快速射电暴实时搜寻技术研究

The Fast Radio Burst (FRB) is one kind of mysterious pulse bursts in the universe discovered in recent years. The reproduction probability of the FRB is extremely low. We are still puzzled about its physical nature. The research on the FRB faces some difficulties, such as very small number (only 16 cases) of FRBs known, few (only 1 case) following up observation information. The real-time FRB searches can solve the difficulties mentioned above. If the real-time FRB searches are used, the data recording hard disk space can be saved greatly, which makes it is possible to discover more cases of FRBs by all-day ceaseless monitoring the universe. On the other hand, follow-up observations can also be arranged on time under the help of real-time FRB search techniques. But it is not easy to carry out real-time FRB searches, as many challenges should be faced. Under the support of the platform of the Shanghai TianMa Radio Telescope, we will do a deep study on how to overcome the challenges in FRB searches in this project, including: (1) techniques on how to build the real-time FRB searching software and hardware platform; (2) techniques on how to eliminate radio interference in real-time;( 3) parallel accelerate dispersion techniques and related algorithms using GPU together with CPU; (4) how to use the pattern recognition techniques to identify the real FRB signals. In order to improve the efficiency of software development and make sure its intelligence and portability, we will call the “Open Computing Language” (OpenCL) and “Machine Learning for Astronomy ” (AstroML) libraries. Ultimately, we will build a platform that can be used to do the real-time FRB searches. This project will provide valuable experience on carrying out FRB real-time searches with the multi-beam receiver and the antenna array in the future.

快速射电暴(FRB)是宇宙中新发现的一类复现概率极低的神秘脉冲暴发,其物理尚不清楚。对它的研究面临样本数目少(仅16例)、缺少暴发后及时观测信息(仅1例有)等困难。FRB实时搜寻可有效解决上述困难。因为它节约硬盘，可长期不断地执行搜寻任务，其在线警报功能也保证了后续观测的及时安排。但FRB实时搜寻面临各种挑战。本课题依托上海天马射电望远镜平台，对FRB实时搜寻所面临的挑战开展针对性技术攻关，主要包括：①FRB实时搜寻软、硬件平台搭建；②实时消除射电干扰技术；③GPU和CPU协同并行加速消色散技术及算法；④备选体智能认证技术。为了提高开发效率并保证系统的智能性和可移植性,我们将调用最近流行的“开放式计算语言”(OpenCL)和“面向天文研究的机器学习和数据挖掘软件库”(AstroML)。最终实现该平台对FRB实时搜寻的目标。本项目将为后续多波束接收机及望远镜阵列的FRB实时搜寻积累经验。

快速射电暴(FRB)是宇宙中新发现的一类复现概率极低的神秘脉冲暴发,尚不清楚其物理本质。目前对其研究面临样本数目少、缺少暴发后及时观测信息等困难。快速射电暴实时搜寻可有效解决上述困难。在本项目支持下，我们依托上海天马射电望远镜平台对FRB实时搜寻所面临的挑战开展针对性技术攻关，成功搭建了快速射电暴实时搜寻软、硬件平台。该平台由模数转换单元（ADC）、Xilinx可编程逻辑门阵列（FPGA）、高性能服务器、图形处理器（GPU）运算单元、数据记录单元组成。针对射电观测中的干扰问题，我们研究了如何利用自适应滤波和零色散滤波技术实时抑制有关干扰。针对色散搜寻中对大量数据进行高密度运算问题，我们研究了GPU和CPU协同并行加速消色散技术及算法，其中包括暴力搜寻算法和树形消色散算法。针对如何准确辨认射电干扰和真实天体信号这一难题，我们综合考虑疑似信号的信噪比随色散量变化情况和脉冲形态等对其进行判断。经过不断调试和优化，我们顺利解决了大量数据传输中的丢包问题、大数据流快速色散搜寻问题、有效甄别干扰和信号问题等。目前，该平台已经能够顺畅运行于上海天马射电望远镜现场。由于该平台可有效甄别天体信号和射电干扰，并且仅记录一系列疑似天体暴发信号相关数据，从而有效缓解硬盘记录压力。在设置快速射电暴搜寻所需恰当色散量搜寻范围情况下，通过对和快速射电暴信号相似的脉冲星巨脉冲、旋转射电暂现源观测，我们充分测试了该系统的性能和稳定性。该设备可长期不断地执行实时搜寻任务并具有警报功能，因而也保证了后续观测的及时安排。此外，我们将快速射电暴的搜寻工作研究从单天线观测数据拓展至甚长基线干涉仪（VLBI）观测数据，通过自相关谱分析、互相关谱分析以及高分辨率成图解决快速射电暴的搜寻定位难题。