基于KICA的旋转机械振动信号消噪方法  

KICA-BASED DENOISING METHOD FOR VIBRATION SIGNAL OF ROTATING MACHINERY

作　　者：蒋玲莉[1] 曾波[1] 冯国全[2] 李学军[1] 

机构地区：[1]湖南科技大学机械设备健康维护省重点实验室,湘潭411201 [2]沈阳发动机设计研究所,沈阳110015

出　　处：《机械强度》2013年第5期566-570,共5页Journal of Mechanical Strength

基　　金：国家自然科学基金(51105138);国家863计划项目(No.2012AA041805);湖南省教育厅重点项目(11A034);湖南省产学研结合创新平台(2010XK6066);湖南省高校科技创新团队支持计划资助~~

摘　　要：原始信号采集过程中总是无法避免引入各种噪声,消噪是信号分析处理的关键环节。已有的大量消噪方法各有优劣,仍需不断寻求新的消噪方法,以推进信号处理技术的发展。依据核独立成分分析(kernel independent component analysis,KICA)可将混合信号源的独立成分一一分离的特性,认为噪声信号和有用信号均可以作为独立成分分离开来,选取有用信号加以重构,即可实现消噪处理。通过引入适配噪声分量,有效实现了旋转机械转子不平衡振动信号的消噪处理,验证了方法的有效性。Denoising is an essential part of signal process, because the inevitable noise are easy to mix in useful signal during raw signal acquisition that always affect the accuracy of analytical results. New denoising methods are studied continuously to develop the signal process technology. The unique feature of kernel independent component analysis （KICA） method, which can separate independent components from mixed-signal source, is used for denoising. The noises and useful signals can be separated as independent components by KICA, then, choosing useful signals to reconstruct, the denoising process can be achieved. By introducing adaptive noise components, the noise of rotor unbalanced vibration signal is cancelled. The results demonstrate the validity of KICA-based denoising method.

关 键 词：核独立成分分析 消噪 旋转机械 故障诊断 

分 类 号：TH133.33[机械工程—机械制造及自动化] TH17