基于岭回归的数控机床温度布点优化及其热误差建模

Temperature Measuring Point Optimization and Thermal Error Modeling for NC Machine Tool Based on Ridge Regression

提出一种基于岭回归分析的数控机床温度布点优化方法。数控机床热误差建模一般采用多元线性回归方法,在多元线性回归模型中,隐含着要求解释变量之间无强相关性的假定。然而在实际的建模中,各自变量与因变量之间的相互关系并不与简单相关系数所反映的情况完全吻合。通过岭迹对温度变量进行优化选择,实现了温度测点优化布置,并选用适当的岭参数k建立了数控机床热误差的多元线性回归优化模型,提高了热误差模型的精确性和鲁棒性。

A kind of NC machine tool thermal error measurement points optimization method was presented based on ridge regression. Usually muhiple linear regression method is used to build the thermal error model of NC machine tool. In multiple linear regression model, a hypothesis is connoted that no strong correlations are existed among the variables. But in real modeling situation, the relationships among the independent variables and the dependent variables are not so perfect matched by the simple correlation coefficient. The ridge trace was used to optimize temperature variables and optimal layout of the temperature measuring points was achieved. The proper ridge parameter was supplied to establish a muhiple linear regression optimization model for NC machine tool thermal errors. The accuracy and robustness of the error modeling are improved.