声信号分析方法在重载货运列车滚动轴承故障诊断中的应用研究
发布时间：2018-08-25 10:54
【摘要】：货运列车滚动轴承故障诊断技术是一门紧密结合实际的工程科学,生产实际的需要是它发展的根本原因,因此研究简便的诊断方法具有广阔的实用价值。本文从货运列车滚动轴承噪声产生和传播的理论分析入手,综合考虑货运列车滚动轴承声音的衰减和背景噪声的影响,针对货运列车滚动轴承声信号非平稳性这一特点,采用现代信号分析技术对声信号进行处理和故障识别,以提高其诊断识别的有效性和可靠性。本文主要研究以下几个方面： 针对货运列车滚动轴承声信号的非平稳性,本文利用小波变换的多分辨率性质来分析声信号,发现冲击成分在小波分解的细节信号中得到放大,对比该频率和各种故障下形成地故障频率找到故障的原因,从而实现对信号波形有效地识别。本文提出一种基于非线性小波变换的去噪方法—分层阈值去噪算法。仿真结果表明,该方法能显著提高滤波精度,在有效去除噪声的同时,能很好地保留信号的主要细节。然后通过对经小波变换后的信号进行自功率谱密度分析,仿真结果表明,基于小波变换的自功率谱密度分析能有效地提取货运列车滚动轴承故障声信号的特征频率,适合声信号这样的非平稳信号的分析与研究。在特征提取方面,本文又提出了一种新方法—基于频段局部能量的区间小波包特征提取,它可以根据需要细分各个频带。经实践验证这些特征因子可以很好地代表滚动轴承的工作状况。 研究了神经网络在货运列车滚动轴承智能诊断方面的应用。基于声音信号分析的故障特征提取方法很多,但每种方法都只在某一方面反映了故障特点,单独应用诊断效果不是很好。本文通过对比以不同方法提取的故障特征组合作为神经网络的输入,最终确定利用小波分析和神经网络相结合的方法对货运列车滚动轴承进行故障诊断。利用神经网络进行滚动轴承故障诊断,可以降低对操作人员的专业知识要求,将故障诊断从传统方法转向人工智能方向。同时,诊断系统中智能技术的应用能大大降低维修人员工作压力。
[Abstract]:The fault diagnosis technology of rolling bearing of freight train is an engineering science which is closely combined with practice. The basic reason of its development is the need of production. Therefore, the research of simple diagnosis method has broad practical value. This paper starts with the theoretical analysis of the noise generation and propagation of rolling bearing of freight train, synthetically considering the influence of sound attenuation and background noise of rolling bearing on freight train, aiming at the non-stationary sound signal of rolling bearing of freight train. In order to improve the effectiveness and reliability of the diagnosis and identification of acoustic signals, modern signal analysis technology is used to process and identify the acoustic signals. This paper mainly studies the following aspects: aiming at the non-stationarity of the sound signal of the rolling bearing of freight train, this paper uses the multi-resolution property of the wavelet transform to analyze the acoustic signal. It is found that the shock component is amplified in the detail signal of wavelet decomposition, and the reason of the fault is found by comparing the frequency with the fault frequency formed under various kinds of faults, so that the waveform of the signal can be effectively recognized. In this paper, a hierarchical threshold denoising method based on nonlinear wavelet transform is proposed. The simulation results show that the proposed method can significantly improve the filtering accuracy and preserve the main details of the signal at the same time of removing noise effectively. Then the self-power spectrum density analysis of the signal after wavelet transform is carried out. The simulation results show that the self-power spectrum density analysis based on wavelet transform can effectively extract the characteristic frequency of the fault acoustic signal of rolling bearing of freight train. It is suitable for the analysis and research of nonstationary signals such as acoustic signals. In the aspect of feature extraction, a new method of feature extraction based on local energy of frequency band is proposed in this paper, which can subdivide every frequency band according to the need. It is proved by practice that these characteristic factors can well represent the working condition of rolling bearings. The application of neural network in intelligent diagnosis of rolling bearing of freight train is studied. There are a lot of fault feature extraction methods based on sound signal analysis, but each method only reflects the fault characteristics in one aspect, and the diagnosis effect is not very good. In this paper, the fault features extracted by different methods are compared as the input of neural network, and the method of combining wavelet analysis and neural network is used to diagnose the fault of rolling bearing of freight train. The fault diagnosis of rolling bearing based on neural network can reduce the requirement of professional knowledge for the operator and turn the fault diagnosis from traditional method to artificial intelligence. At the same time, the application of intelligent technology in diagnosis system can greatly reduce the working pressure of maintainers.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TN912.3;TH165.3
本文编号：2202665
[Abstract]:The fault diagnosis technology of rolling bearing of freight train is an engineering science which is closely combined with practice. The basic reason of its development is the need of production. Therefore, the research of simple diagnosis method has broad practical value. This paper starts with the theoretical analysis of the noise generation and propagation of rolling bearing of freight train, synthetically considering the influence of sound attenuation and background noise of rolling bearing on freight train, aiming at the non-stationary sound signal of rolling bearing of freight train. In order to improve the effectiveness and reliability of the diagnosis and identification of acoustic signals, modern signal analysis technology is used to process and identify the acoustic signals. This paper mainly studies the following aspects: aiming at the non-stationarity of the sound signal of the rolling bearing of freight train, this paper uses the multi-resolution property of the wavelet transform to analyze the acoustic signal. It is found that the shock component is amplified in the detail signal of wavelet decomposition, and the reason of the fault is found by comparing the frequency with the fault frequency formed under various kinds of faults, so that the waveform of the signal can be effectively recognized. In this paper, a hierarchical threshold denoising method based on nonlinear wavelet transform is proposed. The simulation results show that the proposed method can significantly improve the filtering accuracy and preserve the main details of the signal at the same time of removing noise effectively. Then the self-power spectrum density analysis of the signal after wavelet transform is carried out. The simulation results show that the self-power spectrum density analysis based on wavelet transform can effectively extract the characteristic frequency of the fault acoustic signal of rolling bearing of freight train. It is suitable for the analysis and research of nonstationary signals such as acoustic signals. In the aspect of feature extraction, a new method of feature extraction based on local energy of frequency band is proposed in this paper, which can subdivide every frequency band according to the need. It is proved by practice that these characteristic factors can well represent the working condition of rolling bearings. The application of neural network in intelligent diagnosis of rolling bearing of freight train is studied. There are a lot of fault feature extraction methods based on sound signal analysis, but each method only reflects the fault characteristics in one aspect, and the diagnosis effect is not very good. In this paper, the fault features extracted by different methods are compared as the input of neural network, and the method of combining wavelet analysis and neural network is used to diagnose the fault of rolling bearing of freight train. The fault diagnosis of rolling bearing based on neural network can reduce the requirement of professional knowledge for the operator and turn the fault diagnosis from traditional method to artificial intelligence. At the same time, the application of intelligent technology in diagnosis system can greatly reduce the working pressure of maintainers.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TN912.3;TH165.3
【参考文献】
相关期刊论文 前10条
1 张武军,徐金梧,杨德斌,周艳玲,王海峰;声信号分析方法在轴承故障诊断中的应用[J];北京科技大学学报;2002年01期
2 唐德尧,王定晓,杨政明,宋辛辉,王巍松;共振解调技术与机车车辆传动装置故障诊断[J];电力机车技术;2002年05期
3 王铁,张国忠,侯荣涛;基于倒谱和小波变换的驱动桥故障特征提取[J];计算机测量与控制;2003年08期
4 刘庆升,徐霄鹏,黄文浩;一种语音端点检测方法的探究[J];计算机工程;2003年03期
5 莫琦,陈立定,冯太合;小波包在旋转机械故障振动信号处理中的应用[J];计算技术与自动化;2005年02期
6 许陇云,卞青青,朱善安,郑剑锋,张慧才,金凤钊;滚动轴承异音的数学模型和评估参数[J];机械工程学报;2002年01期
7 石华峰,尹国华,贺旭军,叶秋文;机车轴承故障诊断[J];机车电传动;2004年02期
8 桂普江,林建中;滚动轴承故障诊断的神经网络方法[J];机械;2004年10期
9 陈四根;基于熵函数的语音端点检测方法[J];声学与电子工程;2001年01期
10 张生玉,冯庚斌,段全胜,丁福焰;JL-601机车走行部检测系统[J];铁道机车车辆;2000年05期
相关硕士学位论文 前1条
1 崔宝珍;基于小波分析的滚动轴承故障诊断方法研究与应用[D];中北大学;2005年
，本文编号：2202665
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2202665.html
