齿轮箱复合故障诊断特征提取的若干方法研究

发布时间：2019-06-08 12:35

【摘要】：在机械设备中齿轮箱是最重要的动力传动部件，其健康状况直接影响着机械设备能否正常工作，若能准确的预测故障的位置，就可以有效的避免故障所带来的巨大人力和财力损失，因此研究新型复合故障诊断方法对齿轮箱的正常运行具有举足轻重的作用。通过振动加速度传感器所采集到的振动信号通常是非平稳信号，尤其是在工作现场采集到的信号更是受到各种背景噪声的干扰，导致微弱故障特征经常被噪声所淹没。此外，当齿轮箱出现故障时，往往产生了位置不同、形式不同、程度不同的复合故障，每个故障之间相互干扰、相互影响、相互耦合。尤其是在强背景噪声条件下，微弱故障还极易被噪声淹没，从而给故障诊断带来了挑战。因此对强背景噪声下复合故障进行诊断是当今的技术难点。本论文针对以上问题，在国家自然基金（50775157）；山西省基础研究项目(2012011012-1)；山西省高等学校留学回国人员科研资助项目(2011-12)的资助下，把齿轮箱作为研究对象，以近几年比较新的降噪方法作为研究手段，同时以齿轮箱复合故障作为研究目标，对强背景噪声环境下，从复合故障振动信号中准确的提取故障特征信息，进一步对故障特征进行分离进行了深入的研究。论文主要研究结论如下： (1)用EEMD(Ensemble Empirical Mode Decomposition)对强噪声的多调制源多载波频率的仿真信号进行分解，发现单一的白噪声幅值直接影响着EEMD的分解效率。针对这个问题，论文提出了CMF(Combined ModeFunction),即将EMD(Empirical Mode Decomposition)分解得到的与原信号相关性较强的IMFs按高低频进行叠加，形成两个新的组合模态函数ch和c L，然后通过添加不同的白噪声幅值对c h和cL分别进行EEMD分解，最后对敏感的IMFs分别进行循环自相关函数解调分析，将提出的方法应用于仿真信号和复合故障齿轮箱试验台，成功提取了多故障特征，验证了此方法的有效性。 (2)针对强噪声环境下滚动轴承故障信号微弱、故障特征难以提取等问题，本文提出基于最小熵反褶积（Minimum entropy deconvolution，MED）和EEMD相结合的方法来提取复合故障中滚动轴承微弱故障特征。通过对仿真信号分析发现：在强背景噪声下EEMD对微弱信号的特征提取具有很大的局限性。为了剔除噪声干扰，提取微弱故障的特征信息，本文选取MED作为EEMD的前置滤波器，验证了其强大的降噪功能。同时将MED与EEMD相结合的方法用于复合故障的微弱故障特征提取，即先用MED对强背景噪声下风电齿轮箱试验台进行降噪处理，然后再对降噪后的信号进行EEMD，最后对敏感的本征模态函数(IMFs)进行循环自相关函数解调分析。这种方法与EEMD进行对比分析，表明了此方法有效性，从而为多故障共存并处于强背景噪声下的微弱特征提取提供了一种新的方法。 (3)循环平稳信号具有非平稳性特点，因此用循环平稳的特征来研究循环统计量是很有必要的。循环二阶谱适用于周期性振动信号，但通过仿真信号发现在强背景噪声下，时域的离散化并没有导致循环自相关函数在循环域内周期化。此外，多载波频率共存或比较接近时在高频处不可避免的出现了混迭现象。 (4)研究了最大相关峭度反褶积（Maximum correlated Kurtosisdeconvolut on，MCKD）的降噪特点，同时对它的参数（位移数、周期和迭代次数）进行了讨论和分析。 (5)针对多调制源、多载波信号的循环自相关函数解调分析存在交叉项的干扰,这使循环自相关函数解调方法的实际应用产生了局限性。本文提出了基于最大相关峭度反褶积（Maximum correlated Kurtosis deconvolution，MCKD）和循环自相关解调方法，先通过MCKD对原信号进行降噪，，以便提取感兴趣的周期成分，再对降噪后的周期信号通过循环自相关解调分析，有效地抑制了多调制源、多载波对循环平稳结果带来的交叉项干扰,提高了分析的可靠性。将该方法运用于复合齿轮箱故障诊断中，成功地从振动信号中分离出故障源。针对旋转机械在强噪声背景下的复合故障诊断是当前机械故障诊断领域的难点。本文以风电齿轮箱为研究对象，对齿轮点蚀、轴承内外圈点蚀等复合故障的振动信号进行分析。通过仿真信号和工程实例表明将EEMD、MED、MCKD、CMF、循环域解调等方法相结合，可以成功提取强背景噪声下复合故障的特征频率，实现了由单一故障到多故障的突破，应用前景广阔。
[Abstract]:in the mechanical equipment, the gear box is the most important power transmission component, the health condition of the gear box directly influences whether the mechanical equipment can work normally, and if the position of the fault is accurately predicted, the huge human and financial loss caused by the failure can be effectively avoided, Therefore, the new compound fault diagnosis method plays a very important role in the normal operation of the gear box. The vibration signal acquired by the vibration acceleration sensor is usually a non-stationary signal, especially the signal acquired at the work site is interfered by various background noise, and the weak fault characteristic is often inundated by the noise. In addition, when the gear box fails, a complex fault with different positions, different forms and different degrees is often generated, and each fault is mutually interfered, influenced and coupled with each other. In particular, in that condition of strong background noise, the weak fault is extremely easy to be inundated with noise, thus posing a challenge for fault diagnosis. Therefore, the diagnosis of composite fault under strong background noise is the difficult point of the present technology. In view of the above problems, in the national natural fund (50775157), the basic research project of Shanxi Province (2012011012-1), the research object of the gear box is taken as the research object under the support of the research and financing project (2011-12) of the research and support project of the university of higher learning in Shanxi Province (2011-12). In recent years, the new noise reduction method is used as the research means, and the compound fault of the gearbox is used as the research target, and the fault characteristic information is accurately extracted from the composite fault vibration signal under the strong background noise environment, and the fault characteristic is further separated. The main conclusions of the paper are as follows: (1) Using EEMD (EEMD) to decompose the multi-carrier frequency of the multi-modulation source with strong noise, it is found that the single white noise amplitude directly affects the division of the EEMD. In order to solve this problem, the paper presents a combined mode function (CMF), and the IMFs with strong correlation with the original signal obtained by the EMD (EMD) decomposition are combined at high and low frequencies to form two new combined modal functions. and finally, carrying out cyclic autocorrelation function demodulation analysis on the sensitive IMFs, The method is verified by the barrier feature The paper presents a method of combining minimum entropy deconvolution (MED) and EEMD to extract the micro-fault of the rolling bearing in the composite fault with the method of combination of minimum entropy deconvolution (MED) and EEMD. The analysis of the simulation signal shows that the extraction of the weak signal by the EEMD is very important in the strong background noise. In order to eliminate the noise interference and extract the characteristic information of weak fault, this paper selects MED as the pre-filter of EEMD, and verifies its power. The method for combining the MED and the EEMD is used for the weak fault feature extraction of the composite fault, namely, firstly, performing noise reduction processing on the wind power gearbox test bed under the strong background noise by the MED, and then EEMD for the last pair of sensitive intrinsic mode functions (IMFs) The method is compared with EEMD, which shows the effectiveness of this method, so as to provide a weak feature extraction for multi-fault co-existence and under strong background noise. (3) The circular stationary signal has the characteristics of non-stationarity, so the cycle statistics are studied with the characteristics of the cycle stability The cyclic second-order spectrum is suitable for periodic vibration signals, but it is found that the discretization of the time domain does not cause the cyclic autocorrelation function to be in the strong background noise by the simulation signal. in addition, that multi-carrier frequency is unavoidable at high frequency when the multi-carrier frequency coexist or is close (4) The noise reduction characteristics of the maximum correlation kurtosideconvolon (MCKD) are studied, and its parameters (the number of displacement, the period and the number of iterations) are also studied. (5) For multiple modulation sources, the cyclic autocorrelation function of the multi-carrier signal is used to demodulate and analyze the interference of the cross term, which makes the loop self-correlation function demodulation method In this paper, the maximum correlatedKurtosis (MCKD) and the cyclic autocorrelation demodulation method based on the maximum correlation are proposed in this paper. Firstly, the original signal is denoised by the MCKD, so as to extract the periodic component of interest and the periodic signal after noise reduction. Through the self-correlation demodulation analysis of the cycle, the cross terms of the multi-modulation source and the multi-carrier on the smooth result of the loop are effectively suppressed. The method is used in the fault diagnosis of the compound gear box, and successfully The fault source is separated from the vibration signal. The difficulties in the field of current mechanical fault diagnosis are as follows: the wind power gearbox is used as the research object, the pitting of the gears, the inner and outer ring points of the bearing, etc. The vibration signal of the composite fault is analyzed. By combining the simulation signal and the engineering example, the characteristic frequency of the composite fault under strong background noise can be successfully extracted by combining the methods of EEMD, MED, MCKD, CMF and cyclic domain demodulation.
【学位授予单位】：太原理工大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TH165.3

【参考文献】