学院,河南郑州450001

出　　处：《机械设计与制造》2022年第5期143-146,共4页Machinery Design & Manufacture

基　　金：河南省重大科技专项—数控车床智能化关键技术研究(171100210300-01)。

摘　　要：为了减小热误差对数控机床精度的影响,提出了基于DBSCAN聚类算法的温度传感器测点优化方法和基于BP神经网络的数控机床建模方法。通过DBSCAN对特征数据进行聚类分析消除部分线性相关传感器数据。求解聚类后的每个类别中的传感器数据与主轴误差值的皮尔森相关系数,将类别内相关系数按从大至小进行排序,选取类别内相关系数最大的作为优化后的传感器数据,以此将温度测量点从16个减少到5个。添加动态随机数完成数据增强,提高模型泛化性。建立了温度和主轴位移的BP神经网络模型,其准确度可达0.94,为机床热误差补偿提供了重要的理论依据。To reduce the influence of thermal error on the accuracy ofCNC machine tool this paper proposed a temperature sensor measuring point optimization method based on DBSCAN clustering algorithm and a BP neural network modeling method for CNC machine tool. T Cluster analysis of feature data by DBSCAN to eliminate some linear correlation sensor data. Pearson correlation coefficient for calculating the difference between temperature data and spindle error in each category,The coefficients are sorted from large to small.In each category,the largest correlation coefficient is selected as the optimized temperature measurement point,so as to reduce the temperature measurement points from 16 to 5. Add dynamic random number to enhance the data and improve the generalization of the model.The BP neural network model of temperature and spindle displacement is established,and its accuracy can reach 0.94,which provided an important theoretical basis for the machine tool thermal error compensation.

关 键 词：机床 DBSCAN聚类算法 BP神经网络 热误差 

分 类 号：TH16[机械工程—机械制造及自动化] TG502.15[金属学及工艺—金属切削加工及机床]