CAJ | 학술논문

针对现有监督局部线性嵌入算法在低维输出向量重构过程中监督学习能力弱，不利于故障特征提取的问题，通过利用训练样本类标签信息扩大不同类样本间平均距离的方式，增加低维输出向量重构模型的监督学习能力，强化同类样本的聚集性和异类样本的互斥性。基于规范切割准则和低维输出向量重构误差，应用离散粒子群优化算法优化折中系数α和β、以及嵌入维数和邻域等参数，提高故障特征提取精度。将改进的监督局部线性嵌入方法应用于轴承故障特征提取，结果表明推荐方法的特征提取精度较高。
침대현유감독국부선성감입산법재저유수출향량중구과정중감독학습능력약，불리우고장특정제취적문제，통과이용훈련양본류표첨신식확대불동류양본간평균거리적방식，증가저유수출향량중구모형적감독학습능력，강화동류양본적취집성화이류양본적호척성。기우규범절할준칙화저유수출향량중구오차，응용리산입자군우화산법우화절중계수α화β、이급감입유수화린역등삼수，제고고장특정제취정도。장개진적감독국부선성감입방법응용우축승고장특정제취，결과표명추천방법적특정제취정도교고。
Aiming at the shortage of weak learning ability of the supervised locally linear embedding (SLLE ) algorithm being unfavorable to fault feature extraction in reconstructing lower-dimensional output vectors,the learning ability of reconstructed model of output vectors was improved via utilizing the information of class labels of training samples to increase the average distance between samples with different class labels.The aggregation of the same class samples and the mutual exclusion of samples with different class labels were enhanced.In order to enhance extraction precision of fault features,the binary particle swarm optimal (PSO)algorithm,the normalized cut or Ncut criterion and the reconstruction error were employed to optimize compromise coefficients,embedding dimension and neighborhood size.The improved SLLE was employed in the fault feature extraction of rolling bearings.The test results for fault diagnosis of rolling ball bearings showed that compared with other approaches,ISLLE is more effective to extract the fault features form vibration signals,and enhance the classification ability of failure pattern.