首页|云边协同低延迟故障预测算法研究

云边协同低延迟故障预测算法研究

扫码查看
云计算拥有灵活性、高效率和低成本的优点,而边缘计算拥有解决传输响应延迟、带宽资源浪费、传输成本增加以及数据隐私风险等问题的能力.针对云计算与边缘计算各自的特点,研究云边协同故障预测算法.首先,利用小波变换对原始信号进行时频分析,将一维信号转换为时频图像数据.然后,在边缘端利用MobileNetV3模型对时频图像进行故障预测.如果预测结果的最大概率值达到事先设定阈值,则认为边缘端预测可靠;否则,时频图像将会被上传到云端,由云端GoogLeNet模型重新进行预测.实验结果表明,提出的云边协同故障预测算法能准确高效地预测设备故障,并在一定程度上降低了预测的延迟.
Research on low latency fault prediction algorithm for cloud-edge collaboration
Cloud computing possesses the benefits of flexibility,high efficiency and low cost,while edge computing possesses the ability to solve the problems of delayed transmission response,wastage of bandwidth resources,increased transmission costs and data privacy risks.Therefore,the fault prediction algorithm of cloud-edge collaboration was studied for the respective characteristics of cloud compu-ting and edge computing.Firstly,the wavelet transform is used to analyze the time-frequency analysis of the original signal and convert the one-dimensional signal data into time-frequency image data.Then,the time-frequency image obtained by wavelet transform is first predicted by using the MobileNetV3 model at the edge.If the maximum probability value of the prediction result reaches the pre-defined threshold,the edge prediction is considered reliable.Otherwise,the time-frequency image will be uploaded to the cloud and re-predicted by the GoogLeNet model in the cloud.Experiment results show that the proposed cloud-edge collaborative fault prediction algorithm can accurately and efficiently predict equipment faults,and reduce the latency of prediction to a certain extent.

fault predictionconvolutional neural networkswavelet transformcloud-edge collaboration

张文康、赵伟、刘德成、曹阳、王涛、王志晓

展开 >

河南龙宇能源股份有限公司,河南永城 476600

中国矿业大学计算机学院,江苏徐州 221116

故障预测 卷积神经网络 小波变换 云边协同

国家自然基金面上项目

61876168

2024

能源与环保
河南省煤炭科学研究院有限公司 河南省煤炭学会

能源与环保

CSTPCD
影响因子:0.221
ISSN:1003-0506
年,卷(期):2024.46(10)