Kalman滤波在刀具磨损预测模型中的应用

Application of Kalman filter in the model of tool wear prediction

基于LS-SVM建立刀具磨损预测模型,描述铣削过程中输入向量(进给率、切削速度、主轴转速、切削深度、切削时间及磨损位置)和输出向量(刀具磨损)之间的映射关系,并引入Kalman滤波技术,建立LS-KF模型,考虑加工条件及环境变化引起的刀具磨损量的变化,结合刀具的实际磨损量更新LS-SVM的预测结果,并用该更新结果调整训练模型,以使更新后的刀具磨损量能够反映出由于加工条件及环境的变化引起的刀具磨损的变化,提高LS-SVM模型的预测精度,最后用实验验证所建立模型的预测精度。结果表明,LS-SVM模型和LS-KF模型的预测精度均较高,且LS-KF模型的预测精度更高。

In this study, the least squares support vector machines （LS-SVM） was used to establish the tool wear prediction model which described the mapping relationship between input vectors （feed rate, cutting speed, spindle speed, cutting depth, cutting time, and wear location） and output vector （tool wear）. In light of the change in tool wear resulting from the change in cutting conditions and environ- ment, Kalman filter was employed to build LS-KF model, which considered the actual tool wear and updated the prediction of LS-SVM. The updated data was used to optimize the training model so as to raise the accuracy of prediction by LS-SVM model. Experiments show that both LS-SVM model and LS-KF model boast high precision yet the latter displays a higher precision than the former.