%0 Journal Article
%T 基于流形学习与隐马尔可夫模型的刀具磨损状况识别<br>Tool Wear Condition Monitoring Based on Manifold Learning and Hidden Markov Model
%A 张栋梁
%A 莫蓉
%A 孙惠斌
%A 李春磊
%J 西北工业大学学报
%D 2015
%X 为了提高金属铣削过程中的刀具磨损状态识别的自动化程度与精度,提出了基于局部切空间排列(LTSA)方法与隐Markov模型(HMM)来识别刀具的不同磨损状态的方法。该方法首先利用小波分析技术对铣削过程中的切削进给方向力信号进行处理,构造了高维特征空间。然后使用基于流形学习方法实现了高维特征空间的维数约简。最终利用约简后的低维特征向量训练HMM,从而实现刀具磨损状态的识别。实验结果说明该方法能够有效地识别铣削过程的刀具磨损状态。与未经特征维数约简的识别方法相比,新方法能够提高刀具磨损状态的识别效率与准确率。<br>In order to improve the automation and the precision of tool wear condition recognition in the process of metal milling, we proposed the method based on the manifold learning——the local tangent space alignment (LTSA) method——and the hidden Markov model (HMM) to identify tool wear conditions. First, this method used the time domain and the wavelet analysis technique for signal processing of the milling cutting axial force to construct the high dimensional feature space. Then, the local tangent space alignment (LTSA) method was used to achieve the dimensionality reduction. At last, the low dimensional feature vector was used to train the HMM in order to recognize tool wear conditions. Additional tests were conducted to check the feasibility of the method. Comparison of the performance of the proposed method with that of the method of identification without the feature dimension reduction shows that the proposed method can improve the efficiency and the accuracy of tool wear condition recognition
%K 维数约简
%K 刀具磨损状态识别
%K 流形学习
%K 隐马尔可夫模型(HMM)&lt
%K br&gt
%K condition monitoring
%K conformal mapping
%K design of experiments
%K efficiency
%K eigenvalues and eigenfunctions
%K errors
%K experiments
%K feature extraction
%K flowcharting
%K hidden Markov models
%K matrix algebra
%K milling (machining)
%K monitoring
%K probability
%K signal processing
%K time domain analysis
%K wavelet analysis
%K wear of materials
%K dimension reduction
%K local tangent space alignment (LTSA)
%K the manifold learning
%K tool wear conditions recognition
%U http://journals.nwpu.edu.cn/xbgydxxb/CN/abstract/abstract6427.shtml