融合参数化模型与LightGBM的机床进给轴动态位置预测方法  

A Dynamic Position Prediction Method for Machine Tool Feed Axis by Integrating Parameterized Models and LightGBM

作　　者：孙健 姬帅[2] 倪鹤鹏 叶瑛歆 Sun Jian;Ji Shuai;Ni Hepeng;Ye Yingxin(School of Mechanical and Electrical Engineering,Shandong Jianzhu University,Jinan 250101,China;School of Mechanical Engineering,Shandong University,Jinan 250061,China)

机构地区：[1]山东建筑大学机电工程学院,济南250101 [2]山东大学机械工程学院,济南250061

出　　处：《机电工程技术》2024年第4期267-272,共6页Mechanical & Electrical Engineering Technology

基　　金：山东省自然科学基金青年项目(ZR20210E128)。

摘　　要：实现进给轴动态位置的准确预测对于优化机床加工轨迹,补偿轮廓误差,进而提高加工精度具有重要意义。为此,提出了一种融合参数化模型与Light Gradient Boosting Machine(LightGBM)的进给轴动态位置预测方法,能够根据目标加工轨迹实现进给轴位置的高精度预测。首先,建立了包含伺服驱动系统和机械系统的进给轴参数化模型,作为位置预测的基础模型。然后,针对参数化模型预测存在的残差,引入集成学习策略,提出了基于LightGBM的残差补偿模型建模方法,提高补偿精度和训练效率。最后,实验结果表明,相比于传统的参数化模型,所提出的预测方法具有更高的精度。与多种典型数据驱动模型的对比实验表明,提出的基于LightGBM的残差补偿模型具有更高的精度与稳定性,平均误差和均方根误差均减小了40%以上。Accurately predicting the dynamic position of the feed axis is of great significance for optimizing the machining trajectory of machine tools,compensating for contour errors,and improving machining accuracy.To this end,a dynamic position prediction method for the feed axis is proposed,which integrates a parameterized model with the Light Gradient Boosting Machine(LightGBM).It can achieve high-precision prediction of the feed axis position based on the target machining trajectory.Firstly,a parameterized model of the feed axis is established,which includes a servo drive system and a mechanical system,as the basic model for position prediction.Then,in response to the residuals predicted by parameterized models,an ensemble learning strategy is introduced,and a residual compensation model modeling method based on LightGBM is proposed to improve compensation accuracy and training efficiency.Finally,the experimental results indicate that the proposed prediction method has higher accuracy compared to traditional parameterized models.Comparative experiments with various typical data-driven models show that the residual compensation model based on LightGBM proposed has higher accuracy and stability,with an average error and root mean square error reduction of over 40%.

关 键 词：进给轴动态位置预测 参数化模型 集成学习 LightGBM 

分 类 号：TH161[机械工程—机械制造及自动化]