基于联邦学习的轴承故障检测与诊断

Bearing Fault Detection and Diagnosis Based on Federated Learning

首先通过研究轴承故障的理论基础,介绍传统的振动信号分析方法,总结轴承故障算法缺陷,在此基础上进行卷积神经网络(CNN)的学习以用于模型训练、联邦学习;然后通过FedAvg算法将服务机接收的各客户端模型聚合,得到全局模型参数来检测轴承故障;最后进行验证,实现基于联邦学习的轴承故障检测。经大量实验论证,该算法准确有效。

Firstly,by studying the theoretical basis of bearing fault,the traditional vibration signal analysis method is introduced,and the defects of bearing fault algorithm are summarized.On this basis,convolutional neural network(CNN)is studied for model training and federated learning.Then,the FedAvg algorithm is used to aggregate the client models received by the server to obtain the global model parameters for bearing fault detection.Finally,it is verified that the bearing fault detection based on Federated learning is realized.A large number of experiments show that the algorithm is accurate and effective.