基于开放域适应网络的主风机故障诊断方法研究  

Fault Diagnosis of Main Fan Based on Open-domain Adaptive Networks

作　　者：刘晓莉 何健聪 Xiao-li Liu;Jian-cong He(Natural Gas Purification Plant;Hunan Eagle Eye Online Electronic Technology Co.,Ltd.)

机构地区：[1]天然气净化总厂信息科技部 [2]湖南省鹰眼在线电子科技有限公司

出　　处：《风机技术》2025年第2期90-96,共7页Chinese Journal of Turbomachinery

摘　　要：为了提高天然气净化厂中主风机等动设备运维检修效率,以该厂K1401B主风机为例,提出了基于开放域适应网络(Open Set Domain Adaptation,OSDA)的多重交叉故障预测方法。采集主风机振动数据,并运用小波变换去噪,将模型样本映射到源域,并进行目标域划分,目标域中已知故障类型,通过网络进行特征映射到源域完成解耦,根据预测概率分析当前风机故障类型。结果显示网络结构解耦出了90.9%的概率为角度不对中故障,79%的概率为轴承磨损故障,41.7%的概率为停机故障。结论认为,基于开放域适应网络的方法能利用源域中的故障特征库,准确预测目标域故障类型,确保动设备运转过程的安全运行。In order to improve the overhaul efficiency of the main fan and other dynamic equipments in the natural gas purification plant,a multiple cross-fault prediction method based on Open Set Domain Adaptation is proposed to analyze the fault type for the K1401B main fan.Collected the main fan vibration data and mapped the model samples to the source domain and target domain division after denoising by wavelet transform,the target domain of the known faults class can be mapped to the source domain through the network features to complete the decoupling,analyzed the current fan failure conditions according to the predicted probability.The results show that the network model decoupled 90.9%probability of angular misalignment failure,79%probability of bearing wear failure and 41.7%probability of downtime failure.It is concluded that the open-domain adaptive network can use the fault feature database in the source domain to the fault types in the target domain to ensure the safe operation of the rotated equipments.

关 键 词：风机 开放域适应网络 多重交叉故障预测 小波变换 安全运行 

分 类 号：TH165.3[机械工程—机械制造及自动化] TP306.3[自动化与计算机技术—计算机系统结构]