全信息小波分析及设备诊断工程应用研究

【作者】 冯彩红；

【导师】 韩捷；

【作者基本信息】 郑州大学 ， 机械电子工程， 2006， 硕士

【摘要】 转子系统中振动信息通常由相互垂直的两个探头提取，而传统的信息处理方法大多是以单通道信号来进行分析。在同一截面的不同位置测得的振动信息可能存在很大的差异，因此传统分析方法不能全面反映该截面信息。而用全信息技术对转子同一截面的双通道信号进行融合，可以实现转子运转信息的全方位描述。全信息技术之一的全矢谱技术能全面地确定转子振动的频谱结构和强度，还可以得到转子的进动方向和相位角等诊断信息。小波分析具有时频局部化特征，能够实现信号在不同频带、不同时刻的合理分离。基于此，本文结合全矢谱技术和小波分析技术的优点，提出了全信息小波分析技术。主要工作如下： 1．将全矢谱技术和小波分析技术相结合，提出全信息小波分析技术，该方法首先根据故障信号的特点，选择适当的小波函数和分解层数，将两通道的信号分别进行分解，将包含特征频率的分解信号进行单支重构，然后用全矢谱技术将两通道的对应重构信号进行融合。并以转子支承系统松动故障为例进行实验研究，实验结果表明，该方法不但能够实现故障特征频率的分离，而且融合了转子同源两通道的信息，能比较全面地反映转子的振动特征。 2．小波分析虽然可以对信号进行有效的时频分解，但在高频频段其频率分辨率较差，而在低频频段其时间分辨率较差。而小波包分析将频带进行多层次划分，对多分辨分析没有细分的高频部分进一步分解，提高了时频分辨率。因此，本文在全信息小波分析的基础上，又提出了全信息小波包分析技术。并以油膜涡动故障为例进行实验研究，该方法成功地提取了信号中的微弱频率成分，并且融合了双通道的信息，保证了信息量的完整，这充分说明了该方法的有效性和实用性。 3．不论是小波分析，还是小波包分析，都存在一个问题：小波的构造与所分析的信号无关，不能保证有效地匹配信号的特征。而在小波分析基础上提出的第二代小波变换，用提升方法对一个已知小波进行修正，可以获得具有某种期望特性的新的小波基函数，很好地克服了小波变换的不足。为此，本文在前面研究的基础上，将全信息小波(小波包)分析技术的思想推广到第二代小波变换领域，提出了全信息第二代小波变换技术，该技术能够根据信号的特征构造新小波，又能更多还原

【Abstract】 The vibration information of rotor is picked up from two channels, but the traditional information processing is mostly based on the signal coming from single channel. The vibration information picked up from different channels may be completely different, thus the traditional analysis method cannot integrally embody the characters of this section. Two channels information of one rotor’s section are fused by full information technology, which can integrally describe the vibration characters of this section. Vector spectrum, one of full information technology, can fully confirm the spectrum structure and vibration intension, and can receive information of precession direction and phase characters. Wavelet analysis has the character of time-frequency location, and can make signal separate at different frequency and at different time. Based on this, the merits of vector spectrum and wavelet analysis are united and Full Information Wavelet Analysis (FIWA) is proposed. The main content in this article is as follows:1. The vector spectrum is combined with wavelet analysis thus the new theories of FIWA is proposed in this article. According to the fault character, proper wavelet basis and decomposed level is selected, and dual signals are respectively decomposed, and domains that contain character frequency are reconstructed, the corresponding reconstructed signals of two channels are fused by vector spectrum technology. Loose coupling fault signal is employed to experiment, results indicate: FIWA can not only separate different character frequency but also fuse two channels’ information and acquire the united vibration characters.2. Though wavelet analysis has property of time-frequency location and can decompose signal effectively, but the frequency resolution is poor at high frequencies and the time resolutions is poor at low frequencies. Wavelet packet analysis offer a more complex and flexible analysis, the details as well as the approximations are split. Thus wavelet packet analysis can improve the poor frequency resolution at high frequencies. So based on FIWA, Full Information Wavelet Packet Analysis (FIWPA) is also proposed. Oil whirl fault signal is applied to study, weak vibration signals are extracted successfully, and two channels information are fused. Study results show FIWPA is effective and usable.3. Whether wavelet analysis or wavelet packet analysis has a problem: The constructing of wavelet is irrelevant to the analyzed signal and can not effectively更多还原