CAJ | 학술논문

针对早期故障微弱特征难以提取的问题,提出了一种基于非线性流形学习的滚动轴承早期故障特征提取方法.在由时域指标和小波频带能量组成的原始特征空间中,结合局部切空间排列学习算法的特点,采用散布矩阵分类测度指标,实现了局部邻域的优化选取,从而提取出最优的敏感故障特征.通过实例应用,表明该方法有效地克服了主分量分析和非线性核主分量分析方法的不足,提取的融合特征敏感性更好,从而提高了故障模式的分类性能,实现了轴承的早期故障诊断.
침대조기고장미약특정난이제취적문제,제출료일충기우비선성류형학습적곤동축승조기고장특정제취방법.재유시역지표화소파빈대능량조성적원시특정공간중,결합국부절공간배렬학습산법적특점,채용산포구진분류측도지표,실현료국부린역적우화선취,종이제취출최우적민감고장특정.통과실례응용,표명해방법유효지극복료주분량분석화비선성핵주분량분석방법적불족,제취적융합특정민감성경호,종이제고료고장모식적분류성능,실현료축승적조기고장진단.
To extract the incipient fault feature of rolling bearings, a novel feature extraction approach based on the nonlinear manifold learning algorithm is proposed. Constructing the original feature space with the time domain indexes and the wavelet frequency domain energies, the local target space alignment algorithm is employed for extracting nonlinear low dimensional manifold. According to the characteristics of the scatter matrix classification measure, the selection criterion of local neighborhood parameter is introduced tO acquire the sensitive fault features. The experimental results for fault diagnosis of rolling bearings show that this approach, compared with the linear principal component analysis and nonlinear kernel principal component analysis, is more effective to extract the fault features from vibration signals, and enhances the classification ability of failure pattern.