摘要
针对液阀故障微弱信号特征识别问题,提出一种结合非抽样提升小波包(Undecimated lifting scheme packet,ULSP)及奇异值分解(Singular value decomposition,SVD)的降噪方法。确定信号的分解层次及各层初始算子的长度后,通过拉格朗日插值公式算出初始算子,用非抽样算法对原始信号进行分解。对最后一层各频带信号进行奇异值分解降噪处理,根据奇异熵增量曲线确定降噪阶次。用非抽样提升小波包的重构算法对信号进行重构,最终获得降噪后的信号。对降噪后的信号再进行非抽样提升小波包分解,以提取故障特征。对仿真信号的降噪表明,所提方法降噪获得较高的信噪比及较低的均方差,且能保留信号中应有的高频成分。提出的方法成功提取某往复式注水泵排水阀弹簧失效的微弱故障特征。
Aiming at the extraction of failure character signal for liquid valve,a novel method to combine undecimated lifting scheme packet(ULSP) with singular value decomposition(SVD) is developed.Initial operators are calculated by using Lagrange interpolation formula after determining decomposition level and the lengths of initial operators,and then the original signal is decomposed by using undecimated algorithm.All frequency bands signals of the last layer are denoised by using SVD thresholding,a reasonable order for noise reduction is selected according to the singular entropy of singular spectrum.Then signal is reconstructed by using reconstruction algorithm,and the denoised signal is decomposed by using ULSP again in order to extract fault feature.Simulative signal and engineering results confirm the better noise reduction of ULSP-SVD.The weak fault signal of the spring on the drain valve of a reciprocating water-flood pump is extracted from the strong vibration background.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2011年第9期72-77,共6页
Journal of Mechanical Engineering
基金
国家高技术研究发展计划(863计划
2008AA06Z209)
中国石油天然气集团公司创新基金(07E1005)资助项目
关键词
非抽样提升小波包
奇异值分解
液阀
故障诊断
Undecimated lifting scheme packet Singular value decomposition Liquid valve Fault diagnosis