基于CEEMD-WPT的刀具磨损状态识别研究

发布时间：2018-06-06 21:39

  本文选题：刀具磨损状态 + 互补总体平均经验模态分解　； 参考：《南京信息工程大学》2017年硕士论文

【摘要】：随着机床在自动化、集成化和无人化方向发展的越来越快,如何保证加工产品的质量和生产效率就显得尤为重要。而刀具作为加工过程的直接执行者,不可避免地存在着磨损现象。因此为了保证产品质量的同时又实现对刀具的高效利用,就有必要对刀具状态监测技术展开研究。针对刀具加工特点,本文选择对刀具声发射信号进行监测,声发射监测技术作为一种有效的无损检测技术因其灵敏度高、抗干扰能力强、无需停机等优点已经得到广泛的应用,但由于采集得到的声发射信号频率高,数据量大且频率成分复杂,无法直接进行刀具状态识别,故本文为了准确定性的掌握切削过程中刀具的磨损状态,提出了一种基于互补总体平均经验模态分解(Complementary Ensemble Empirical Mode Decomposition, CEEMD)和小波包变换(Wavelet Package Transform, WPT)的刀具状态监测方法。首先利用CEEMD将声发射信号自适应地分解成多个固有模态函数(Intrinsic Mode Function, IMF),每个IMF内都包含有原信号的不同时间尺度特征,针对依然存在的模态混叠问题的IMF,利用WPT良好的局部处理能力予以修正,从而实现对特征分量的精确提取,然后选取修正后能量值较大的前几个IMF分量,求其占总能量的比重组成特征向量,最后输入到支持向量机(Support Vector Machine, SVM)中进行训练与测试,从而建立起由6个SVM二值分类器组成的4类刀具状态识别系统。文章通过与CEEMD特征提取方法进行比较,说明CEEMD-WPT提取的特征更加精确,更具有表征性,将两种时频分析方法结合起来,既有效的解决了CEEMD分解后依然存在的模态混叠问题,又消除了单独进行WPT处理后产生虚假频率分量、频率混淆现象的影响,为后期精确地识别出刀具磨损状态奠定了好的基础。
[Abstract]:With the development of automatic, integrated and unmanned machine tools, it is very important to ensure the quality and production efficiency of processed products. As the direct executor of machining process, tool wear is inevitable. Therefore, it is necessary to study the tool condition monitoring technology in order to ensure the product quality and realize the efficient use of cutting tools. According to the characteristics of tool machining, this paper chooses to monitor the tool acoustic emission signal. As an effective nondestructive testing technology, acoustic emission monitoring technology has been widely used because of its high sensitivity, strong anti-interference ability and no need to stop. However, due to the high frequency of acoustic emission signal collected, the large amount of data and the complexity of frequency components, it is impossible to recognize the cutting tool state directly, so in order to accurately and qualitatively grasp the tool wear state in the cutting process, A tool condition monitoring method based on complementary Ensemble empirical Mode decomposition (CEEMDM) and Wavelet package transform (WPT) is proposed. At first, acoustic emission signals are decomposed adaptively into Intrinsic Mode functions (IMFMs) by CEEMD. Each IMF contains different time scale characteristics of the original signals. Aiming at the IMF of the existing modal aliasing problem, this paper uses the good local processing ability of WPT to correct the feature components, and then selects the first few IMF components with large corrected energy. Finally, it is input into support vector machine support Vector Machine (SVM) for training and testing, and four types of tool state recognition system composed of six SVM binary classifiers are established. Compared with CEEMD feature extraction method, this paper shows that the feature extraction of CEEMD-WPT is more accurate and more characterizing. The combination of two time-frequency analysis methods can effectively solve the modal aliasing problem that still exists after CEEMD decomposition. The effect of false frequency component and frequency confusion after WPT processing alone is eliminated, which lays a good foundation for accurate identification of tool wear state in later stage.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG71

【参考文献】

中国期刊全文数据库 前10条

1 王丽华;陶润U，

本文编号：1988201