Research On Data Processing Method Of Pulsar Search

In September 2016,the Five-hundred-meter Aperture Spherical radio Telescope(FAST)has been built and the pulsar survey will be one of the important scientific projects for FAST.It is expected that the amount of annual survey data will generate hundreds of millions of pulsars candidates,rely on artificial identification and selection which cannot meet the requirement of the aging needs,which makes it more impossible to accomplish further requirements of real-time data processing.Therefore,this paper will summarize and forecast the existing select methods in pulsar search data processing,and realize the automatic verification of candidate,as well as the FAST team will be based on the pattern recognition screening candidates work,they will use the simulation observation data program simPulseTrain to generate simulation candidates,to achieve the screening work,here we will optimize the method in simPulseTrain,and realize the parallelization of the program.First of all,this article combs the FAST indicators and observation performance,introduced the pulsar survey mode.Secondly,we introduce the data processing process of pulsar search,and summarize the application of artificialintelligence method in pulsar candidate screening,statistical analysis of pulsar candidate screening methods has been carried out,and look forward to the FAST pulsar candidate screening work.In addition,the traditional pulsar candidate verification is done by manually querying the data in PSRCAT,its efficiency is relatively low.This article will use the Web and database technology to achieve the pulsar candidate verification automation,and the development of the corresponding Web interface.Using the Parkes multi-beam pulsar survey in the FAST area data processing(about 60G),access to about 30,000 candidate documents for verification.Finally,27 known pulsars were obtained,corresponding to 247 pulsar candidates,the results show that the proposed method can match the candidate of the corresponding cycle in the calculation.The FAST team simulates the generated observations for simulating FAST pulsar drift surveys and estimating the sensitivity of the survey,and they will soon be carried out to generate specific types of candidates for machine learning training set,to achieve the pulsar candidate screening intelligent.Simulation program—simPulseTrain,it is in the simulation of the dispersion signal needs more channels,theprogram uses a single-threaded approach,it is running too slow.Therefore,the simulation method needs to be parallelized.In this paper,the simulation method of pulsar observation data is summarized and the simulation method is optimized.The parallel program is designed on the multi-core CPU to realize parallel acceleration,to achieve parallel acceleration.The final experimental data show that the use of CPU multi-core parallel can improve the performance of the algorithm to enhance about 14 times.