基于小波分析提取风机噪声特征

Extracting the noise feature of the fan through wavelet analysis

利用窗口傅里叶分析对噪声信号进行频域分析可提取噪声的部分特征 ,但窗口傅氏变换对不同的频率成分在时域上的取样步长是不变的 ,不能聚焦到对象的任意细节 ,不利于噪声信号特征的提取 .本文利用小波分析 (Wavelet Analysis) ,对高频成分采用逐渐精细的时域或空域取样步长 ;利用 Mallat算法根据半波带滤波器 H、G实现信号的多尺度分解 .本文的目的是提取噪声特征 ,故选取 Daubechies构造的有限脉冲响应滤波器 { hn(k) } 2 n- 1 k=0 ,n=2 ,3,… ,作为低通滤波器 ,它对应的Φ (x)和Ψ (x)是紧支的 .考虑系统工作的实时性 ,取n=2 ,它对应的是四抽头滤波器 ,对风机在断相和正常运行下辐射噪声特征进行提取 .为了进一步分析噪声源辐射噪声信号在各频率段的能量特征 ,用 Bartlet平均周期图法对各级分解中的近似信号和细节信号进行功率谱估计 .

Analysis of the noise spectrum was dealt with through adding window Fourier transform in the field of frequencies.Some characteristics of noise was extracted.But sampling step to the different frequency in the field of the time was similar through adding window Fourier transform.It couldn′t focus on all the details of object and is disadvantageous to extracting the characteristic of noise.Wavelet analysis was employed to decompose noise signal.In order to extract the characteristic of noise,the authors chose finite pulse response filter provided by Daubechies as low pass filter, where Φ(x) and Ψ(x) were compact branches.The features of the noise source were acquired in the normal state and in the broken phase.Furthermore,energy characteristics of noise signal on the every frequency band were estimated in view of the power frequency of approximate signals and detail signals through the method of Bartlet average period.