Research Of Service Robot Cloud Fault Diagnosis Based On Machine Learning

With the development of artificial intelligence technology,service robots have gradually entered millions of households.It provide many conveniences for people's lives.Service robots work in the home environment,and their users include the elderly and children,so the safety of service robots is'very important.In order to improve the safety of service robots,this paper designs a cloud-based fault diagnosis system,and studies the fault method based on machine learning.On the basis of summarizing the fault diagnosis technology of robots,we designe and implement the cloud fault diagnosis framework,cloud fault diagnosis service,fault feature selection and fault diagnosis model according to the fault analysis of the motion control system.In the aspect of system framework design,the motion control system of service robot is designed and fault analysis is carried out.On this basis,the system framework of "cloud-robot" is designed.Firstly,the robot collects sensor data and uploads it to the cloud through WebSocket,and preprocesses the data in the cloud.Then,feature selection is implemented based on the integrated tree model,and the selected data input into the fault diagnosis model to calculate the fault diagnosis results.Finally,the cloud fault diagnosis results are fed back to the robot through the cloud interactive interface.In the aspect of platform building,fault diagnosis data set is established,and the cloud database is designed and implemented.four-tier fault diagnosis cloud service architecture is constructed,and Tornado framework is used to program the fault diagnosis cloud service.In the aspect of fault feature selection,firstly,data is standardized and backward differential pretreatment is carried out.Secondly,fault diagnosis and fault feature selection based on random forest and gradient boost decision tree algorithm are studied.Finally,RF-GBDT fault feature selection algorithm is proposed,which combines the advantages of random forest and gradient boost decision tree.Experiments show that the proposed RF-GBDT algorithm can effectively realize feature selection and improve the accuracy of fault diagnosis.In the aspect of fault classification algorithm,aiming at the problem that traditional fault diagnosis algorithms only focus on the current state data,a fault diagnosis method based on time series neural network model is proposed.Firstly,sliding window is used to generate time series samples.Then,an improved mixed fault diagnosis model based on GRU neural network is proposed.The time series features are extracted by GRU neural network and the current data state features are extracted by BP neural network.Several groups of comparative experiments show that the improved fault diagnosis model based on GRU neural network has higher accuracy.Finally,the system test in the actual scene shows that the cloud-based fault diagnosis system designed in this paper can achieve the expected fault diagnosis target,and can effectively improve the safety of the robot.