基于高阶累计量 Root-MUSIC 法和 Prony 法的轧辊偏心谐波参数估计

摘要

针对轧辊偏心信号是混杂在各种随机干扰中含有多次谐波的复杂高频周期信号，以及FFT （Fast Fourier Transform）法对轧辊偏心信号分析的局限性，研究了一种基于四阶累积量的Root-MUSIC（Multiple Signal Classification）法和Prony法相结合的轧辊偏心信号估计新方法。利用基于四阶累积量的Root-MUSIC法准确估计出偏心谐波的频率及谐波的个数，同时由Root-MUSIC求得的根直接使用Prony方法估计出偏心信号的各次谐波幅值和相位。仿真结果和实验结果也验证了结合方法的可行性和有效性，在信噪比较低的情况下仍具有较高的频谱分辨率和估计精度，能准确地同时估计出偏心谐波的频率、幅值及相位，尤其在频率分辨率和抗噪声上具有FFT法无法比拟的优越性。%The roll eccentricity signal was complex high frequency periodic signal which is mixed in a variety of random dis-turbance .As the method of FFT (Fast Fourier Transform ) is limitations for analyzing the roll eccentricity signal ,in this article a new method is presented that is the combination of Root-Multiple Signal Classification (Root-MUSIC ) based on higher-order cumulans (HOC ) and Prony method .Using Root-MUSIC of HOC accurately estimate the harmonic frequencies and harmonic number in the meantime the roots which were obtained by the Root-MUSIC directly estimate the harmonic amplitude and phase by using Prony method .The simulation results and experimental results showed that the method is feasibility and effectiveness .In the case of low SNR it still has high spectral resolution and estimation accuracy and can estimate eccentric harmonic frequency amplitude and phase .The advantages of this approach is outstanding than FFT in the frequency resolution and restraining noise .