基于DSP和小波分析的转子故障检测方法研究

发布时间：2019-01-17 09:52

【摘要】：转子作为旋转机械的核心部件,对其进行故障检测是非常重要的。转子故障检测是一门综合多学科的技术。要对转子进行故障检测,必须进行信号的采集、处理和特征提取。转子故障检测主要依赖振动信号,转子运行中产生的振动信号包含了丰富的信息。转子振动信号由于受其它机械设备周期性振动及干扰信号的影响必须进行降噪。本文以转子作为研究对象,选取转子的典型故障作为例证,对其进行信号采集、分析和特征提取。本文主要工作如下：采用DSP TMS320F2812处理器作为平台来设计转子信号采集系统。介绍了TMS320F2812和AD7606-4的功能结构及GPIO功能脚的设置,详细阐述了它们之间引脚的连接方法和作用,以及采集系统的软件功能。讲述了并行模式下AD7606-4四通道同步采样时序图,采用了CPU定时器0中断,控制A/D转换器启停。通过改变分频系数的寄存器和预定计数常数的寄存器的值,实现改变采样频率。采用Bayes样本估计法获取小波阈值,利用软阈值函数对每层小波分解的细节系数进行处理,并逐层重构处理后的系数获得降噪后的信号,得到了一种实用的、具有自适应能力的转子振动信号的小波降噪新方法。将Bayes阈值小波降噪处理结果与Donoho阈值法、Penalty阈值法、Birge-Massart阈值法的小波降噪处理结果进行比较,结合一些仿真和实例的验证,表明使用此方法处理的信号较好地保留了原信号的细节部分,信噪比也有明显提高。转子特征提取是转子检测的一个重要环节。研究了小波频带能量、小波包络两种特征量提取方法。用仿真的方法验证了应用Mallat算法的正交小波分解频率范围的计算公式。对转子信号小波分解,计算其各个频带的能量,对其进行归一化处理作为特征量。在小波分解的故障频带内采用Hilbert包络谱提取故障特征频率值及其频谱获取特征信号。根据特征向量判断转子故障。
[Abstract]:As the core component of rotating machinery, it is very important to detect its faults. Rotor fault detection is a multi-disciplinary technology. Signal acquisition, processing and feature extraction are necessary for rotor fault detection. Rotor fault detection mainly depends on vibration signal, which contains abundant information. The rotor vibration signal must be de-noised because of the periodic vibration and interference signal of other mechanical equipment. In this paper, the rotor is taken as the research object, the typical fault of the rotor is selected as an example, and the signal collection, analysis and feature extraction are carried out. The main work of this paper is as follows: using DSP TMS320F2812 processor as the platform to design the rotor signal acquisition system. The functional structure of TMS320F2812 and AD7606-4 and the setting of GPIO functional pin are introduced. The connecting method and function of pin between them and the software function of acquisition system are described in detail. The AD7606-4 four channel synchronous sampling sequence diagram in parallel mode is described. The CPU timer 0 interrupt is used to control the start and stop of the A / D converter. The sampling frequency is changed by changing the value of the register of the division coefficient and the register of the predetermined counting constant. The wavelet threshold is obtained by using Bayes sample estimation method, and the detail coefficients of each wavelet decomposition are processed by soft threshold function, and the de-noised signal is obtained by reconstructing the coefficients layer by layer, and a practical method is obtained. A new wavelet denoising method for rotor vibration signal with adaptive capability. The results of Bayes threshold wavelet denoising are compared with those of Donoho threshold method, Penalty threshold method and Birge-Massart threshold method. It shows that the signal processed by this method retains the details of the original signal, and the signal-to-noise ratio (SNR) is improved obviously. Rotor feature extraction is an important part of rotor detection. Wavelet band energy and wavelet envelope extraction methods are studied. The calculation formula of the frequency range of orthogonal wavelet decomposition using Mallat algorithm is verified by simulation. The rotor signal is decomposed by wavelet, the energy of each frequency band is calculated, and the signal is normalized as the eigenvalue. In the fault frequency band of wavelet decomposition, the Hilbert envelope spectrum is used to extract the characteristic frequency value of the fault and the spectrum to obtain the characteristic signal. The rotor fault is judged by eigenvector.
【学位授予单位】：华中农业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】