滚动轴承故障特征提取与诊断方法研究
发布时间：2019-01-03 10:49
【摘要】：随着科学技术的迅猛发展和现代化大生产的日益普及,旋转机械不断朝着大型化、复杂化、高速化和自动化方向发展,这对设备的运行安全提出了更高的要求。滚动轴承作为旋转机械中应用最为广泛的部件之一,直接决定着整个机械系统能否正常可靠运行,深入开展滚动轴承故障诊断和状态检测技术的研究,对有效避免生产中重大事故的发生,具有重要的学术意义和工程应用价值。本文在详细论述滚动轴承故障信号降噪、特征提取、复合与智能故障诊断研究现状的基础上,从振动信号分析与处理方法着手,针对滚动轴承故障特征提取与诊断中所涉及的几个关键问题进行了深入研究,在滚动轴承故障特征提取、微弱故障诊断、复合故障特征分离、故障模式智能识别和运行状态检测方面取得了一些研究成果,论文的创新点及主要工作如下：(1)传感器采集的滚动轴承故障振动信号频率成分比较复杂,在无关频率成分及噪声的干扰下,轴承故障特征常常难以准确提取。针对此问题,本文基于倒谱预白化和奇异值分解重构提出了一种故障特征提取方法。该方法通过倒谱预白化处理轴承故障信号,消除了信号中离散频率成分和谐波分量的干扰；然后进行奇异值分解,并基于奇异值最大差分谱重构信号,有效滤除了信号中的干扰噪声。实验证明,该方法能准确提取滚动轴承的故障特征。(2)在恶劣的工作环境下,滚动轴承振动信号中常混杂有强烈的背景噪声,尤其是故障特征较为微弱时,极易被噪声所掩盖,轴承故障难以诊断。因此,本文基于自适应多尺度自互补Top-Hat变换提出了一种轴承微弱故障诊断方法。形态学自互补Top-Hat变换滤波器处理轴承故障信号时,能够抑制信号中的强背景噪声,并有效增强轴承的故障冲击特征。同时,为达到兼顾抗噪性和信号细节保持性的目的,构建了多尺度形态学滤波器,通过比较不同尺度下滤波信号的故障特征能量比,自适应确定了最优结构元素的尺度。(3)滚动轴承出现复合故障时,在单通道振动信号中轴承不同元件的故障特征彼此混杂,难以分离。为解决此问题,本文基于改进谐波小波包分解提出了一种轴承复合故障特征分离方法。该方法可以根据需要对信号频带进行任意划分,克服了传统谐波小波包分解后子信号个数及带宽范围受二进制划分的缺陷,通过计算子信号中各单一故障信号的权重因子,重构分离出轴承各单一故障信号,有效实现了滚动轴承复合故障特征的分离。(4)针对以故障模式识别与运行状态检测为主要内容的滚动轴承智能诊断问题,本文采用Hermitian小波对轴承信号进行连续小波变换,再结合样本熵理论,提出以时间-小波能量谱样本熵作为特征参数,对轴承智能诊断进行研究。该方法将时间-小波能量谱样本熵作为轴承不同工况下样本信号的特征向量,通过支持向量机分类算法实现了轴承不同故障模式的智能识别。之后将时间-小波能量谱样本熵用于滚动轴承运行状态检测,计算全寿命周期实验数据的时间-小波能量谱样本熵,按照时间顺序排列,绘制出了轴承运行状态曲线,通过判断曲线走势可有效诊断出轴承早期故障的发生。
[Abstract]:With the rapid development of science and technology and the increasing popularity of modern production, the rotating machinery has been developing in the direction of large-scale, complicated, high-speed and automatic. As one of the most widely used parts in the rotating machinery, the rolling bearing directly determines whether the whole mechanical system can operate normally and reliably, and the research of the fault diagnosis and the state detection technology of the rolling bearing is carried out, and the occurrence of a major accident in the production can be effectively avoided, and has important academic significance and engineering application value. In this paper, on the basis of the present situation of noise reduction, feature extraction, compound and intelligent fault diagnosis of rolling bearing fault signal, this paper proceeds from the analysis and processing method of vibration signal, and studies the key problems involved in the feature extraction and diagnosis of rolling bearing. Some research achievements have been made in the fault feature extraction, weak fault diagnosis, compound fault feature separation, fault mode intelligent identification and operation state detection of rolling bearing, and the innovation point and main work of the paper are as follows: (1) The frequency component of the fault vibration signal of the rolling bearing collected by the sensor is more complex, and the fault characteristics of the bearing are often difficult to be extracted accurately under the interference of independent frequency components and noise. In this paper, a fault feature extraction method is proposed based on inverse spectrum pre-whitening and singular value decomposition reconstruction. The method eliminates the interference of the discrete frequency component and the harmonic component in the signal through the cepstrum pre-whitening treatment bearing fault signal, then performs singular value decomposition, and reconstructs the signal based on the singular value maximum difference spectrum, and effectively filters out the interference noise in the signal. The experimental results show that the method can accurately extract the fault features of the rolling bearing. (2) In the severe working environment, the vibration signal of the rolling bearing is often mixed with strong background noise, especially when the fault characteristic is weak, it is very easy to be covered by the noise, and the bearing fault is difficult to diagnose. Therefore, based on the self-adaptive multi-scale self-complementary Top-Hat transformation, a method of bearing weak fault diagnosis is proposed. When the morphology self-complementary Top-Hat transform filter is used to process the bearing fault signal, the strong background noise in the signal can be suppressed, and the fault impact characteristic of the bearing can be effectively enhanced. At the same time, the multi-scale morphological filter is constructed for the purpose of achieving both anti-noise and signal detail retention, and the scale of the optimal structural element is determined by comparing the fault characteristic energy ratio of the filtered signal at different scales. (3) The fault features of different components of bearing in single-channel vibration signal are mixed with each other, and it is difficult to separate. In order to solve this problem, a method for separating a bearing composite fault feature based on improved harmonic wavelet packet decomposition is presented in this paper. According to the method, the signal frequency band can be arbitrarily divided according to needs, the defects that the number of the sub-signals and the bandwidth range of the traditional harmonic wavelet packet decomposition are subjected to binary division are overcome, the weight factors of each single fault signal in the sub-signal are calculated, and the single fault signals of the bearing are reconstructed and separated, and the separation of the composite fault characteristic of the rolling bearing is effectively realized. (4) According to the intelligent diagnosis of rolling bearing with fault pattern recognition and operation state detection as the main content, this paper uses Hermitian wavelet to perform continuous wavelet transform on the bearing signal, and then combines the sample entropy theory, and puts forward the time-wavelet energy spectrum sample entropy as the characteristic parameter. The intelligent diagnosis of bearing is studied. The method takes the time-small-wave energy spectrum sample entropy as the characteristic vector of the sample signal under different working conditions of the bearing, and realizes the intelligent identification of different fault modes of the bearing by supporting the vector machine classification algorithm. then, the time-small-wave energy spectrum sample entropy is used for detecting the running state of the rolling bearing, the time-small-wave energy spectrum sample entropy of the whole life cycle experimental data is calculated, and the running state curve of the bearing is drawn according to the time sequence, and the occurrence of the early fault of the bearing can be effectively diagnosed by judging the trend of the curve.
【学位授予单位】：华北电力大学(北京)
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TH133.33
[Abstract]:With the rapid development of science and technology and the increasing popularity of modern production, the rotating machinery has been developing in the direction of large-scale, complicated, high-speed and automatic. As one of the most widely used parts in the rotating machinery, the rolling bearing directly determines whether the whole mechanical system can operate normally and reliably, and the research of the fault diagnosis and the state detection technology of the rolling bearing is carried out, and the occurrence of a major accident in the production can be effectively avoided, and has important academic significance and engineering application value. In this paper, on the basis of the present situation of noise reduction, feature extraction, compound and intelligent fault diagnosis of rolling bearing fault signal, this paper proceeds from the analysis and processing method of vibration signal, and studies the key problems involved in the feature extraction and diagnosis of rolling bearing. Some research achievements have been made in the fault feature extraction, weak fault diagnosis, compound fault feature separation, fault mode intelligent identification and operation state detection of rolling bearing, and the innovation point and main work of the paper are as follows: (1) The frequency component of the fault vibration signal of the rolling bearing collected by the sensor is more complex, and the fault characteristics of the bearing are often difficult to be extracted accurately under the interference of independent frequency components and noise. In this paper, a fault feature extraction method is proposed based on inverse spectrum pre-whitening and singular value decomposition reconstruction. The method eliminates the interference of the discrete frequency component and the harmonic component in the signal through the cepstrum pre-whitening treatment bearing fault signal, then performs singular value decomposition, and reconstructs the signal based on the singular value maximum difference spectrum, and effectively filters out the interference noise in the signal. The experimental results show that the method can accurately extract the fault features of the rolling bearing. (2) In the severe working environment, the vibration signal of the rolling bearing is often mixed with strong background noise, especially when the fault characteristic is weak, it is very easy to be covered by the noise, and the bearing fault is difficult to diagnose. Therefore, based on the self-adaptive multi-scale self-complementary Top-Hat transformation, a method of bearing weak fault diagnosis is proposed. When the morphology self-complementary Top-Hat transform filter is used to process the bearing fault signal, the strong background noise in the signal can be suppressed, and the fault impact characteristic of the bearing can be effectively enhanced. At the same time, the multi-scale morphological filter is constructed for the purpose of achieving both anti-noise and signal detail retention, and the scale of the optimal structural element is determined by comparing the fault characteristic energy ratio of the filtered signal at different scales. (3) The fault features of different components of bearing in single-channel vibration signal are mixed with each other, and it is difficult to separate. In order to solve this problem, a method for separating a bearing composite fault feature based on improved harmonic wavelet packet decomposition is presented in this paper. According to the method, the signal frequency band can be arbitrarily divided according to needs, the defects that the number of the sub-signals and the bandwidth range of the traditional harmonic wavelet packet decomposition are subjected to binary division are overcome, the weight factors of each single fault signal in the sub-signal are calculated, and the single fault signals of the bearing are reconstructed and separated, and the separation of the composite fault characteristic of the rolling bearing is effectively realized. (4) According to the intelligent diagnosis of rolling bearing with fault pattern recognition and operation state detection as the main content, this paper uses Hermitian wavelet to perform continuous wavelet transform on the bearing signal, and then combines the sample entropy theory, and puts forward the time-wavelet energy spectrum sample entropy as the characteristic parameter. The intelligent diagnosis of bearing is studied. The method takes the time-small-wave energy spectrum sample entropy as the characteristic vector of the sample signal under different working conditions of the bearing, and realizes the intelligent identification of different fault modes of the bearing by supporting the vector machine classification algorithm. then, the time-small-wave energy spectrum sample entropy is used for detecting the running state of the rolling bearing, the time-small-wave energy spectrum sample entropy of the whole life cycle experimental data is calculated, and the running state curve of the bearing is drawn according to the time sequence, and the occurrence of the early fault of the bearing can be effectively diagnosed by judging the trend of the curve.
【学位授予单位】：华北电力大学(北京)
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TH133.33
【相似文献】
相关期刊论文 前10条
1 杨建刚;;旋转机械摩擦故障特征的多重性分析[J];电力设备;2007年10期
2 秦文娟;张骏;宁宁;;基于多类别信息量的模糊Petri网故障推理策略[J];计算机测量与控制;2008年07期
3 梅检民;陈祥龙;肖云魁;杨万成;乔龙;;基于稳健全息阶比谱提取微弱故障特征[J];振动与冲击;2012年12期
4 马辉;孙伟;王学军;闻邦椿;;转子系统松动故障特征分析[J];东北大学学报(自然科学版);2009年03期
5 李锋;王家序;杨荣松;;有监督不相关正交局部保持映射故障辨识[J];仪器仪表学报;2013年05期
6 邵毅敏;周晓君;欧家福;陈再刚;周本学;;增强型滤波及冲击性机械故障特征的提取[J];机械工程学报;2009年04期
7 曲佳;兴成宏;李迎丽;李r，
本文编号：2399270
本文链接：https://www.wllwen.com/jixiegongchenglunwen/2399270.html
