偏最小二乘回归在刀具磨损试验建模中的应用

Application of Partial Least-squares Regression Analysis Method in Modeling Analysis of Tool Wear

应用灰色关联度筛选变量,将样本变量筛选为部分样本变量和全部样本变量组合用于建模的自变量,以刀具后刀面磨损量作为因变量,对不同切削条件下铣削加工过程刀具后刀面磨损的多组实验数据,采用偏最小二乘回归分析方法,建立对所选自变量的偏最小二乘回归模型。结果表明,该方法提取的成分具有线性无关的特点,对刀具磨损有较好的解释能力,且利于建模和预测,同时可以消除输入因素的多重共线性。刀具磨损的回归模型计算出不同切削条件下刀具后刀面的磨损量,以全部样本变量所建立的模型预测效果更为理想。

Tool wear model based on grey relational degree filtering variable and the partial least-squares regression analysis （PLSRA） was brought forward to forecast tool wear loss. According to grey relational degree analysis, sample variables were filtered as part sample variables and used as independent variable for modeling fit together all sample variables, and the tool wear loss was taken as dependent variable, thus the PLSRA model was built up through several experimental data of tool wear in different cutting conditions of milling process. It indicated that elements picked up in the method have a characteristic of non-linearity, which can account tool wear much for tool wear, and is beneficial for building up model and forecasting, in addition, this method can eliminate multiply jointly linearity. The tool wear loss in different cutting conditions can be figured out using the tool wear regression model. Especially, the model is more effective by using all sample variables.