基于稀疏分解和支持向量机的高速铣削刀具磨损状态监测

发布时间：2018-06-17 14:36

本文选题：高速铣削加工 + 刀具磨损状态监测　；参考：《中国科学技术大学》2017年硕士论文

【摘要】：近年来高速铣削加工技术凭借其出众的加工精度、加工效率和表面质量等优势在众多领域得到了广泛应用。在高速铣削加工中,铣刀在超高转速下进行不连续切削,刀具磨破损迅速,直接影响加工精度与产品质量,严重时甚至损坏机床和工件,引起事故。因此,对高速铣削加工过程中刀具的磨损状况进行实时的在线监测意义重大。本文借助于先进的传感技术,在稀疏表示和模式识别的基础上提出了一种新的故障诊断方法,以达到对刀具磨损状态的实时监测,提高生产系统的安全性。论文的主体工作包含以下几点:(1)学习并总结了近年来在高速铣削加工领域针对于刀具磨破损状态监测的各种科学方法,并介绍了各个科研机构和学者的研究进展,研究了刀具磨损的机理和分类问题,为该课题的开展奠定扎实的理论基础。(2)基于压缩感知和稀疏表示的理论,结合形态分量分析和增广拉格朗日变量分离算法,构造对偶BP优化模型,并使用SALSA算法对优化模型进行求解,达到了对信号的脉冲成分和谐波成分进行分离的目的.随后对该算法的分离效果进行仿真分析和验证。(3)搭建高速铣削加工实验平台,介绍了该平台的理论构成和各模块的使用情况,并进行了传感信号的采集与储存。针对加工过程中振动信号的特点及其在频域上的稀疏特性,对振动信号进行稀疏分解和形态分量分析,分离出振动信号中的脉冲成分和谐波成分。对分离后的信号分量分别提取脉冲密度和高次谐波频率与基频的幅值比等特征,并分别与刀具磨损形成的向量之间的相关性,探索这些特征的物理意义和在刀具磨损状态监测上的实用性。(4)构造多类别支持向量机分类器,将通过稀疏分解得到的特征样本输入分类器中进行训练和学习,使该分类器具备通过多个特征有效地辨识刀具所处的磨损状态的功能,并将新的实验数据输入到分类器中进行刀具磨损状态预测,以进行验证该分类器分类效果。
[Abstract]:In recent years, high speed milling technology has been widely used in many fields because of its outstanding machining accuracy, processing efficiency and surface quality. In high speed milling, the milling cutter is discontinuous cutting at ultra-high speed, and the tool wear is damaged quickly, which directly affects the machining precision and product quality, and even damages the machine tool and workpiece seriously, causing accidents. Therefore, it is of great significance to monitor tool wear in real time during high speed milling. In this paper, a new fault diagnosis method based on sparse representation and pattern recognition is proposed with the help of advanced sensing technology to achieve real-time monitoring of tool wear and improve the safety of production system. The main work of this paper includes the following points: 1) Learning and summing up various scientific methods in the field of high speed milling in recent years, and introducing the research progress of various scientific research institutions and scholars. This paper studies the mechanism and classification of tool wear, which lays a solid theoretical foundation for the development of the subject. It is based on the theory of compressed perception and sparse representation, combined with morphological component analysis and augmented Lagrangian variable separation algorithm. The dual BP optimization model is constructed and the SALSA algorithm is used to solve the optimization model. The aim of separating the pulse component from the harmonic component of the signal is achieved. Then the separation effect of the algorithm is simulated and verified. (3) A high-speed milling experimental platform is built. The theoretical structure of the platform and the usage of each module are introduced, and the sensing signals are collected and stored. According to the characteristics of vibration signal and its sparsity in frequency domain, the vibration signal is decomposed by sparse decomposition and morphological component analysis, and the pulse component and harmonic component of vibration signal are separated. For the separated signal components, the characteristics such as pulse density, amplitude ratio of high harmonic frequency to fundamental frequency are extracted, respectively, and the correlation between them and the vectors formed by tool wear is obtained, respectively. To explore the physical meaning of these features and the practicability of tool wear state monitoring, a multi-class support vector machine classifier is constructed, and the feature samples obtained by sparse decomposition are trained and studied in the classifier. The classifier has the function of identifying the wear state of the tool effectively through several features, and the new experimental data are input into the classifier to predict the tool wear state, so as to verify the classification effect of the classifier.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG54

【参考文献】