支持向量机在机床热误差建模补偿中的机理研究

针对数控机床热误差预测补偿功能,以Leaderway-V450数控机床为试验对象,通过跨季度的5批次数据,比对分析支持向量机(ε-SVR)和多元回归(MLR)两种建模方法的拟合和预测精度。研究得出,环境温度会对数控机床热误差预测模型产生干扰,模型拟合精度和预测精度在这种干扰下不具备等同性。线性ε-支持向量机在试验条件相近、环境温度接近的情况下鲁棒性较好,拟合精度、预测精度要高于多元回归算法,但当环境温度变化较大时,其预测精度较差。将跨季度环境温度条件下的数据进行综合建模,两种模型的预测精度均有所提高,但拟合精度优越的支持向量机算法建立的模型鲁棒性则低于传统的多元回归算法模型。

Study on Mechanism of Support Vector Model in Thermal Error Modeling and Forecasting of CNC Machine Tools

For CNC machine tool thermal error prediction compensation function,this paper takes Leaderway-V450 CNC for example,analyze the fitting and prediction precision of MLR and SVR through 5 data in different quarters of the year.It is demonstrated that the ambient temperature should be taken into account when estimating the thermal error model.Besides,fitting and prediction precision are not same when ambient temperature is different.The robustness of SVR is good when the experiment condition and temperature are close.Both fitting and prediction precision are better than MLR.However,the prediction precision of SVR is not better than MLR when the temperature changes a lot.Combining data from different quarters of the year and modeling again,the prediction precision of them have increased,but the robustness of precision on SVR is worse than MLR.