摘要
为实现时频域高维信息辅助的振动信号微弱瞬态特征增强,提出了一种多尺度时频谱二值化方法.通过高分辨率时频谱切片提取能量突变点,其权重为1,其余点权重为0.进行多次不同尺度的二进制谱分析并降维至时域,得到针对多目标频率的权重谱与权重向量,从而实现微弱冲击特征的增强.用仿真信号分析对该方法的可行性与准确性进行了验证,列车轴承故障诊断试验则进一步验证了该方法在信号微弱特征提取中的有效性.
A multi-scale binaryzation method of time-frequency map was proposed to enhance the weak instant features of vibration signal with high dimensional information assist in time-frequency domain. The method extracts energy fluctuations in slices of high definition time-frequency map by weighting the points of energy fluctuation as 1 andother points as 0. Repeat the process of binary spectrum analysis with different scales and utilize dimensionality reduction to time domain. Then the weight spectrum and vector of weights are obtainedto enhance the weak shock features. Simulated signal processing confirms the validity and accuracy of this method. Train bearing experiment verifies the effectiveness of the method in weak feature extraction of vibration signal.
出处
《天津大学学报(自然科学与工程技术版)》
EI
CSCD
北大核心
2016年第7期667-673,共7页
Journal of Tianjin University:Science and Technology
基金
国家自然科学基金委员会与中国民航局联合资助项目(U1533103)
国家自然科学基金资助项目(51475324)
关键词
集合经验模式分解
二进谱
特征提取
信号处理
故障诊断
ensemble empirical mode decomposition(EEMD)
binary spectrum
feature extraction
signal proc-essing
fault diagnosis