Research On The FAST Node Displacements Prediction Based On Support Vector Machine

FAST(Five-hundred-meter Aperture Spherical Radio Telescope),the most sensitive and internationally largest spherical radio telescope,has significant impact on deep space exploration,environmental research and national security.Since researching on the FAST node displacement control theoretically and practically impact the design and implementation of FAST in industrial control fields,in this project the background,only for FAST node real-time dynamic positioning and control,FAST main reflector can be successfully completed observation mission,thus establishing reasonable sense of motion displacement prediction model node is very important.First,the statistical learning theory and support vector machine related theory summarized,with an emphasis on support vector machine and kernel function basic principles of learning and research;Secondly,the FAST-node system structures and active reflection surface Deformation strategy analysis,focusing on the displacement strategy FAST node,controlling factors influence the displacement principle and accuracy were analyzed to determine the establishment of the control model input and output;once again discussed the SVR established FAST nodal displacement of support vector machine regression prediction model;and finally,for the support vector machine regression model parameter selection method of limitations,respectively,with PSO and genetic algorithms to optimize the selection parameters of the model,two optimization algorithms have advantages shortcomings,the use of genetic algorithms PSO improvements and set out specific implementation steps to improve the algorithm and processes.FAST cable network with multiple input node movement,nonlinear,time-varying,large inertia and other characteristics,the use of support vector machine learning methods,the establishment of FAST node displacement prediction model.Through simulation analysis,the effectiveness and feasibility of establishing a support vector machine regression FAST node model.The main innovation of this paper is to choose the appropriate node displacement prediction model FAST-FAST node displacement prediction model based on support vector machine regression,the parameters of the model selection method improved by simulation verify the validity and feasibility of the model.In this paper,the establishment of FAST displacement prediction model based on support vector machine node,using particle swarm optimization and genetic algorithm parameter selection methods to improve the model,and verified by simulation vector machine algorithm built FAST nodal displacement control model based on support the feasibility and effectiveness of this initiative will be FAST adjustment reflecting surface of the entire network research has important practical significance.