【作者】 王婧；

【导师】 崔玲丽；

【作者基本信息】 北京工业大学 ， 机械工程（专业学位）， 2013， 硕士

【摘要】 滚动轴承是工业传动系统重要组成部分，其作用十分关键，进行滚动轴承状态监测和故障诊断对确保生产安全，预防重大事故，降低生产成本有着重要意义。本文重点关注稀疏分解中匹配追踪和基追踪方法在滚动轴承故障诊断中的算法优化及实际应用效果，主要分为以下几个内容：(1)研究了滚动轴承故障信号的组成，特征及滚动轴承故障诊断的依据。介绍了稀疏分解中的基追踪法和匹配追踪方法原理。分析了稀疏分解方法的组成三要素即原子库模型的构造、算法的实现方式、迭代终止条件，并对每种要素的实现方法进行了列举和说明，结合滚动轴承故障信号特征讨论了各个方法的可行性及其特点；(2)在传统建模方式的基础上引入了转速，滚动轴承尺寸参数，故障大小等参数建立了一种能够反映滚动轴承故障状态特征的新型模型，称之为新型冲击字典模型，该模型提升了原子库模型与滚动轴承故障特征信号之间的相似度。仿真与实验信号分析结果表明了该模型的可行性和有效性；(3)基于所建立的新型冲击字典模型结合遗传算法提出了基于新型冲击字典匹配追踪的滚动轴承故障诊断方法，并通过仿真，实验，工程信号分析验证了该方法的可行性，优越性及不足之处；(4)结合新型冲击字典模型提出了一种自适应字典的稀疏分解实现方法。该方法利用新型冲击字典模型，通过逐步选取参数，建立冗余度小，原子匹配程度高的原子库，继而实现快速稀疏分解，称之为自适应冲击字典的稀疏分解实现方法。该方法与基追踪和匹配追踪相结合在仿真、实验和工程实际中都取得了很好的效果，并且在准确性、稳定性和快速性方面大大优于基于遗传算法的匹配追踪。进一步的分析结果表明基于自适应冲击字典匹配追踪方法在稳定性、准确性和快速性方面都略优于基于自适应冲击字典基追踪方法；(5)利用LabVIEW软件开发了一套基于稀疏分解方法的滚动轴承故障诊断系统。该系统具有信号采集，信号处理，结果保存等功能。使用方便，操作界面友好，为稀疏分解方法在工程实际中应用提供了一种可行的方案。更多还原

【Abstract】 Rolling bears are integral components of the transmission system have importantrole. The condition monitoring and fault diagnosis on the bearing is of greatsignificance to ensure operational safety, prevent serious accidents and reduceproduction cost.This paper focus on the efficiency analysis, optimization and application ofsparse decomposition in bearing fault diagnosis, is divided into several parts：(1) Study the compositions, characters of fault bearing signal and the principle ofbearing fault diagnosis. Than introduce the principle of basic pursuit and matchingpursuit algorithm. Analysis the three essential factors of sparse decomposition anddescription every factor in detail (The establishment of dictionary model, Theimplementation methods and the iteration termination) Feasibility analysis of sparesdecomposition applied to bearing fault diagnosis is carried out combing with bearingfault signal character.(2) A new dictionary model is established according to the characteristics andmechanism of bearing faults based on traditional impulse dictionary. The new modelincorporates the rotational speed of the bearing, dimension of the bearing, and bearingfault status, among other parameters be called New Impulse Dictionary model. Thismodel optimized the accuracy of traditional dictionary. And the comparison ofexperiment and simulation verified the accuracy and effectiveness of New ImpulseDictionary model.(3) Established a matching pursuit algorithm using new impulse dictionarycombining with genetic algorithm. Simulation and experiment suggest that newimpulse dictionary used in matching pursuit algorithm combining with geneticalgorithm have more accurate effect on bearing fault diagnosis than using traditionalimpulse dictionary. But it still has some weak points of poor stability, rapidity andcontrollability by using genetic algorithm.(4) The self-adaptive impulse dictionary is established based on new impulsedictionary. The basic pursuit and matching pursuit algorithm of self-adaptive impulsedictionary is adopted to analyze the simulated experiment and practice signals. Theresults show that basic pursuit and matching pursuit algorithm of self-adaptiveimpulse dictionary have better effect on bearing fault diagnosis than matching pursuitalgorithm combining with genetic algorithm.(5) A fault diagnosis system based on sparse decomposition for bearing fault isdeveloped by using LabVIEW. The system contained several functions of signalacquisition, signal processing and signal storing etc. The structure is reasonable andthe interface is friendly. Provide a feasible scheme for bearing fault diagnosis by using sparse decomposition.更多还原