摘要
为了提高轴承智能故障诊断能力,利用深度迁移自动编码器处理方法来实现轴承故障智能诊断。通过指数型线性缩放与非负约束处理技术能够促使自动编码器达到更优控制效果,设置足够源域数据预训练深度自动编码器模型,并测试各类该模型故障诊断的有效性。研究结果表明:由于存在噪声因素对结果造成的较大影响,使得源域和目标域都呈现差异很大的分布特征。实验测试诊断准确率均值达到88.46%,有助于对目标域新数据达到更好地匹配状态,再对源域知识实施转换转变至目标域。迁移模型诊断测试获得了89.42%的准确率,与其它迁移模型相比具备更高准确率;达到了0.341的标准差,达到了稳定的测试要求。该研究可以适用于其它的机械传动系统,具有很好的理论支撑价值。
In order to improve the bearing intelligent fault diagnosis ability,the deep migration autoencoder processing method is used to realize the bearing intelligent fault diagnosis.Through exponential linear scaling and non-negative constraint processing technology,the auto-encoder can achieve better control performance,set enough source domain data to pre-train the deep auto-encoder model,and test the validity of the fault diagnosis of various models.The results show that the source domain and the target domain have very different distribution characteristics due to the noise factor.The average diagnostic accuracy of experimental tests reached 88.46%,which was helpful to better match the new data in the target domain,and then transform the knowledge from the source domain to the target domain.The accuracy of the migration model diagnostic test is 89.42%,which is higher than other migration models.The standard deviation is O.341,which meets the test requirements of stability.This research can be applied to other mechanical transmission systems and has a good theoretical support value.
作者
刘小娟
夏运东
张明
马利华
史春娥
LIU Xiaojuan;XIA Yundong;ZHANG Ming;MA Lihua;SHI Chune(School of Automotive Engineering,Yellow River Transportation Institute,Jiaozuo Henan 454950,China;School of Mechanical Engineering,Henan University of Technology,Jiaozuo Henan 454000,China;Chery Commercial Vehicle Anhui Co.,Ltd.,Wuhu Anhui 241000,China)
出处
《机械设计与研究》
CSCD
北大核心
2022年第5期134-137,共4页
Machine Design And Research
基金
河南省高等学校重点科研项目(18A460024)。
