振动信号Burg谱估计算法的性能优化研究

摘要

Burg算法是振动信号功率谱估计中常用的方法,具有分辨力性能优良、计算效率高的优点.但当数据为实数样本时会出现谱线分裂、谱峰偏移、信号初始相位和预测滤波器阶数对结果影响大等问题,影响了振动信号谱估计分析的准确性.针对Burg算法存在的问题,提出将最优窗引入到Burg算法中,通过信息论准则确定预测滤波器的最优阶次,通过加最优窗函数改善谱线分裂和谱峰偏移现象,实现对Burg谱估计算法的优化.结合工业现场实测汽轮机振动信号,利用仿真工具进行实验分析和仿真,并与直接Burg算法进行了仿真对比.理论分析和实验结果表明,最优阶次下的加窗Burg算法中谱线分裂、谱峰偏移现象得到明显改善,具有更优良的分辨力性能,提高了振动信号谱估计的准确性和质量.%With the advantages of high resolution and efficiency of calculation,Burg algorithm is a commonly used method in vibration signal power spectral estimation.Aimed at the problems in Burg algorithm,this paper intro duces optimal window into Burg algorithm,and determines the optimal order by AIC.The optimization of the Burg spectral estimation algorithm was realized by using the optimal window function to improve the spectral line splitting and peak shifting.Combing with the turbine's vibration signals which collected industrial site and the simulation tools to analyze and simulate the experiment,the simulation results were compare with the original Burg algorithm.Theoretical analysis and experimental results show that the spectral line splitting and peak shifting in windowed Burg algorithm under the optimal order are improved obviously,which has better resolution performance and improves the accuracy and quality of the vibration signal spectral estimation.