Design And Implementation Of Based Intelligent Monitoring Of Motor State On Federated Learning

With the proposal of "made in China 2025",China has put the innovation of manufacturing industry in an important position.The safe operation of production equipment has been paid more and more attention by enterprises.As the key part of production equipment,motor condition monitoring is very necessary.At the same time,artificial intelligence technology and big data technology are also developing,machine learning is widely used in industrial modeling.However,there are still many problems in the actual industrial scene: the sharing rate of production information data distributed in different edge or data center is low,and the security and privacy of information data in the process of transmission and use are difficult to evaluate.In response to the above problems,this thesis introduces federated learning technology into the field of motor state monitoring,where multiple edge nodes jointly train a global shared model without the need to transmit production data.This is of great significance for the sharing of production information.The main research work of this thesis is as follows1.Aiming at the problem of the difficulty of sharing and reusing motor operating information.Introduce federated learning technology,and use federated mechanism to train global models among multiple edge nodes,so that scattered operating information among different devices can be effectively shared.This method reduces the consumption of data bandwidth and increases data security and privacy.2.Aiming at the data distribution scene in industry,this thesis studies the federated learning algorithm,improves the weighted average algorithm and aggregation strategy of Federated learning,effectively improves the classification effect of global model,and reduces the training time of global model.3.Considering that the global information of single feature model is not complete enough in motor state classification,a method of model pre classification and semantic rule reasoning is proposed.The classification result of Federated learning is added to the motor ontology,and then the rule reasoning described by RDF Prolog and Lisp language is established,and the classification result of Federated learning and the real-time state of motor are considered in the rule language.Finally,the inference engine provided by Allegrograph is used for motor state inference.4.A federal learning cluster is built to test the effectiveness of the intelligent motor state monitoring system.In this paper,we use PC and server in the laboratory to build a verification platform,and test and analyze the front-end page display and model training.The verification results prove the effectiveness of the motor condition intelligent monitoring system based on Federated learning.