数控机床工作台进给系统故障诊断研究

发布时间：2019-06-19 13:09

【摘要】：数控机床是现代工业生产的主力设备，特别是在加工结构复杂、大型和高精密零件时，数控机床发挥了不可替代的作用。但是数控机床通常处于高速、变载以及往复冲击的工作环境下，，长时间的工作数控机床可能会产生故障，特别是一些机械部件如丝杠、轴承、导轨等。开展数控机床故障诊断研究可以及时的发现机床故障并找出故障隐患，从而提高机床的可靠性，并推动数控机床故障诊断技术由故障后维修和定期维修到实时维修的转变，达到降低维修的成本，创造更大经济效益的目的。本文研究了数控机床的常见故障形式及其故障机理并基于BP神经网络设计了数控机床工作台进给系统的故障诊断系统。主要包括故障类型及机理分析、实验方案的设计、数据采集系统的软硬件设计、信号分析与特征值提取和基于神经网络的故障诊断模型设计等内容。重点研究了信号处理技术包括信号预处理技术、特征提取技术和特征选择技术以及两级故障诊断模型的设计和实现等。首先，研究了数控机床的常见故障及其机理，对故障发生比较频繁的机械部件进行了重点研究。并以此为根据设计了实验方案，包括故障件的选择和设置、测点的选择，传感器的选择和安装，以及具体实验流程的设计等。其次，研究了数据采集技术，并设计了数据采集系统，包括硬件系统设计和软件系统设计两大部分。硬件设计是在NI-PXI的基础上选择了数据采集平台和数据采集卡以及相应的线缆和调理设备并对其参数进行了设定；软件系统设计主要基于LabVIEW和MATLAB平台设计了数据采集模块、数据分析模块和数据库管理模块三大模块，并编制了程序。再次，研究了数据处理技术，对本文所采集到的数据的处理共分为三大步。第一步，对采集到的数据进行信号预处理，包括去除奇异点处理和信号零均值处理；第二步，对经过预处理的信号分别进行时域分析、频域分析和小波分析，并提取相应的时频特征值；第三步，对提取的时频特征值进行进一步的选择和提取，包括特征值初步选择和基于核主元分析的特征提取两部分，最终得到用于故障诊断的特征值。最后，建立了基于BP神经网络的数控机床工作台进给系统的两级故障诊断模型。第一级为总网络，用来诊断不同部件的故障；第二级为各个子网络，用来诊断同一部件的不同故障，分为滚动轴承网络和滚珠丝杠网络两个子网络。两级故障诊断模型实现了故障的初步判别和故障的细化诊断功能。
[Abstract]:CNC machine tool is the main equipment of modern industrial production, especially when the machining structure is complex, large and high precision parts, CNC machine tool plays an irreplaceable role. However, CNC machine tools are usually in the working environment of high speed, variable load and reciprocating impact, and long working CNC machine tools may have faults, especially some mechanical components such as screw, bearing, guideway and so on. The fault diagnosis research of NC machine tool can find out the fault of machine tool in time and find out the hidden trouble, so as to improve the reliability of machine tool, and promote the transformation of fault diagnosis technology of NC machine tool from post-fault maintenance and regular maintenance to real-time maintenance, so as to reduce the cost of maintenance and create greater economic benefits. In this paper, the common fault forms and fault mechanism of NC machine tools are studied, and the fault diagnosis system of NC machine tool table feed system is designed based on BP neural network. It mainly includes the analysis of fault type and mechanism, the design of experimental scheme, the design of software and hardware of data acquisition system, signal analysis and eigenvalue extraction, and the design of fault diagnosis model based on neural network. The signal processing technology, including signal preprocessing technology, feature extraction technology and feature selection technology, as well as the design and implementation of two-level fault diagnosis model, are studied in detail. Firstly, the common faults and their mechanisms of NC machine tools are studied, and the mechanical components with frequent faults are studied. According to this, the experimental scheme is designed, including the selection and setting of fault parts, the selection of measuring points, the selection and installation of sensors, and the design of specific experimental flow. Secondly, the data acquisition technology is studied, and the data acquisition system is designed, including hardware system design and software system design. On the basis of NI-PXI, the hardware design selects the data acquisition platform and data acquisition card, as well as the corresponding cable and conditioning equipment, and sets its parameters. The software system design mainly designs three modules based on LabVIEW and MATLAB platform: data acquisition module, data analysis module and database management module, and compiles the program. Thirdly, the data processing technology is studied, and the data processing collected in this paper is divided into three steps. In the first step, the collected data are preprocessed, including the removal of singular points and the zero-mean processing of the signal. In the second step, the preprocessed signals are analyzed in time domain, frequency domain and wavelet, and the corresponding time-frequency eigenvalues are extracted. In the third step, the extracted time-frequency eigenvalues are further selected and extracted, including the preliminary selection of eigenvalues and the feature extraction based on kernel principal component analysis, and finally the eigenvalues for fault diagnosis are obtained. Finally, a two-stage fault diagnosis model of NC machine tool table feed system based on BP neural network is established. The first level is the general network, which is used to diagnose the faults of different components, and the second level is each sub-network, which is used to diagnose the different faults of the same component, which is divided into two sub-networks: rolling bearing network and ball screw network. The two-stage fault diagnosis model realizes the functions of preliminary fault discrimination and fault refinement diagnosis.
【学位授予单位】：青岛理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TG659;TH165.3

【参考文献】