摘要
针对辛几何模态分解(symplectic geometry mode decomposition,SGMD)难以完成强噪声干扰模式下振动信号有效分解的问题,提出一种基于高斯核相似度聚类的辛几何模态分解(symplectic geometry mode decomposition-Gaussian kernel similarity,SGMD-GKS)方法。在SGMD-GKS方法中,首先,对初始分量进行GKS聚类,实现强噪声干扰模式下不同模态准确分割;其次,引入包络谱峰值因子衡量模态的归属类型并作为聚类终止条件;最后,计算模态中各峰值对包络谱峰值因子的贡献度,进一步完成故障特征的深度挖掘。仿真和实际复合故障信号试验结果表明,SGMD-GKS方法可以实现复合故障振动信号不同模态的准确划分,并完成有效诊断。
To address the challenge that symplectic geometry mode decomposition(SGMD)struggles to achieve accurate delineation of different modal components in complex signals,this paper proposes a novel method termed symplectic geometry mode decomposition-Gaussian kernel similarity(SGMD-GKS)clustering.In the SGMD-GKS method,the initial components were first clustered via Gaussian Kernel Similarity(GKS)to achieve precise segmentation of complex data distributions under strong noise interference.Subsequently,the envelope spectrum crest factor was introduced to evaluate the affiliation type of each mode,which served as the termination criterion for clustering.Finally,the contribution degree of individual spectral peaks to the envelope spectrum crest factor was calculated,enabling deep mining of fault characteristics for enhanced feature extraction.Simulation and experimental results on compound fault signals demonstrate that the SGMD-GKS method achieves accurate separation of different modal components in complex vibration signals and enables effective diagnosis of rotating machinery faults.
作者
王自忠
潘海洋
郑近德
程健
童靳于
WANG Zizhong;PAN Haiyang;ZHENG Jinde;CHENG Jian;TONG Jinyu(School of Mechanical Engineering,Anhui University of Technology,Ma’anshan 243032,China)
出处
《振动与冲击》
北大核心
2025年第24期289-297,共9页
Journal of Vibration and Shock
基金
国家自然科学基金(52475080)
安徽省自然科学基金(2408085ME113)。
