本文選題：齒輪故障診斷 + 隱馬爾科夫模型　； 參考：《南昌航空大學(xué)》2014年碩士論文

【摘要】：齒輪傳動在機(jī)械設(shè)備中應(yīng)用廣泛，其運(yùn)行狀態(tài)的好壞直接決定了整個設(shè)備的性能，因此對齒輪運(yùn)行狀態(tài)進(jìn)行在線監(jiān)測和故障診斷尤為重要。齒輪傳動是以輪齒的周期性嚙合傳遞運(yùn)動，這一過程必將產(chǎn)生機(jī)械振動，當(dāng)齒輪出現(xiàn)制造裝配誤差、磨損、裂紋等故障與缺陷時，必然會使機(jī)械振動產(chǎn)生不同的變化，其嚙合振動信號中包含了豐富的齒輪狀態(tài)信息，因此，分析齒輪嚙合振動信號是齒輪故障診斷最有效的方法。 在齒輪的使用中，希望能夠及早的發(fā)現(xiàn)故障并對故障做出診斷，就可以合理的使用齒輪或制定維修計(jì)劃，提高齒輪的利用率，防止生產(chǎn)事故的發(fā)生。為此，本文以齒輪為分析對象，，采用理論分析與實(shí)驗(yàn)研究相結(jié)合，深入研究了基于隱馬爾科夫模型（HMM：Hidden Markov Model）的齒輪故障診斷方法與技術(shù)，主要做了以下幾個方面的工作。 1.分析了齒輪故障診斷的研究意義，綜述了齒輪故障診斷技術(shù)的發(fā)展與HMM在故障診斷中的應(yīng)用，闡述了齒輪故障模式，常見故障的振動機(jī)理與齒輪嚙合振動信號的特征。 2.研究了HMM的基本理論，重點(diǎn)討論了連續(xù)隱馬爾科夫模型（CHMM：Continuous HMM）的理論，針對算法下溢與模型參數(shù)的初始化這兩個實(shí)際應(yīng)用中出現(xiàn)的問題，給出了解決方案。最后給出基于HMM的齒輪故障診斷的思路與流程。 3.提出了基于細(xì)化譜分析的齒輪故障特征提取方法，并將其應(yīng)用在離散隱馬爾科夫模型（DHMM：Discrete HMM）中，該方法首先利用時域同步平均提取目標(biāo)齒輪的振動信號，再進(jìn)行細(xì)化譜分析提取主要頻率及其附近的邊頻帶幅值作為特征向量，量化后輸入到模型中進(jìn)行訓(xùn)練和分類。通過實(shí)驗(yàn)驗(yàn)證了該方法的有效性。 4.應(yīng)用CHMM結(jié)合AR系數(shù)的特征提取方法，進(jìn)行了齒輪故障診斷與齒輪的狀態(tài)識別。在齒輪狀態(tài)識別的研究中，進(jìn)行了齒輪的全生命周期實(shí)驗(yàn)，采用交叉驗(yàn)證尋找最優(yōu)狀態(tài)數(shù)，并用K均值聚類算法對模型進(jìn)行狀態(tài)初始化，成功的對生命周期三個階段進(jìn)行了識別，為齒輪箱的狀態(tài)監(jiān)測提供了科學(xué)依據(jù)。
[Abstract]:Gear transmission is widely used in mechanical equipment, and its running state directly determines the performance of the whole equipment. Therefore, it is very important to monitor and diagnose the running state of gear on line. Gear transmission is transmitted by periodic meshing of gear teeth, which will produce mechanical vibration. When the gear has faults and defects such as assembly error, wear, crack and so on, it will inevitably cause different changes in mechanical vibration. The meshing vibration signal contains abundant information of gear state. Therefore, the analysis of gear meshing vibration signal is the most effective method for gear fault diagnosis. In the use of gears, it is hoped that the faults can be detected and diagnosed as soon as possible, so that the gears can be reasonably used or maintenance plans can be formulated, the utilization ratio of gears can be improved, and the occurrence of production accidents can be prevented. In this paper, the gear fault diagnosis method and technology based on Hidden Markov Model (hmm: Hidden Markov Model) are studied by combining theoretical analysis with experimental research. The main work is as follows. 1. The research significance of gear fault diagnosis is analyzed. The development of gear fault diagnosis technology and the application of HMM in fault diagnosis are summarized. The gear fault mode, the vibration mechanism of common faults and the characteristics of gear meshing vibration signal are expounded. 2. In this paper, the basic theory of HMM is studied, and the theory of continuous Hidden Markov Model (CHM: continuous HMMM) is discussed. The solutions to the problems of algorithm overflow and the initialization of model parameters are given. Finally, the train of thought and flow chart of gear fault diagnosis based on HMM are given. 3. A gear fault feature extraction method based on thinning spectrum analysis is proposed and applied to discrete Hidden Markov Model (DHMM: discrete HMMM). Firstly, the vibration signal of the target gear is extracted by time-domain synchronous averaging. Then the main frequency and the amplitude of the edge band are extracted as the eigenvector by the thinning spectrum analysis, and then the quantization is input into the model for training and classification. The effectiveness of the method is verified by experiments. 4. Gear fault diagnosis and gear state recognition are carried out by using CHMM and AR coefficient feature extraction method. In the research of gear state recognition, the whole life cycle experiment of gear is carried out, the optimal state number is found by cross validation, and the state of the model is initialized with K-means clustering algorithm. The three stages of the life cycle are identified successfully, which provides scientific basis for the condition monitoring of the gearbox.
【學(xué)位授予單位】：南昌航空大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TH132.41;TH165.3

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 李葦營，易克初，胡征;神經(jīng)網(wǎng)絡(luò)與HMM構(gòu)成的混合網(wǎng)絡(luò)在語音識別中應(yīng)用的研究[J];電子學(xué)報;1994年10期

2 金鑫;任獻(xiàn)彬;周亮;;智能故障診斷技術(shù)研究綜述[J];國外電子測量技術(shù);2009年07期

3 童進(jìn),吳昭同,嚴(yán)拱標(biāo);大型旋轉(zhuǎn)機(jī)械升降速過程故障診斷HMM方法研究[J];機(jī)械科學(xué)與技術(shù);1999年06期

4 于德介,楊宇,程軍圣;一種基于SVM和EMD的齒輪故障診斷方法[J];機(jī)械工程學(xué)報;2005年01期

5 李浩;董辛e

本文編號：1981176