基于高阶累积量MUSIC法的轧辊偏心信号提取  

Roller eccentricity signal extraction based on higher-order cumulant MUSIC algorithm

作　　者：王洪希[1,2] 杨卫东[1] 田伟[2] 

机构地区：[1]北京科技大学自动化学院,北京100083 [2]北华大学电气信息工程学院,吉林132021

出　　处：《仪器仪表学报》2013年第10期2279-2286,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61179012)资助项目

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,研究了一种基于高阶累积量的MUSIC法和Prony法相结合的轧辊偏心信号提取新方法。首先采用基于高阶累积量的MUSIC法对偏心信号进行空间分解达到降阶的效果,能够有效地抑制噪声,在信噪比低时仍具有高的频谱分辨率,能准确提取出偏心谐波的频率及谐波的个数,对于频率相近的偏心谐波信号,不存在频带重叠的问题,克服了低信噪比时FFT法出现的频谱泄漏和栅栏效应等缺点。通过高阶累积量MUSIC法对偏心信号进行了消噪和降阶预处理后,使用Prony方法进一步估计偏心信号的各次谐波幅值和相位,弥补了Prony法对噪声敏感的弱点及对信号阶数的要求,充分发挥了其优点。仿真及实验结果验证了该结合方法的可行性和有效性,低信噪比时能够准确提取出偏心信号参数,重构的偏心模型比FFT法精度高,偏心补偿效果明显。Roll eccentricity signal is a complex high frequency periodic signal that is mixed with various random dis- turbances and contains multiple harmonics. In this paper,a new roll eccentricity signal extraction method is presen- ted, which combines the multiple signal classification （MUSIC） algorithm based on the higher order cumulant （HOC） and Prony method. Firstly, the eccentricity signal is decomposed in the space to reduce its order and effec- tively suppress noise using the MUSIC method based on HOC;in low SNR,this method still has high spectral resolu- tion and can accurately extract the frequencies of the eccentric harmonics and the number of the harmonics. At the same time, for the eccentric harmonic signals with similar frequencies, there is no band overlapping problem. This method overcomes the shortcomings of frequency spectrum leakage and fence effect that the fast Fourier transform （FFT） method has in low SNR. After the pretreatment of de-noising and order reduction, Prony method is used to es- timate the amplitudes and phases of the harmonics,which remedies the weaknesses of sensitive to noise and the de- mand for the signal order of the Prony method, and makes full use of the advantages of the Prony method. The simula- tion and experiment results show that the combined method is effective and feasible, and this method can accurately extract the eccentricity signal parameters in low SNR. The reconstructed eccentric model with the proposed method has higher precision than that with the FFT method, and the eccentricity compensation effect is obvious.

关 键 词：多信号分类 高阶累积量 PRONY法 轧辊偏心 参数估计 

分 类 号：TP334.9[自动化与计算机技术—计算机系统结构]