Research On The Cloud/Fog/Edge Collaborated Method For Bearing Fault Diagnosis

Nowadays,in order to meet the increasing requirements for product quality,product complexity,and production efficiency,electromechanical equipment is required to have a better performance in reliability and stability.As an indispensable part of electromechanical equipment,bearing has a great risk of breaking down during running.If the operators cannot predict and handle these faults timely,there would be a serious damage,which probably leads to high maintenance costs,severe economic losses,and even safety issues.The current technology of bearing fault diagnosis still has the following problems.Firstly,Traditional diagnostic methods need to extract features manually,which is time-consuming and the diagnosis result is unstable.Secondly,the Convolutional Neural Networks(CNN)based diagnostic methods take huge amount of computing resources and long training time,which cannot meet the real-time response demand of fault diagnosis.To solve above problems,this paper proposed a cloud/fog/edge collaborated method for bearings fault diagnosis,which realizes accurate and real-time fault diagnosis of the bearings in the device cluster.There are both advantages and disadvantages of realizing the fault diagnose totally on cloud,fog or edge.Therefore,this paper establishes the collaborated sub tasks of the cloud,fog and edge to maximize their advantages and realize the accurate and real-time diagnosis of bearing fault.To implement these tasks,this paper improved an existing one-dimensional convolutional neural network and realized the collaboration of cloud task,fog task and edge task using transfer learning.At the same time,a real-time data management system for bearing running information is constructed based on Flume and Kafka tools.The system realizes real-time collection,transmission and storage of data generated by each device in the cluster.In addition,it will match all kinds of collected data to the corresponding consumers according to the type of bearings and working conditions in the cluster.Data driven for the tasks in cloud,fog and edge is realized through this system.Then a cloud/fog/edge collaborated platform for bearing fault diagnosis is built in this paper.A testbed is used as the equipment end,which can simulate different bearing operating conditions and complete the collection of real-time operation information for bearings.A complete computing platform including cloud,fog and edge is built,which can simulate cloud server,fog server and edge controller.Based on the framework of the testbed and computing platform,the realization and verification of the real-time data management system for the running status of the equipment in the cluster are completed.The accuracy,real-time,practicality and efficiency of the cloud/fog/edge collaborated method for bearing fault diagnosis are verified by constructing a comparative experiment on the above constructed experimental platform.