基于边云协同的数控机床故障诊断联邦学习研究OA北大核心CSTPCD

In order to solve the problem when the spindle bearing and cutter fault diagnosis model of NC machine tool is trained,it needs massive data and takes a long time.A federated mean learning fault diagnosis model based on edge-cloud collaborative architecture is proposed.Firstly,one-dimensional convolutional neural network model architecture is designed,and local model training is carried out on each edge client to reduce data upload scale and share computing pressure on cloud server side.Then,the model aggregation algorithm is optimized based on the accuracy on the cloud server side,and the edge client screening algorithm is improved to accelerate the convergence rate and improve the accuracy of the model.Thirdly,the KubeEdge cloud collaborative platform based on Kubernetes is built on the cloud server side to shorten the communication time of data transmission.Finally,the experimental results show that the accuracy of fault diagnosis in each edge client of the model is stable at about 87.5％.Compared with the control group,the convergence speed and training time are optimized.