旋转机械振动信号奇异值子空间分解滤波  

Singular Value Subspace Decomposition Filtering for Rotating Machinery Vibration Signal

作　　者：张峰[1] 杨尚君[1] 石现峰[1] 

机构地区：[1]西安工业大学电子信息工程学院,西安710021

出　　处：《西安工业大学学报》2017年第8期634-640,共7页Journal of Xi’an Technological University

基　　金：国家自然科学基金项目(60972095;61271362);陕西省科技厅自然基金项目(2012JQ8008);陕西省教育厅科技专项(2012JK0545)

摘　　要：为了消除振动信号噪声对旋转机械状态检测和故障诊断的干扰.基于奇异值分解构建奇异值子空间法,通过构造Hankel矩阵对振动信号进行了奇异值分解.根据奇异值子空间的分布特性,将奇异值子空间分为信号奇异值子空间和噪声奇异值子空间.通过保留信号奇异值子空间,去除了噪声奇异值子空间,给出了基于奇异值分解振动信号滤波算法.结合汽轮机振动实测信号,对算法进行分析与仿真.结果表明:该算法计算量小于维纳滤波算法,保证了振动信号在线去噪处理的实时性;滤波前后振动信号线性相位特性保持不变,滤波后信号未产生相位失真.In order to eliminate the interference of vibration signal＇s noise during the state detection and fault diagnosis for rotating machinery, singular value decomposition of vibration signal is obtained by constructing Hankel matrix based on singular value decomposition to construct the singular value subspace. According to the distribution characteristics of singular value subspace, the singular value subspace is divided into the signal singular value subspace and the noise singular value subspace. By preserving the signal singular value subspace and removing the noise singular value subspace, the vibration signal filtering algorithm based on singular value decomposition is proposed. The algorithm is analyzed and simulated on the basis of the measured vibration signal of steam turbine. The results show that the calculation amount of this algorithm is less than Wiener filtering algorithm, also the real time performance of on-line vibration signal denoising is guaranteed, and during the filtering process, the linear phase characteristic of the vibration signal remains unchanged. After filtering the signal does not produce phase distortion.

关 键 词：振动 奇异值子空间 滤波 噪声 

分 类 号：TH113.1[机械工程—机械设计及理论]