基于EEMD与LLE的旋转机械设备特征融合算法  

Feature fusion algorithm of rotating machinery based on EEMD and LLE

作　　者：董志远 刘淑杰[1] 张洪潮[1,2] DONG Zhiyuan;LIU Shujie;ZHANG Hongchao

机构地区：[1]大连理工大学机械工程学院,辽宁大连116024 [2]德克萨斯理工大学工业工程系,美国德克萨斯卢伯克79409

出　　处：《现代机械》2022年第6期19-25,共7页Modern Machinery

摘　　要：对旋转机械运行性能进行有效的监测和评估可以避免突发事故,最大程度地降低损失,是进行准确的故障诊断与寿命预测的关键。本文首先基于旋转设备的运行数据进行多域退化特征提取,提取了包括峰峰值、峭度因子在内的五个时域特征和包括中心频率、频率方差在内的四个频域特征;随后基于经验集合模态分析EEMD算法对螺旋桨的时频域特征进行提取,计算提取的各个时频域特征与原信号的相关程度,选取相关系数较高的四个特征向量,与前面提取的时域与频域特征构建多域特征集。考虑到仅依靠单个的时域或频域特征难以完全描述其性能的退化特征和趋势。本文提出了一种基于局部线性嵌入LLE的特征融合方法,对包含基于EEMD算法所提取的时频域特征在内的多域特征集合进行降维融合,并结合切比雪夫不等式对设备的不同运行状态进行识别,实现性能退化检测,并通过实验轴承数据验证了本文方法的准确性。Effective monitoring and evaluation of the operation performance of the rotating machinery can avoid unexpected accidents and reduce losses to the maximum extent,and is the key to accurate fault diagnosis and life prediction.In this paper,the multi-domain degradation features are extracted based on the operation data of the rotating equipment.Five time-domain features,including peak-to-peak value and kurtosis factor,and four frequency-domain features,including center frequency and frequency variance,are extracted.Then,the time-frequency domain features of the propeller are extracted based on the empirical set modal analysis(EEMD)algorithm,and the correlation between the extracted time-frequency domain features and the original signal is calculated.Four feature vectors with high correlation coefficients are selected to construct multi-domain feature sets combining with the previously extracted time-domain and frequency-domain features.Considering that it is difficult to completely describe the degradation characteristics and performance trends only depending on a single time-domain or frequency-domain feature,a feature fusion method based on local linear embedding(LLE)is proposed to reduce the dimensionality of the multi-domain feature sets,and Chebyshev inequality is adopted to identify different operating states of the device to realize performance degradation detection.The accuracy of the proposed method is verified by experimental bearing data.

关 键 词：特征融合 LLE 性能退化 切比雪夫不等式 

分 类 号：TH133.33[机械工程—机械制造及自动化]