基于深度学习的选煤厂振动筛故障诊断方法  

Fault Diagnosis Method for Vibrating Screen in Coal Preparation Plant Based on Deep Learning

作　　者：何长林 李越 王斌[1] 李彪[1] 李敬兆[2] HE Changlin;LI Yue;WANG Bin;LI Biao;LI Jingzhao(Huainan Mining Group Coal Preparation Branch,Huainan,Anhui 232000,China;School of Computer Science and Engineering,Anhui University of Science and Technology,Huainan,Anhui 232001,China)

机构地区：[1]淮南矿业集团选煤分公司,安徽淮南232000 [2]安徽理工大学计算机科学与工程学院,安徽淮南232001

出　　处：《自动化应用》2024年第21期8-11,共4页Automation Application

摘　　要：振动筛是选煤厂的关键设备。为及时发现振动筛可能出现的故障,通过非接触式智能感知及智慧处理装置采集并处理振动筛的音频信号,提出基于长短期记忆网络(LSTM)和时空图神经网络(ST-GNN)的故障诊断模型。首先利用经验模态分解(EMD)对音频进行预处理,其次使用ST-GNN-LSTM模型分别提取信号的时间和空间特征,融合时空特征最终实现故障诊断。将该模型算法嵌入到边缘端,最终由智能算法与软硬件结合共同组成选煤厂振动筛的故障诊断监测系统。实验表明,该方法能有效提取并利用振动筛音频信号特征,实现振动筛的故障诊断。The vibrating screen is a key equipment in coal preparation plants.To detect possible faults in the vibrating screen in a timely manner,a fault diagnosis model based on long short-term memory(LSTM)and spatiotemporal graph neural network(ST-GNN)is proposed by collecting and processing the audio signals of the vibrating screen through a non-contact intelligent sensing and intelligent processing device.Firstly,the audio is preprocessed using empirical mode decomposition(EMD).Secondly,the ST-GNN-LSTM model is used to extract the temporal and spatial features of the signal,and the spatiotemporal features are fused to achieve fault diagnosis.Embedding the model algorithm into the edge end,and ultimately combining intelligent algorithms with software and hardware to form a fault diagnosis and monitoring system for the vibrating screen in the coal preparation plant.The experiment shows that this method can effectively extract and utilize the audio signal features of the vibrating screen to achieve fault diagnosis of the vibrating screen.

关 键 词：振动筛 长短期记忆网络 时空图神经网络 故障诊断 

分 类 号：TD67[矿业工程—矿山机电]