基于异类传感器融合的数控机床伺服系统故障诊断关键技术研究

发布时间：2018-09-06 09:11

【摘要】：数控机床加工精度高、质量稳定、加工速度快、生产效率高,是实现制造技术和装备现代化的基石。伺服系统是数控机床及众多复杂数控设备的关键部分,其性能好坏直接决定了整个设备的精度、工作效率及可靠性等。随着数控机床的不断发展,朝着结构复杂化和高度自动化方向发展的数控机床伺服系统,各部分之间的关联更加密切,微小的故障往往会爆发连锁反应,严重时会导致整个机床的性能异变、缩短寿命甚至报废,后果危害性极大。机械设备的预测维修理念核心是通过对机械设备运行过程的工况监测、正确估计故障发展趋势和演变规律,找出故障原因及时采取措施进行维修保养,达到“维修出效益”到“预测出效益”的理念变革。数控机床伺服系统与国防军工、航空航天行业中的重型设备的伺服系统具有相同特征,研究成果相互通用。因此,开展数控机床伺服系统故障诊断的基础理论与基本方法研究,对于提高我国数控机床在监测、诊断、维护等方面的科技水平,十分必要。本文紧紧围绕异类传感器融合“采集什么信息”、“如何采集信息”、“如何利用信息”的三个关键技术,结合目前数控机床伺服系统故障诊断研究存在的问题,即研究对象多为单一零部件、对内置传感器的价值挖掘不足、一直被作为故障分类和模式识别的问题来研究,层层推进、步步深入,逐一展开了重点研究工作。以数学建模手段建立了整个伺服系统的复杂数学模型并进行了稳定性判别,理论分析了伺服系统各种典型故障机理与表现,建立了故障表现与内部参数之间的映射关系,并通过仿真进行了验证,为利用内置传感器获取机床本体信息的可靠性提供理论依据。以此为基础,结合外置传感器检测某一部件的传统方法,提出了利用内外置交叉互补的异类传感器融合的新方法。搭建了异类传感器融合的试验系统,研究了利用内外置传感器获取伺服系统本体信息的关键技术即数据对准技术,结合现有的试验基础条件,提出了一种适用于802DSL数控系统和NI数据采集系统同步采集的时间对准方案。与单单利用外置传感器采集典型故障信息或者单单利用内置传感器获取本体信息相比,丰富了信源、拓宽了信道。针对试验发现的通过外置传感器检测滚动轴承故障信号频率与利用故障特征频率计算公式计算的故障特征频率存在着试验误差及间谐波倍频误差这一问题,重点研究了误差产生机理和积累与传递过程。通过对误差改善和提高频率分辨率各种技术手段的研究,得出基于特征频率计算的滚动轴承故障诊断方法本身具有不可消除的模糊性。接着,提出一种基于数据驱动的滚动轴承故障诊断新方法。然后,对该方法模糊证据获取得到的不确定度概率参数的理论分析和物理意义的挖掘,在故障诊断领域引入了直觉模糊集的概念,提出了将随机集框架下的模糊证据获取与匹配转变为直觉模糊证据获取及多元决策融合的新思路,并进行了理论分析和试验研究。研究证明:一直作为故障分类和模式识别问题研究的多源信息融合的故障诊断,也可以看作多元决策融合问题。建立了基于直觉模糊决策加权融合的数控机床伺服系统故障分级诊断模型。首先,研究了时域、频域及小波包降噪与EMD分解相结合的多域特征参数提取方法和基于极值间距的特征筛选及特征相关分析的数据降维。然后,构建了基于遗传BP网络、RBF网络与SVM的多分类器分级故障识别模型,并对三种智能识别模型的诊断能力进行了比较分析,提出了利用单一分类器模型的诊断准确率作为权重系数,构建基于加权集结算子的直觉模糊决策融合的数控机床伺服系统智能分级诊断模型,试验证明该方法对不同分类器之间存在分歧的样本识别能力强、准确率高,体现了方法本身的容错和自纠正能力。
[Abstract]:NC machine tools are the cornerstone of the modernization of manufacturing technology and equipment with high precision, stable quality, fast processing speed and high production efficiency. Servo system is the key part of NC machine tools and many complex NC equipment. Its performance directly determines the accuracy, efficiency and reliability of the whole equipment. With the development of the CNC machine tool servo system, which is developing toward the direction of complex structure and high automation, the relationship between the parts is more close. Small faults often break out chain reaction, which will lead to the performance variation of the whole machine tool, shorten the life and even scrap. The consequences are very harmful. By monitoring the operating conditions of mechanical equipment, correctly estimating the development trend and evolution law of the fault, finding out the causes of the fault and taking timely measures for maintenance, the concept transformation from "repairing benefit" to "predicting benefit" can be achieved. The servo system has the same characteristics, and the research results are common to each other. Therefore, it is necessary to study the basic theory and method of fault diagnosis for CNC machine tool servo system to improve the scientific and technological level of monitoring, diagnosis and maintenance of CNC machine tools in China. The three key technologies of "how to collect information" and "how to use information" are combined with the existing problems in fault diagnosis research of CNC machine tool servo system, that is, the research object is mostly a single component, and the value of the built-in sensor is not excavated enough, which has been studied as a problem of fault classification and pattern recognition. The complex mathematical model of the whole servo system is established by means of mathematical modeling and its stability is discriminated. The typical fault mechanism and performance of the servo system are analyzed theoretically. The mapping relationship between fault performance and internal parameters is established and verified by simulation. Based on the reliability of the built-in sensor to obtain the ontology information of the machine tool, combining with the traditional method of detecting a part with the external sensor, a new method of fusion of heterogeneous sensors with the internal and external cross-complementary is proposed. The key technology of servo system ontology information is data alignment technology. Combining with the existing experimental basic conditions, a time alignment scheme for synchronous acquisition of 802DSL CNC system and NI data acquisition system is proposed. It can collect typical fault information with external sensors or acquire ontology information with internal sensors. In order to solve the problem that there are test error and inter-harmonic frequency doubling error between the fault characteristic frequency calculated by the formula of fault characteristic frequency and the fault signal frequency detected by the external sensor, the mechanism of error generation and the process of accumulation and transmission are emphatically studied. Through the study of various technical means to improve the error and enhance the frequency resolution, it is concluded that the fault diagnosis method of rolling bearing based on the calculation of characteristic frequency has its own indelible fuzziness. Then, a new method of fault diagnosis of rolling bearing based on data-driven is proposed. The concept of intuitionistic fuzzy sets is introduced in the field of fault diagnosis. A new idea is proposed to transform the fuzzy evidence acquisition and matching under the framework of random sets into intuitionistic fuzzy evidence acquisition and multi-decision fusion. The theoretical analysis and experimental research are carried out. As a problem of fault classification and pattern recognition, multi-source information fusion can also be regarded as a problem of multi-decision fusion. A hierarchical fault diagnosis model of CNC machine tool servo system based on intuitionistic fuzzy decision weighted fusion is established. Firstly, the multi-domain features of time domain, frequency domain and wavelet packet denoising combined with EMD decomposition are studied. Feature parameter extraction method and data dimension reduction based on extremum distance feature selection and feature correlation analysis. Then, a multi-classifier hierarchical fault identification model based on genetic BP network, RBF network and SVM is constructed, and the diagnostic ability of the three intelligent recognition models is compared and analyzed, and the diagnosis based on single classifier model is proposed. Accuracy is taken as weight coefficient, and an intelligent hierarchical diagnosis model of CNC machine tool servo system based on intuitionistic fuzzy decision fusion with weighted aggregator is constructed. The experiment proves that the method has strong ability of identifying samples with different classifiers and high accuracy, which reflects the fault tolerance and self-correction ability of the method itself.
【学位授予单位】：青岛理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TG659

【参考文献】