摘要
主要介绍了3种基于小波包分解的以不同方式进行提取刀具磨损振动信号特征向量的方法。刀具振动信号通过小波包分解后重构成不同频段的信号系数。在此基础上,首先提取各个频段能量基于总能量比值的特征向量;其次对其进行功率谱分析,提取特定频段幅值的特征向量;最后,利用奇异值分解将不同频段的信号映射到正交子空间中,从中选取信号的奇异值作为特征向量。最终将得到的特征向量组合成一个特征向量输入支持向量机中进行刀具磨损识别。
Three different ways are described based on wavelet packet to extract the feature vector of vibration signals of tool wear. Tool vibration signal is decomposed by wavelet packet, and then re-constituted different frequency band signal coeffi- cients. On this basis, firstly, extracting the ratio of energy of each frequency hand based on total energy as a feature vector. Secomlly, extracting the amplitude of specific fl-equency bands in its power spectrum as a feature vector. Finally, different fre- quency bands of signal are mapped to the orthogonal subspace by using the singular value decomposition （SVD） , The singular values of the signal are selected as a feature vector in the subspace. A signal feature vector composed with three feature vectors is finally inputted into support vector machines tot tool which being recognized.
出处
《机械研究与应用》
2013年第4期58-60,共3页
Mechanical Research & Application
基金
北京市人才强教深化计划项目(PHR201106227)
北京市教委科研计划项目(SQKM201211232003)
校重点课程建设项目
