摘要
在数控机床的热误差补偿技术中,机床温度信息的提取对改善机床的加工精度至关重要。首先对广泛使用的模糊聚类多元线性回归模型在变工况下的性能进行了试验,结果证明:试验工况变化后,该模型预测值失准。通过方差膨胀因子判断,这种现象是由模型自变量的复共线性引起。为了改进上述模型,提出了一种温度特征提取的建模方法,通过特征提取算法,提取模糊聚类优化测点的综合特征,从而得到综合特征自变量,最后利用综合自变量进行回归建模。试验表明,该方法有效消除了复共线性对模型预测精度和鲁棒性的影响,优化后的回归模型均方根误差在4μm以内,可有效预测76%以上误差,相较于其他方法表现出优良的预测性能,易于在其他机床热误差补偿中推广使用。
In thermal error compensation technology on CNC machine tool,it is crucial for improving the accuracy of machine to extract temperature information. Firstly, analyzing the performance of popular model whose temperature point is optimized by fuzzy clustering when operating condition varied. The analysis shows the difference of the operating conditions leads to poor forecast per- formance of the model. Further analysis shows the poor performance is resulted from muhi-collinearity. For the sake of improving the model, a method of optimizing independent variables is proposed. Through a feature extraction algorithm, the method obtains comprehensive feature variables which are extracted from temperature points optimized by fuzzy clustering. Finally, the compre- hensive variables are utilized to establish a multiple linear regression model. The experiment results indieated that the proposed method effectively eliminated the influence of muhi-collinearity on the accuracy and robustness of the model. The root mean square error of the optimized model was within 4 ~n and can forecast over 76 % thermal errors. The method shows a superior performance comparing with other methods and is easier to be applied widely to thermal error modeling of other CNC machine tool.
出处
《现代制造工程》
CSCD
北大核心
2018年第2期64-69,共6页
Modern Manufacturing Engineering
基金
国家自然科学基金项目(51605381)
四川省科技厅科技支撑计划项目(2016GZ0205)
四川省教育厅重点项目(16ZA0415)
关键词
数控机床
热误差
测点优化
特征提取
复共线性
变量优化
CNC machine tool
thermal error
critical point optimization
feature extraction
multi-collinearity
variable optimization