基于GRU-1DFCN的硝化机搅拌系统故障诊断研究  

Research on Fault Diagnosis of Nitrifying Machine Stirring System Based on GRU-1DFCN

作　　者：孙俪榕 孙琤 SUN Li-rong;SUN Cheng(China Wuzhou Engineering Group Corporation Ltd.,Beijing 100053,China)

机构地区：[1]中国五洲工程设计集团有限公司,北京100053

出　　处：《工程建设与设计》2025年第7期55-58,共4页Construction & Design for Engineering

基　　金：西城区科技专项项目(XCSTS-YI2024-29)。

摘　　要：为提升涉火企业现场对硝化机等设备的维保效率,论文提出一种基于GRU-1DFCN的硝化机搅拌系统故障诊断方法。首先,采用GRU和1DFCN的并列式结构构建特征提取器;其次,采用Concat方法融合故障特征;最后,利用多分类算法Softmax实现对搅拌系统关键部件不同位置和不同故障类型的识别。故障诊断实例结果表明,论文模型在4种负载下的平均故障诊断准确率可以达到99.26%,相对于GRU、1DFCN、LSTM、CNN-LSTM、BP、SVM、KNN模型分别提高了0.57%、0.49%、3.5%、2.5%、16.14%、18.73%、19.31%,并且具有良好的泛化性和抗噪性能。In order to improve the maintenance efficiency such as nitrifying machine at fire-related enterprises'sites,the paper proposes a fault diagnosis method for nitrifying machine stirring system based on GRU-1DFCN.Initially,a parallel architecture integrating GRU and 1DFCN is established to serve as the feature extractor.Subsequently,fault features are integrated by Concat,and ultimately,the Softmax multi-classification algorithm is employed to pinpoint different fault types and locations within the mixing system's key components.The results from fault diagnosis exemplars demonstrate that the proposed model attains an average fault diagnosis accuracy of 99.26%across four distinct loads.The accuracy surpasses that of the GRU,1DFCN,LSTM,CNN-LSTM,BP,SVM,and KNN models by 0.57%,0.49%,3.5%,2.5%,16.14%,18.73%,and 19.31%,respectively.Furthermore,the model exhibits robust generalization capabilities and noise resistance.

关 键 词：故障诊断 门控循环单元 全卷积神经网络 硝化机搅拌系统 

分 类 号：TH133.3[机械工程—机械制造及自动化]