一种基于深度残差网络的旋转机械故障诊断方法  

A Rotating Machinery Fault Diagnosis Method Based on Deep Residual Networks

作　　者：刘芷源 王文权 杨鹏祺 周章玉 段昶[3] 廖飞龙 朱策 LIU Zhiyuan;WANG Wenquan;YANG Pengqi;ZHOU Zhangyu;DUAN Chang;LIAO Feilong;ZHU Ce(School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China;CNPC,CCDC,Safety Environment Quality Supervision and Testing Research Inspection,Deyang 618300,China;School of Electrical Engineering and Information,Southwest Petroleum University,Chengdu 610599,China)

机构地区：[1]电子科技大学信息与通信工程学院,四川成都611731 [2]中国石油集团川庆钻探工程有限公司安全环保质量监督检测研究院,四川德阳618300 [3]西南石油大学电气信息学院,四川成都610599

出　　处：《通信与信息技术》2024年第6期1-6,共6页Communication & Information Technology

基　　金：中国石油集团川庆钻探工程有限公司安全环保质量监督检测研究院支持(项目编号:CQ2023B-19-1-3)。

摘　　要：针对传统钻采设备故障诊断方法存在的成本高、缺乏实时性和普适性不足的问题,提出了一种基于深度残差网络的旋转机械故障诊断方法。该方法通过固定窗口长度的随机采样来构建充足的训练数据集,利用深层架构和残差学习机制,结合一维卷积神经网络提取时序信号中的多尺度特征,并通过全连接神经网络完成故障的识别与分类。实验结果表明,对公开数据集(CWRU)添加3dB噪声后,所提方法在测试集上的准确率依然保持在99%以上。在PT800实采数据集上,该方法的准确率也超过了98%。这些结果验证了该方法在变工况条件下对旋转机械设备故障进行高效、准确诊断的能力,为旋转机械故障的自动诊断提供了一种新的可行方法。To address the issues of high cost,lack of real-time capability,and insufficient generality in traditional drilling equip⁃ment fault diagnosis methods,a fault diagnosis method for rotating machinery based on deep residual networks is proposed.This meth⁃od constructs a sufficient training dataset through random sampling with a fixed window length,utilizes deep architecture and residual learning mechanisms,and employs one-dimensional convolutional neural networks to extract multi-scale features from time-series sig⁃nals.Fault recognition and classification are accomplished through a fully connected neural network.Simulation results indicate that after adding 3dB noise to a public dataset,the proposed method maintains an accuracy rate of over 99%on the test set.Additionally,on the PT800 real dataset,the accuracy exceeds 98%.These results demonstrate the method's capability for efficient and accurate diagnosis of rotating machinery faults under varying conditions,providing a new feasible approach for automatic fault diagnosis in such equipment.

关 键 词：钻采设备 故障诊断 深度残差网络 卷积神经网络 

分 类 号：TH165.3[机械工程—机械制造及自动化] TH17