摘要
针对主要基于受综合因素影响的机床本体温度所建立的热误差模型鲁棒性较差的问题.综合考虑机床本体温度、动力源转速、冷却液温度及环境温度提出了多变量关联热误差组合模型.将最小二乘支持向量机(LS-SVM)的方法运用到热误差建模中,并利用偏最小二乘(PLS)方法提取输入变量的主成分作为LS-SVM的输入,形成PLS-LSSVM组合热误差模型.此外根据数控加工过程及材料热变形原理,将相对起始温度的差温值作为温度输入,使热误差补偿更加准确.在某型号精密加工中心进行实验验证,结果表明:PLS-LSSVM模型比LS-SVM更稳定,比PLSR预测精度高;考虑差温多变量的PLS-LSSVM模型较单纯考虑机床本体测量温度值的PLS-LSSVM~*模型,热误差预测值的均方根误差(RMSE)平均减少了5.5μm.
To solve the problem that based on machine tool temperature that is comprehensively influenced by other factors,and thus the thermal error models had poor robustness,a multivariate correlative and combined thermal error model was put forward,which overall considers machine tool temperature,speed of power source,and temperature of coolant and environment. Least squares support vector （LS-SVM） method was applied to the thermal error model,and partial least squares （PLS） method was applied to extract the principal components as the input of LS-SVM, and the PLS-LSSVM thermal error combined model was then formulated. In addition,this model set the differential temperatures,relatively with initial temperatures,as the temperature variable,which is based on the process of numerical control machining and the principle of material thermal deformation,to make the thermal error compensation more accurate. It was tested on a precision machining center,whose results showed that the PLS-LSSVM thermal error model is more stable than the LS-SVM model,and more accurate than the partial least squares regression （PLSR） model. Besides,the root mean square error （RMSE） of the predictive thermal error with the PLS-LSSVM model is 5. 5 μm on average less than that with the PLS-LSSVM ＊ model, which only takes into account the temperature measurements of the machine tool.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2017年第5期700-705,共6页
Journal of Northeastern University(Natural Science)
基金
辽宁省科技创新重大专项(201301002)
辽宁省科学技术计划重大项目(2015106016)
关键词
数控机床
热误差模型
热误差影响因素
偏最小二乘
最小二乘支持向量机
CNC machine tool
thermal error model
influence factor of thermal error
partial least squares
least squares support vector machine
