栈式稀疏加噪自编码深度神经网络的滚动轴承损伤程度诊断

发布时间：2018-01-11 01:30

本文关键词：栈式稀疏加噪自编码深度神经网络的滚动轴承损伤程度诊断　出处：《振动与冲击》2017年21期 　论文类型：期刊论文

【摘要】：针对滚动轴承损伤程度的特征自学习提取与智能诊断问题,提出栈式稀疏加噪自编码深度神经网络的滚动轴承损伤程度诊断方法。滚动轴承损伤特征受到工况、环境噪声等干扰,浅层自编码网络对损伤特征的自学习、提取能力不足。为此,论文将稀疏项限制和加噪编码融入自编码网络,同时将自编码网络堆栈并添加分类层,构建出栈式稀疏加噪自编码深度神经网络,进行轴承损伤特征非监督自动提取与损伤程度智能诊断。稀疏项限制和深度神经网络的构建提高了特征学习能力,加噪编码的融入改善了网络的鲁棒性。所构建深度神经网络通过多层无监督逐层自学习和有监督微调,完成损伤特征自动提取与表达,并实现了损伤程度智能诊断。不同工况下轴承损伤程度诊断的实验验证证明了所提方法的可行性和有效性。
[Abstract]:Aiming at the problem of feature self-learning extraction and intelligent diagnosis of rolling bearing damage degree, a new method based on stack sparse noise-adding self-coding depth neural network is proposed to diagnose the rolling bearing damage degree, and the damage feature of rolling bearing is subjected to working conditions. Due to the interference of environment noise and the ability of self-learning of shallow self-coding network to extract damage features, the sparse item restriction and noise-added coding are incorporated into the self-coding network in this paper. At the same time, the self-coding network stack is added and the classification layer is added to construct the stack sparse noisy self-coding depth neural network. Unsupervised automatic extraction of bearing damage features and intelligent diagnosis of damage degree are carried out. Sparse term limitation and the construction of depth neural network improve feature learning ability. The integration of noise-coding improves the robustness of the network. The depth neural network is constructed through multi-layer unsupervised self-learning and supervised fine-tuning to complete the automatic extraction and expression of damage features. The intelligent diagnosis of damage degree has been realized, and the feasibility and effectiveness of the proposed method have been proved by the experimental results of bearing damage degree diagnosis under different working conditions.
【作者单位】：重庆交通大学机电与车辆工程学院;四川大学空天科学与工程学院;重庆交通大学经济与管理学院;
【基金】：国家自然科学基金(51305471) 中国博士后科学基金(2014M560719;2016M590861) 重庆市基础与前沿研究计划资助项目(cstc2014jcyj A70009;cstc2015jcyj A70012) 重庆市教育委员会科学技术研究(KJ1400308) 国家留学基金(201408505081)
【分类号】：TH133.33;TP183
【正文快照】： 滚动轴承是旋转机械的重要部件,其性能退化或失效影响整机性能及其运行可靠性。目前滚动轴承故障诊断多集中于故障分类研究,而工程中仅判断是否发生故障以及故障类型难以进行预防维护,只有对损伤程度进行准确评估和诊断才能有效指导轴承维护工作。所以,轴承损伤程度的诊断研究

【相似文献】