论文部分内容阅读
对声发射信号进行5层小波分解提取6个频段的能量值,把它与切削速度、切削深度、进给量和切削时间一起作为刀具状态的特征向量.通过主元分析进行降维、消除特征向量间的相关性后,把得到的主元作为BP(Back Propagation)神经网络的输入向量.BP神经网络应用改进的LM(Levenberg-Marquart)算法进行学习,利用输入向量对网络进行训练后,实现对刀具后刀面磨损量VB的预测.实验结果显示:基于主元分析和LM算法改进的BP神经网络建立的预测系统,网络输出与实测VB值的误差0.03以内;根据预测VB值的范围可判别出刀具的不同状态.
The acoustic emission signal is decomposed by five layers to extract the energy values of six frequency bands and use it as the eigenvector of the tool state along with the cutting speed, depth of cut, feed rate and cutting time, and reduce the dimension by principal component analysis Vector, the principal component obtained is used as the input vector of BP (Back Propagation) neural network.The BP neural network is studied by the improved Levenberg-Marquart (LM) algorithm and the input vector is used to train the network The tool flank wear VB prediction.The experimental results show that: based on the principal component analysis and LM algorithm improved BP neural network to establish the prediction system, the network output and the measured VB value error of 0.03 or less; according to the range of predicted VB value Discerning the different states of the tool.