摘要
针对基于单通道声信号的机械故障诊断信号干扰成分大,故障特征难以提取的问题,提出了一种结合改进自适应噪声完备经验模态分解(ICEEMDAN)和快速独立分量分析(FastICA)的方法。依据峭度与信号相关性准则设定本征内模分量(IMF)选取系数,对ICEEMDAN自适应分解的IMF进行有效筛选,实现信号降噪和粗提取,并以所选IMF作为虚拟通道,应用FastICA成功实现信噪的盲源分离。通过内外圈故障轴承实验数据对算法实行对比验证,结果表明,所提算法大幅降低了噪声及干扰,有效提取了故障特征。
Aiming at the problem in mechanical fault diagnosis based on single-channel acoustic signals that collected signals always have the intensive interference,from which fault features are difficult to be extracted,a method combining the improved complete ensemble empirical mode decomposition with adaptive noise(ICEEMDAN)and the fast independent component analysis(FastICA)were proposed.Setting an intrinsic mode function(IMF)selection coefficient according to both the kurtosis and the signal correlation as criterion,IMFs adaptively decomposed by ICEEMDAN were effectively screened to reduce the noise and roughly extract the signal features.FastICA was applied to the selected IMFs regarded as virtual channels,by which blind source separation could be successfully achieved.The as-proposed method was verified by experimental data of inner and outer bearing ring faults.The results show that the as-proposed method can greatly reduce the noise and interference,and is efficacious for extracting fault features.
作者
李篪
陈长征
LI Chi;CHEN Changzheng(School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110870,Liaoning,China)
出处
《沈阳工业大学学报》
CAS
北大核心
2023年第6期672-679,共8页
Journal of Shenyang University of Technology
基金
国家自然科学基金项目(51675350,51705337)
辽宁省“揭榜挂帅”科技攻关项目(2022JH1/10400008)。
关键词
机械故障诊断
声学诊断
声信号
滚动轴承
改进自适应噪声完备经验模态分解
快速独立分量分析
特征提取
盲源分离
mechanical fault diagnosis
acoustic diagnosis
acoustic signal
rolling bearing
improved complete ensemble empirical mode decomposition with adaptive noise
fast independent component analysis
fault feature extraction
blind source separation
