【作者】 王正英；

【导师】 王太勇；

【作者基本信息】 天津大学 ， 机械制造及其自动化， 2007， 博士

【摘要】 故障的发生和发展、设备工况的变化以,使机械设备的动态信号表现出非平稳性和强噪声特性,给故障诊断带来困难。本论文以机电设备为对象,对复杂振动信号中的时频特征提取、净化及回转机械的智能诊断方法进行了研究。针对机电设备工况信号中的非平稳特征,提出了基于SVD降噪算法对原始振动信号进行预处理,然后再进行EMD分解,抑制了异常数据或高频噪声产生的影响,分解所得基本模式分量更能反应设备运行状态的本质特征。研究了基于Hilbert变换的相位解调算法的不足,提出了基于EMD的相位解调方法,仿真数据验证了该方法的有效性;基于工程实践,指出了HHT在工程应用中的两个局限性。针对机电设备运行中的信号混合特性,研究了信号混合的数学模型和常用的信息分离方法,提出将盲源分离方法和Hilbert-Huang变换相结合,实现了旋转机械转子系统非平稳故障特征的独立化提取,故障诊断案例明该方法的有效性和实用性。研究了机械振动信号变换域盲源分离的可行性,提出了机械振动信号频域内的独立分量提取方法,涡流传感器失效故障诊断和早期碰摩故障诊断的成功应用,表明了该方法的有效性,结合设备振动的先验知识可以准确判定各独立分量的物理内涵,提高诊断信息的质量。针对汽轮机智能诊断中知识表达的不确定性问题,在研究了基于联合树的贝叶斯网络精确推理方法和贝叶斯网络的学习算法的基础上,结合机械故障特有的表现形式,提出了基于贝叶斯网络的汽轮机故障诊断方法,并建立了基于贝叶斯网络的机械故障诊断模型,通过对一些设备的实测数据进行了故障诊断分析,验证了该模型的有效性。研究了基于振动信息的自动诊断方法,建立了故障模式类的概念,根据模糊聚类算法确定了故障标准模式类及其频谱特征,为在故障模式类层次上的识别提供了理论基础;研究了模糊关系诊断结果的分布规律、多征兆模糊产生式规则的诊断知识表示以及规则结论的可信度组合,工程诊断实例验证了其可靠性。最后,提出了监测诊断与设备管理信息系统的集成化设计思想,给出了系统框架和技术方案,并在此基础上开发出了“基于网络和状态监测的设备管理信息化系统”,消除了状态监测和设备管理之间的“信息孤岛”现象。该系统在天津大港广安津能发电有限责任公司得到应用,并顺利通过了天津市科委的验收。更多还原

【Abstract】 The dynamic signals of mechanical equipment often possess nonstationarities and strong noise due to occurrence of fault, variance of operation and inherent nonlinearity of equipment, which brings difficulties to fault diagnosis. Aiming at the electric- mechanical equipment, this dissertation focuses on the theroy and applications of time-frequency feature extraction for complex signals and intelligent fault diagnosis mentod for rotating machinary.Aiming at the non-stationary signals in engineering, the original vibration signal was preprocessed by singular value decomposition (SVD) to reduced noise, the influence induced by singularity data or high frequency noise was restrained, and the cleaning signal was decomposed by EMD to extract the intrinsic mode functions (IMFs). The results show that SVD can effectively increase the signal noise ratio (SNR) and emphasize the fault characteristic of the original vibration signal, the IMFs extracted from the denoised signal have clear physical meaning and will increase the precision of fault information. The insufficiency of phase demodulation method based on Hilbert transformation was researched, and a new demodulation method based on EMD was introduced . Two limitations of this method when used in fault diagnosis of mechanical equipments are proposed.Aiming at the mixing signals in engineering measurement, mixing model and separation method in common use are summarized. The combined ICA-HHT method is applied in the sensor failure detection and incipient fault diagnosis of rotor system. The results show that this method can extract the fault information from mixed vibration signals, and distributed in the HHT spectrum. The feasibility of the blind separation for mechanical vibration signals in transformation domain is demonstrated. Successful applications of BSS are achieved in the detection of eddy-current sensor failure and the diagnosis of incipient impact-rub fault. The results show that BSS has widely prospect for application in the condition monitoring and fault diagnosis of mechanical equipment, and transcendental knowledge of equipment’s vibration are helpful for us to analyse the independent components.In order to achieve the intelligent fault diagnosis, aiming at some common vibration failure in rotary machine, vibration features were described and the basic theory on Bayesian Network was elaborated. The accurate reference approaches based on junction trees and some learning methods were discussed, and the mechanical failure diagnosis model based on Bayesian Network was established. The result of an application of this model in engineering shows the validity of this method.The automatic diagnosis mehtod based on vibration information was proposed. The acduisition of spectrum feature of fault pattern class by the fuzzy clustering algotithm is the bisis of pattern recognition in the level of fault class. The result acduired by fuzzy relationship between vibration symptom and fault shows the feature of fuzzy-relation-based method. A kind of fuzzy fault diagnosis expert system of multiple symptoms based on Clips was introduced. The expression of multiple symptom fuzzy diagnosis knowledge and subordinate degree of rules was discussed.Finally, the precept, the system truss and prototype of“Plant Management Information System based on Internet and Condition Monitoring(PMIS)”are proposed to eliminate the“information island”between the plant management and condition monitoring. the integration technologies between different modules and different systems are discussed. A perfect PMIS is built and applied in the Dagang Power Plant.更多还原