针对提取频率相近信号特征的经典算法Zoom-FFT(ZFFT)在分离差频1Hz以内的微速差信号时,需要足够的数据长度才能达到良好的信号分离效果的问题,为了在不增加数据长度的基础上提高信号频率分辨率,在ZFFT算法的基础上产生了ZFFT+FT算法.通过LabVIEW软件仿真试验研究了该算法对频率差1Hz和0.5 Hz的微速差信号提取精度,并与chirp Z transform(CZT)算法、ZFFT算法以及FFT算法的提取精度作了对比,通过提取结果的相对误差及方差来衡量提取精度.在同轴微速差双转子试验台进行现场试验,利用FFT对微速差信号28s数据集的提取结果与CZT,ZFFT及ZFFT+FT算法对4s数据集的提取结果作对比,ZFFT+FT与FFT算法的提取结果最大相差0.56 μm,验证了ZFFT+FT的实时性和精确性.结果表明,ZFFT+FT算法能够实时精准提取频率差1Hz以内的微速差信号.研究结果可为旋转机械微速差信号的提取提供参考.
Research on the accuracy of micro-speed difference dual-rotor signal separation method
The separation of micro-speed difference signals within differential frequency 1 Hz is a challenge in the field of dual-rotor vibration feature extraction.Zoom-FFT(ZFFT)is a classical algorithm for extracting the features of signals with similar frequencies.However,it needs sufficient data length to achieve a good signal separation effect.In order to improve the signal frequency resolution without increasing the data length,the ZFFT+FT algorithm is generated on the basis of ZFFT algorithm.The extraction accuracy of the algorithm for micro-speed difference signals with frequency difference of 1 Hz and 0.5 Hz was studied through LabVIEW software simulation experiments,and compared with that of chirp Z transform(CZT),ZFFT and FFT algorithms.The extraction accuracy was measured by the relative error and variance of the extraction results.Field experiments were carried out in the coaxial micro-speed difference dual-rotor test stand,and the extraction results of 28s dataset of micro-speed difference signal by FFT were compared with those of 4s dataset extracted by CZT,ZFFT and ZFFT+FT algorithms.The maximum error between ZFFT+FT algorithm and FFT algorithm is 0.56 μm,which verifies the real-time performance and accuracy of ZFFT+FT.The results indicate that the ZFFT+FT algorithm can accurately extract the micro-speed difference signals within frequency difference of 1 Hz in real time.The research results can provide reference for the extraction of micro-speed difference signals in rotating machinery.
国家科技期刊平台
NSTL
万方数据
