As the fault vibration signal characteristics presented non-stationary and the fault fre-quencies were hard to extracted,a new feature extraction method was proposed .This approach com-bined LCD and LTSA which was one of the typical manifold learning methods to extracting fault fre-quencies.Firstly,the vibration signals were decomposed into multiple intrinsic scale components in multidimensional feature vectors using LCD.Secondly,LTSA method was applied to compress the high-dimensional vectors into low-dimensional vectors,the low-dimensional vectors were used to re-construct and the new fault signals were obtained.Finally,the new fault signal's spectrum were ana-lysed and the fault characteristic frequencies were acquired.The rolling bearing fault experimental re-sults show that this new technique may extract the inner and outer ring fault frequencies,it verifies the effectiveness of this new approach.%为了从非线性、非平稳的振动信号中提取故障特征频率,提出了一种故障特征频率提取新方法.该方法将局部特征尺度分解和流形学习算法局部切空间排列相结合,首先利用局部特征尺度分解将振动信号分解成若干个内禀尺度分量,将其组成多维特征向量;其次采用流形学习算法中的局部切空间排列对多维特征向量进行降维处理,得到低维特征向量,对得到的低维特征向量进行信号重构;最后采用频谱分析方法对重构信号进行故障特征频率的提取.在滚动轴承故障试验中,所提出方法能够准确提取出内圈和外圈故障的特征频率,验证了该方法的有效性.
展开▼