构建基于脉冲星搜索数据库的FAST多波束干扰消除和快速射电暴实时触发系统

Radio Frequency Interference (RFI) will significantly affect the upcoming FAST pulsar drift scan, and also to other science data. For periodic pulsar search program, RFI will reduce our telescope sensitivity, and for single-pulse search, RFI will greatly increase the number of candidates, which need to be inspected by scientists or specialists one by one. We know that the elimination of RFI, Effectively, for the search of pulsar and Fast Radio Burst (FRB), are the key technical difficulties, and that matters the efficient and accurate FRB event trigger. But, there is no special success in the international in this area. Real-time RFI cancellation and FRB signal real-time triggering is conducive to the FRB events in a timely by multi-telescope, multi-band follow-up observation, and also help us capture the valuable observed characteristics of FRB in different bands..The project will use our already built pulsar-search database (by NAOC and GZNU) and multi-beam signal cross-validation technology, to improve the speed and accuracy of FAST site RFI elimination and FRB search program significantly. For other Observatory, our work will provide accurate and effective trigger signal. FAST multi-beam RFI interference cancellation based on pulsar search database and the establishment of real-time trigger system for Fast Radio Burst will help scientists reduce FRB triggering misjudgment rate and improve FRB detection capability.

射电干扰(RFI)的存在会显著影响即将开展的FAST脉冲星漂移巡天效果。对于周期性脉冲星搜索,RFI将导致望远镜灵敏度降低;对于单脉冲搜索,RFI会极大增加认证候选信号数量。有效消除RFI对于脉冲星和快速射电暴(FRB)搜索影响和高效准确的FRB事件触发,都是关键性技术难点,目前国际上还没有特别成功先例。实时RFI消除和FRB信号的实时触发,有利于对FRB事件进行及时的多望远镜、多波段跟进观测,捕捉FRB在其他波段的观测特征。.本项目针对FAST设备，拟采用已有的FAST脉冲星搜索数据库和多波束信号交叉检验技术，大幅提高台址RFI消除和FRB等暂现信号搜索的认证速度和准确性，为其他天文台提供精确有效的触发信号。基于脉冲星搜索数据库的FAST多波束RFI干扰消除和快速射电暴实时触发系统的搭建，将有利于科学家减少FRB触发误判率，增强FRB探测能力。

项目背景：.本项目在前期脉冲星搜索数据库的工作基础开展构建基于脉冲星搜索数据库的FAST多波束RFI干扰消除和快速射电暴检测系统研究，提升FAST在实时探测快速射电暴等具体科学目标的数据处理能力，实现国内学者利用中国设备，系统发现新快速射电暴的突破。.主要研究内容：.多波束巡天数据中RFI信号的交叉检验和消除、多波束中FRB候选体信号的交叉检验、构建搜寻数据库，提供数据分析服务功能。.重要结果：.提高并行化是检测搜索系统优化一种实现途径。针对现有搜索软件存在的问题，结合FAST数据搜索特点和实际应用，搜索系统可以从以下方面提高并行度并进行优化：任务并行、计算节点间并行、软件程序并行。（论文7，8）.多波束巡天数据中RFI信号消除模型与交叉检验可以有效提高候选体质量。各波束间数据的相似性，或者某些信号之间的相似性，可以检测、识别甚至提取一些特定成分的信息在算法的实现上，尽可能的利用多核心GPU的并行性是提高计算速度的有效方法。针对基于定义计算信号相似度方法运行效率低的问题,提出了2种CPU与GPU异构并行的方式处理数据,最高提速约7倍。（论文5，10）.通过GPU加速的单脉冲搜索过程及候选体数据库，可以高效地进行FRB检测。开发了GPU的单脉冲搜索流程（GPU based single-pulse search pipeline,GSP）并将其应用于CRAFTS数据搜索，计算效率较PRESTO的单机版本提高了120倍加速，较PRESTO的MPI版本提高了约60倍。（论文1，2，3，4，6，8）.FRB数据统计。快速射电暴平均色散量为584.5pc·cm^-3，暗示着快速射电暴来自河外。通过最佳拟合估算红移幂律分布谱dN/dFobs=4.14±1.30×Fobs-1.120 55 sky^-1·day^-1。（论文9）