基于GPU集群的空间VLBI射电源条纹搜索研究OA北大核心CSTPCD

The fourth phase of the Lunar Exploration Project will launch the Chang'e-7 relay satellite,carrying a 4.2-meter parabolic telescope.Together with ground-based radio telescope,it will form the first Lunar Orbit Space VLBI system.Space VLBI is influenced by the relay satellite's orbit perturbations and onboard equipment delays,and the predicted time delay model for radio source may not effectively guide the normal operation of the correlator.It is necessary to employ a fringe search based on actual measurement signals to find a high-precision time delay model that meets the requirements.Currently,fringe searches are accelerated using CPU cluster,but the search time is prolonged,lasting up to 2700s,which fails to meet the practical engineering requirements.A space VLBI radio source fringe search algorithm based on GPU cluster is proposed.The research involves studying task distribution for load balancing on GPU cluster and parallel implementation of corre-lation modules and residual value search modules.Validated with RadioAstron observation data,the processing speed of GPU cluster has improved about 27.0 times compared to CPU cluster.The research outcomes can be applied to subsequent space VLBI projects.