摘要
鉴于在实际的应用中滚动轴承的故障信号所属的类别往往是未知的,而且为了得到一定的测试数据需要花费大量的时间,甚至对机械设备造成了一些损害。利用极限学习机训练速度快且泛化能力强的特点,提出了一种基于半监督极限学习机的滚动轴承故障诊断方法,该方法允许在有少量带标签的轴承故障数据的情况下,将带标签的历史数据与新采集到的部分未带标签的数据一起用来训练得到一个最优的诊断模型。首先通过相空间重构将原始一维信号映射到一个高维的相空间,在相空间中提取初始的轴承特征集,然后将特征集输入半监督的极限学习机中进行训练和测试。实验结果表明,这种基于半监督算法的诊断模型简单,在神经元个数较少的情况下仍然具有很好的泛化能力,具有一定的应用价值。
In the actual application of the fault diagnosis of rolling bearing,the category of the fault signal is often unknown.And,obtaining labels for fully supervised learning is time consuming and expensive,it even causes some loss of mechanical equipment.A semi-supervised extreme learning machine is proposed for fault diagnosis on behalf of the fast training speed and strong generalization ability of extreme learning machine.For semi-supervised method,both few labeled data and plenty of unlabeled data are used for training to get an optimum diagnosis model.In the experiment,the original one-dimensional signal is mapped into a high dimensional phase space,and then the initial feature set can be obtained in the phase space.Then the feature set is inputted into the semi-supervised extreme learning machine for training and testing.The experimental results show that this semi-supervised diagnosis model is simple,and it still has good generalization ability in the case of less number of neurons,and has certain application value.
出处
《测控技术》
CSCD
2016年第3期13-16,21,共5页
Measurement & Control Technology
基金
国家自然科学基金项目(51375037
51135001)
教育部新世纪优秀人才支持计划(NCET-12-0759)
关键词
滚动轴承
故障诊断
极限学习机
半监督学习
rolling bearing
fault diagnosis
extreme learning machine
semi-supervised learning
