加权动态时间扭曲的不平衡结构转子故障诊断

Fault Diagnosis of Unbalanced Rotor Structure Based on Weighted Dynamic Time Warping

为了弥补实际工业数据和实验实验室数据之间的差距,提出了一种基于加权动态时间扭曲的不平衡结构转子故障诊断方法。通过使用基于故障信息内容的加权方案改进的软动态时间扭曲方法处理数据安全性和不平衡问题。在故障分类阶段,引入了一种结构转子故障的早期分类方法,通过仅将准确度作为目标来开发序列深度学习分类器,然后通过考虑准确性和早期性来定义早期决策策略。在试验台数据集上产生的结果证明了提出方法能够有效提升结构转子故障诊断的精度,并且有效同化数据之间的差异。

In order to fill the gap between the actual industrial data and the experimental laboratory data, a fault diagnosis method of unbalanced rotor structure based on weighted dynamic time distortion is proposed. By using the weighted scheme based on fault information content, the improved soft dynamic time warping method is used to deal with the problems of data security and imbalance. In the stage of fault classification, an early classification method of structural rotor fault is introduced. The sequential deep learning classifier is developed by taking only the accuracy as the goal, and then the early decision strategy is defined by considering the accuracy and early. The results on the test-bed data set show that the proposed method can effectively improve the accuracy of structural rotor fault diagnosis and effectively assimilate the differences between data.