基于全矢谱理念的现代谱分析方法研究

发布时间：2018-01-01 01:34

本文关键词：基于全矢谱理念的现代谱分析方法研究　出处：《郑州大学》2014年硕士论文　论文类型：学位论文

【摘要】：经典谱估计方法，是信号处理学科，以及故障诊断领域中非常重要、非常常用的方法。经典谱估计是基于傅里叶变换的，对于样本数据之外的数据假设为零进行谱估计，得到的谱估计与真实谱并不是一致估计；经典谱估计方法由于受到窗函数的影响，导致处理得到的频谱图，其谱分辨率低，方差性能差。旧方法的不足必然会诱导出新的方法，因此很多专家学者开始研究，分辨率更高的谱估计方法，现代谱估计方法得以出现。对于经典谱估计或者是现代谱估计，都是基于单源信号进行采集处理，便很容易引起误判，造成严重后果。基于同源全信息融合技术的全矢谱分析方法，可以有效对同源双通道信号，进行信息融合，其计算简单、快速且稳定，能够真实地、完全地反应设备当前的运行振动状态。本文结合现代谱估计和全矢谱技术的优点，研究新的故障诊断的方法——全矢最大熵倒谱分析方法，是以全矢谱技术为基础，通过最大熵谱估计进行倒频谱的分析，在故障诊断中应用。主要工作如下： 1.介绍了现代谱估计： AR模型谱估计和最大熵谱估计；介绍了全矢谱技术的基本理论：全矢谱技术不仅可以有效融合双通道数据，，还能够传统的谱估计方法（倒频谱、最大熵等），进行有效地结合提出一些新的方法。 2.研究全矢最大熵倒谱分析方法：根据倒频谱的定义及全矢谱技术的成熟发展应用，结合基于AR模型的最大熵谱估计的高分辨率，研究了全矢最大熵倒谱分析方法，通过实验验证新方法的有效性及相比传统方法新方法得到的谱图具有更高的分辨率。 3.研究最大熵倒谱分析方法其它形式：基于熵倒谱的思路，对于全矢最大熵倒谱分析方法的其它形式进行了深入探讨，给出了基于功率倒频谱的三种全矢最大熵倒谱分析方法的形式及计算公式，通过实验验证了它们的有效性。 4.研究了基于双通道数据的AR模型的阶次选取的实用性问题。AR模型阶次选取是进行谱估计的关键，通过实验探讨了基于双通道数据的AR模型的阶次选取的实用性问题，给出了具有一定参考价值的经验公式。
[Abstract]:The classical spectrum estimation method, signal processing is a very important subject, and the fault diagnosis method is very common in the field. The classical spectrum estimation is based on the Fourier transform, the sample data for data beyond the assumption of zero spectral estimation, the spectrum estimation is not consistent with the real spectrum estimation method; due to the effect of the window function estimation of classical spectrum, spectrum processing result obtained, its spectral resolution is low, the variance of poor performance. The deficiency of the old method will induce new methods, so many experts and scholars began to study, high resolution spectrum estimation method, modern spectral estimation methods to appear. For the classical spectrum estimation and modern spectral estimation. Is the single source based on the signal acquisition and processing, it is very easy to cause the miscarriage of justice, causing serious consequences. The full vector spectrum analysis method based on homologous information fusion technology, can be effective for homologous Dual channel signal, information fusion, its calculation is simple, fast and stable, can truly and completely run vibration state of the equipment at present. Based on modern spectral estimation and full vector spectrum technology advantages - vector maximum entropy cepstrum analysis method research on fault diagnosis of the new method, is based on the full vector spectrum technology as the foundation, the maximum entropy spectral analysis cepstrum estimation used in fault diagnosis. The main work is as follows:
The 1. introduces the modern spectral estimation: AR model spectrum estimation and maximum entropy spectral estimation; introduces the basic theory of the full vector spectrum technology: vector spectrum technology can not only effective integration of dual channel data, but also to the traditional spectrum estimation method (cepstrum, maximum entropy, etc.) are effectively combined to put forward some new methods.
2. of the maximum entropy vector cepstrum analysis method: according to the definition of the cepstrum and mature development and application of the full vector spectrum technology with high resolution, maximum entropy spectrum estimation based on AR model, the maximum entropy vector cepstrum analysis method is verified by the experiment spectra of new method to get the validity of the new approach and compared with the traditional the method has higher resolution.
3. of the maximum entropy cepstrum analysis method in other forms: entropy cepstrum based on the idea of the other form of maximum entropy vector cepstrum analysis method are discussed, and calculation formula of power cepstrum three vector maximum entropy cepstrum analysis method is proposed based on the experimental results validate the effectiveness of them.
4. AR model of dual channel data of the order selection of practical problems of.AR model order selection is a key spectrum estimation based on the experiment data of the dual channel AR model order selection problem of utility based on the proposed has a certain reference value for the empirical formula.

【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TH165.3

【参考文献】