摘要
轴承故障诊断作为轨道交通技术的研究热点,是保障安全运行的重要一环。针对传统神经网络出现的特征抓取不足,池化层信息丢失引起的识别率较低的问题,给出了基于胶囊神经网络进行轴承故障诊断的方法。以凯斯西储大学的滚动轴承数据作为样本,提出通过改进胶囊神经网络提取数据的全部特征和局部特征,实现轴承状态识别。算法在基准数据集上获得97.58%的识别准确率,超过了当前轴承故障诊断的主流方法,该文算法具有一定的先进性。
As a research hotspot of rail transit technology,bearing fault diagnosis is an important part of ensuring safe operation.Aiming at the problems of insufficient feature capture in traditional neural network and low recognition rate caused by information loss in pooling layer,a method for bearing fault diagnosis based on capsule neural network is presented.Taking the rolling bearing data of Case Western Reserve University as a sample,it is proposed to extract all features and local features of the data by improving the capsule neural network to realize the bearing state recognition.The algorithm obtained a recognition accuracy of 97.58%on the benchmark data set,which exceeded the current mainstream methods of bearing fault diagnosis.The algorithm in this paper has a certain degree of advancement.
作者
郭占广
尹帅
谢敬玲
宫辉
GUO Zhan-guang;YIN Shuai;XIE Jing-ling;GONG Hui(Qingdao JARI Industrial Control Technology Co.,Ltd.,Qingdao 266520,China)
出处
《自动化与仪表》
2022年第12期49-53,共5页
Automation & Instrumentation
关键词
人工智能
胶囊神经网络
轴承故障诊断
artificial intelligence
capsule neural network
bearing fault diagnosis