摘要
针对数控(Computer numerical control,CNC)机床主轴热漂移误差建模及预测问题,提出一种基于贝叶斯推断的最小二乘支持向量机(Least squares support vector machine,LS-SVM)建模方法。以一台双转台五轴加工中心为研究对象,进行热误差测量试验,利用非接触式激光位移传感器及温度传感器同步测量机床主轴各运动方向热漂移误差及温度变化数值,获取建模数据。模型训练过程运用贝叶斯推断方法对LS-SVM的正规化参数、核函数参数进行优化选择,获取基于参数后验概率最大化的最优参数组合,进而构建可准确预测机床主轴热漂移误差的优化模型。分别利用基于贝叶斯推断的LS-SVM模型、传统LS-SVM模型以及BP神经网络(Back propagation artificial neural networks,BP-ANN)模型对机床变工况条件下主轴热漂移误差进行预测,通过预测效果对比,基于贝叶斯推断的LS-SVM模型具有更高的预测精度,在机床变工况条件下仍具有较高鲁棒性与泛化能力,可以很好地弥补现有建模方法的部分局限性。
A new least squares support vector machine （LS-SVM） modeling method based on Bayesian inference is presented to predict the spindle thermal distortion of computer numerical control （CNC） machine tool. With laser displacement sensors and thermal resistor sensors, a series of measurements on a 5-axis CNC machining center with double rotate tables are carried out to acquire spindle thermal distortions of different directions and temperature values for modeling. During the modeling process, the parameters of LS-SVM are chosen by Bayesian inference method, which can obtain the optimized parameters of maximum posterior probability. With these optimized parameters, the established model can predict the spindle thermal distortion accurately. The LS-SVM model based on Bayesian inference, the conventional LS-SVM model and the Back propagation artificial neural networks （BP-ANN） model are respectively used to predict the spindle thermal distortion under variable working conditions. With the prediction results comparison, the LS-SVM model based on Bayesian inference has the best prediction accurate, and is more effective and robust especially when the spindle runs under variable working conditions, it can smoothlg over some shortages of the traditional modeling methods.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2013年第15期115-121,共7页
Journal of Mechanical Engineering
基金
国家科技重大专项资助项目(2011ZX04015-031)
关键词
数控机床
主轴热漂移
建模
支持向量机
贝叶斯推断
CNC machine tool Spindle thermal distortion Modeling Support vector machine Bayesian inference
