CAJ | 학술논문

故障识别是确定故障类型的重要方式。传统方法不能直观识别故障类型，忽略了水平和垂直方向的信息之间的关系，很难准确提取故障特征。二维全息谱融合了水平和垂直方向的振动信息，反映了一个支承面上转子的振动情况。但在某些情况下不能准确识别主要故障，无法通过分倍频、工频和高倍频的椭圆信息确定故障类型。选择流形学习的 LE 算法与全息谱技术结合，弥补了二维全息谱算法的缺陷，提高了流形学习处理信号的优越性。通过实验验证了方法的准确性。
고장식별시학정고장류형적중요방식。전통방법불능직관식별고장류형，홀략료수평화수직방향적신식지간적관계，흔난준학제취고장특정。이유전식보융합료수평화수직방향적진동신식，반영료일개지승면상전자적진동정황。단재모사정황하불능준학식별주요고장，무법통과분배빈、공빈화고배빈적타원신식학정고장류형。선택류형학습적 LE 산법여전식보기술결합，미보료이유전식보산법적결함，제고료류형학습처리신호적우월성。통과실험험증료방법적준학성。
Fault identification is an important way to determine the form of fault. Traditional methods can not intuitively identi-fy fault types, ignoring the relationship between the information of the horizontal and vertical direction and it is difficult to ac-curately extract the fault feature. Two dimensional holographic spectrum has blend the vibration of horizontal and vertical di-rection information, reflecting the vibration of the rotor on a supporting surface. But in some cases it can not accurately identify major failure and determine the failure types through elliptic information of points frequency doubling, power frequency and high frequency. Choosing the manifold learning LE algorithm and combined with holographic spectrum technology, making up for the defects of the two dimensional holospectrum algorithm and improving the advantages of manifold learning signal process-ing. the correct result have been got by experiment.