Aiming at online predicting tool wear accurately,a method based on the regression algo-rithm of LS-SVM was proposed.First the acoustic emission signals were decomposed into several in-trinsic mode functions(IMF)employing empirical mode decomposition.Then,an AR model of each IMF was established respectively.AR model coefficients were extracted to construct feature vector.Fi-nally,the feature vectors were feed into LS-SVM and prediction of tool wear was realized.The experi-mental results show that it can predict the amount of tool wear after 10s according to the current cut-ting conditions and the proposed method has better accuracy compared with neural network algo-rithm.%提出了基于最小二乘支持向量机回归算法的刀具磨损量预测方法。该方法首先利用经验模态分解算法对非线性、非平稳的声发射信号进行平稳化处理,得到了若干个固有模态函数;然后建立了每个固有模态函数的自回归模型,并提取模型系数构造特征向量;最后采用最小二乘支持向量机回归算法实现了刀具磨损量的预测。该方法与神经网络预测算法相比,具有更高的预测准确率,可有效预测当前切削状态下10 s后的刀具磨损量。
展开▼