基于多层超限学习机的滚动轴承故障诊断方法

发布时间：2018-01-13 06:36

本文关键词：基于多层超限学习机的滚动轴承故障诊断方法　出处：《科学技术与工程》2017年14期 　论文类型：期刊论文

【摘要】：针对目前轴承故障诊断领域存在的海量数据问题及快速学习、实时监测的诊断要求,采用一种多层超限学习机方法对滚动轴承故障数据进行诊断测试。该方法直接学习轴承故障振动时域信号,与传统诊断方法相比,省去了复杂的信号处理过程,更加简便。将多层超限学习机方法的诊断结果分别与单层超限学习机、深度神经网络方法的诊断结果进行比较,多层超限学习机具有明显优势:(1)与单层超限学习机相比,多层超限学习机具有更好地学习和特征提取能力,其诊断准确率可达到98.29%;(2)与深度神经网络相比,多层超限学习机能够在保证较高诊断准确率的前提下,获得较快的训练速度,其训练速度较深度神经网络提高了41倍。结果表明,所采用的方法在滚动轴承故障诊断方面具有很好的效果和应用价值。
[Abstract]:Aiming at the problems existing in the field of mass data bearing fault diagnosis and diagnosis of fast learning, real time monitoring, using a machine learning method for multi gauge diagnostic tests on rolling bearing fault data. The method of fault vibration signal of bearing direct learning, compared with the traditional method, complex signal processing process, saves more convenient the multilayer learning machine. The diagnosis results were overrun and overload of single machine learning, compare the depth of the neural network method of diagnosis results, learning machine has obvious advantages over multi: (1) compared with single gauge machine learning, machine learning has better multilayer transfinite learning and feature extraction ability, its accuracy can be reached 98.29%; (2) compared with the depth of the neural network, multilayer learning machine can guarantee the gauge accuracy of diagnosis, obtain a faster training speed The training speed is 41 times higher than that of deep neural network. The results show that the method has good effect and application value in rolling bearing fault diagnosis.

【作者单位】：东北大学机械工程与自动化学院;
【基金】：辽宁重大装备制造协同创新中心国家自然科学基金(61573093)资助
【分类号】：TH133.33;TP181
【正文快照】： 引用格式:郝丽娜,王风立,曹瑞珉.基于多层超限学习机的滚动轴承故障诊断方法[J].科学技术与工程,2017,17(14):86—91Hao Lina,Wang Fengli,Cao Ruimin.Multi-layer extreme learning ma-chine methods based fault diagnosis of rolling bearing[J].Science Tech-nology and E

【相似文献】