基于一维卷积神经网络的钻杆故障诊断  

Drill pipe fault diagnosis method based on one-dimensional convolutional neural network

作　　者：金列俊 詹建明 陈俊华 王涛 JIN Lie-jun;ZHAN Jian-ming;CHEN Jun-hua;WANG Tao(Institute of Mechanical Engineering,Zhejiang University,Hangzhou 310027,China;School of Mechanical and Energy Engineering,Ningbo Institute of Technology,Zhejiang University,Ningbo 315100,China)

机构地区：[1]浙江大学机械工程学院,浙江杭州310027 [2]浙江大学宁波理工学院机电与能源工程学院,浙江宁波315100

出　　处：《浙江大学学报（工学版）》2020年第3期467-474,共8页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金重点资助项目(U1813223);浙江省自然科学基金资助项目(LY17E050012).

摘　　要：为了在钻杆故障早期诊断出钻杆的故障类型,提出一种基于一维卷积神经网络的钻杆故障诊断模型,对模型的结构和参数进行详细地设计与分析.参考现有的卷积神经网络模型,结合钻杆的工作特性以及感受野的原理,设计模型的卷积层和池化层的层数、卷积核的大小以及滑动步长.该模型省去了对故障信号特征提取的过程,比先前的钻杆故障诊断有更高的诊断准确率.该模型在不同转速工况下和不同土质工况下均具有较强的适应性和抗噪能力.A drill pipe fault diagnosis model based on one-dimensional convolutional neural network was proposed in order to diagnose the fault type of drill pipe failure early;the structure as well as the parameter of the model were designed and analyzed in detail. Referring to the existing convolutional neural network model, the layer number of the convolutional layer as well as the pooling layer of the model, the size of the convolution kernel and the sliding step length were designed combining with the working characteristics of the drill pipe and the principle of the receptive field. The model eliminates the process of extracting fault signal features and has higher diagnostic accuracy than previous drill stem fault diagnosis. Meanwhile, the model has strong adaptability and anti-noise ability under different speed conditions and different soil conditions.

关 键 词：钻杆故障诊断 一维卷积神经网络 感受野 适应性 抗噪能力 

分 类 号：TH165.3[机械工程—机械制造及自动化] TN911.23[电子电信—通信与信息系统]