摘要
自动特征提取在机械系统智能状态监测中起着至关重要的作用,可以自适应地从原始数据中学习特征并发现新的状态敏感特征。本文重点研究了不同深度的卷积神经网络(CNN)模型在没有先验知识的情况下从激励响应信号中挖掘代表信息和敏感特征的能力,并将螺栓连接结构的特征提取和装配紧度分类过程融合在一起。通过车架试验台螺栓连接转子激振实验数据验证该方法的有效性。结果表明,CNN方法自适应学习的特征可以表示信号与装配状态之间的复杂映射关系,并且比其他方法具有更高的准确率。
Automatic feature extraction plays a crucial role in the intelligent state monitoring of mechanical systems,which can adaptively learn features from raw data and discover new state-sensitive features.This research focuses on the ability of different depth convolution neural network(CNN)models to mine representative information and sensitive features from the excitation response signal without prior knowledge,and combine the feature extraction with assembly tightness classification process of the bolted structure.The effectiveness of the method is verified by excitation test data of car frame test bench bolt connection rotor.The results show that the feature learned adaptively by CNN can represent the complex mapping relationship between response signal and assembly state,and has higher accuracy than other methods.
作者
赵俊锋
张小丽
闫强
申彦斌
杨吉
Zhao Junfeng;Zhang Xiaoli;Yan Qiang;Shen Yanbin;Yang Ji(Ministry of Education Key Laboratory of Road Construction Technology and Equipment,School of Construction Machinery,Chang′an University,Xi′an 710064,China)
出处
《机械科学与技术》
CSCD
北大核心
2019年第3期351-357,共7页
Mechanical Science and Technology for Aerospace Engineering
基金
装备预研教育部联合基金项目(6141A02033111)
陕西省自然科学基础研究计划项目(2016JQ5030)
陕西省高校科协青年人才托举计划项目(20170509)
陕西省博士后科研项目资助
关键词
螺栓
装配紧度
卷积神经网络
特征提取
bolt
monitoring
convolution neural network
features extraction
