Research On Performance Prediction Method Of Time-space Two-domain Clustering For Fast Hydraulic Actuator Cluster System

Aiming at the problem of lack of reference samples and performance prediction in the reliability of FAST hydraulic actuator cluster system,this paper proposes a clustering method using the big data of FAST health monitoring system to solve the problem of FAST hydraulic actuator cluster system reference samples.At the same time,it provides research ideas for the research on small sample problems and performance prediction of similar complex systems.The main work of this paper is as follows:(1)The investigation of the research background of the FAST radio telescope leads to the research purpose and significance of this paper,and has an in-depth understanding of the research status of small sample reliability prediction and clustering algorithms at home and abroad,and proposes the main research content and research of this paper plan.(2)In this paper,the research object of the FAST hydraulic actuator used in the cluster system theory is introduced in detail,introduced the definition of the cluster system,subsystem and family system and the application of the cluster system theory to the FAST system.The current common clustering algorithms are studied,and the advantages and disadvantages of different clustering algorithms are analyzed,as well as their respective application scopes and evaluation indexes of clustering algorithms.(3)A polar coordinate spatial clustering method for cluster systems is proposed.In view of the reference sample problem of FAST hydraulic actuator cluster system,data preprocessing of FAST health monitoring system is carried out,polar coordinate system is established,and the actual position of actuators in FAST hydraulic actuator cluster system is corresponding to the position in polar coordinate system.Spatial clustering of FAST hydraulic actuators is carried out,and the clustering results show that most of the problem of actuator reference samples has been solved,but there are still some hydraulic actuators in the class without reference samples.(4)A time domain clustering method for constructing historical virtual samples is proposed.Aiming at the problem that there is still no reference sample for hydraulic actuators in some classes of spatial clustering,a method of re-clustering virtual sample data for hydraulic actuators by introducing time-domain data is proposed.The time domain data of single hydraulic actuator is preprocessed,and the time domain clustering algorithm of hydraulic actuator is selected and clustered.To solve the problem that some special hydraulic actuators cannot generate reference samples through spatial clustering.(5)The spatial-temporal coupling cluster analysis strategy is constructed,and verified and optimized in the performance prediction of FAST hydraulic actuator cluster system.By combining the time-domain data to construct virtual samples with the spatial-temporal data,the spatial-temporal two-domain coupling clustering is constructed,and the spatial-temporal two-domain clustering is carried out for the FAST hydraulic actuator cluster system,so as to solve the small sample problem of the FAST hydraulic actuator cluster system due to missing fault data or zero sample problem.The spatiotemporal two-domain clustering is validated by the performance prediction of hydraulic actuator.According to the performance prediction results of hydraulic actuator,the performance prediction scheme is optimized to improve the accuracy of performance prediction.