基于卡尔曼滤波算法的刀具磨损状态预测

Prediction of Tool Wear Based on Kalman Filter Algorithm

针对机加工过程中的刀具状态预测问题,利用随机森林算法良好的多分类能力,提出一种基于卡尔曼滤波算法刀具状态分类监测模型。以铣削力信号作为状态监测信号,通过卡尔曼滤波以及前一次走刀过程的铣削力建立状态预测方程,对下一次走刀过程中的铣削力进行预测;将预测的铣削力分别在时域、频域与小波包分析中提取特征值,然后利用训练好的随机森林模型对提取的特征值进行识别,预测下次走刀过程的刀具磨损状态。数据处理的结果表明,该方法可以有效地预测出下次加工是否会进入磨损阶段,在整体状态识别中准确率为98.73%。

Aiming at the problem of tool state prediction in the machining process, a tool state classification monitoring model based on the Kalman filter algorithm is proposed by using the good multi-classification ability of the random forest algorithm.Using the milling force signal as the condition monitoring signal, the state prediction equation is established by Kalman filtering and the milling force of the previous pass process to predict the milling force in the next pass process.The predicted milling force is extracted in the time domain, frequency domain, and wavelet analysis respectively, and the extracted feature values are identified by using the trained random forest model to predict the tool wear status of the next pass process.The results of data processing show that the method can effectively predict whether the next process will enter the wear stage, the accuracy rate in the overall state recognition is 98.73%.