Research On GPU Acceleration Of Particle-in-cell Simulation In Ion Propulsion

The human desire for space exploration continues to promote the advancement of space science and technology.This advancement has expanded the scope of human knowledge of the world and brought convenience to human life.The in-depth exploration of space and the use of its space is one of the key development directions of humanity’s future science and technology.Deep space exploration requires the use of advanced technology propellers.Ion propulsion has the advantages of long life,high specific impulse,convenient thrust adjustment,etc.It has been widely used in spacecraft such as communication satellites and deep space probes.Including spacecraft attitude control,position maintenance,orbit maneuvering and interplanetary flight.In order to describe in detail the core parameters of the ion thruster and evaluate and improve the performance of the Ion propulsion,researchers increasingly use numerical simulation methods(currently based on the PIC method)to perform the physical mechanism inside the ion thruster.the study.However,the PIC method is inefficient.By studying the application of GPU high-performance computing in PIC simulation,the PIC simulation efficiency can be greatly improved.This dissertation is based on the self-developed ion accelerator PIC/MCC software BUMBLEBEE-EP developed by the University of Electronic Science and Technology of China,and implements the parallelization of its charged particle and electromagnetic field interaction solver,and applies it to the study of the physical mechanism of the Ion propulsion.The main work and innovation of this article include:1.The research background,research significance,working principle,and main research directions of the ion thruster are summarized.It is concluded that the physical mechanism of the discharge chamber is one of the main research directions of the ion thruster.The study of the discharge cell is mainly a numerical simulation method.It is proposed to use GPU to increase the simulation efficiency in large-scale simulation problems.2.Based on the existing GPU architecture and CUDA programming technology,combined with the characteristics of the BUMBLEBEE-EP software interaction algorithm between the charged particle and electromagnetic field,the various modules of the solver are implemented in parallel,mainly including the field solution module and the motion solution module.And current source solving module;3.Optimization of parallel programs,mainly including memory optimization(mainly constant memory,shared memory,registers),thread structure optimization,and atomic operation optimization.Among them,a new reduction method is proposed in atomic operation optimization to optimize the current source solution.Finally,the parallel speedup ratios of various optimized versions are given,and the key problems that restrict the accelerating efficiency are analyzed.4.Verify the consistency of serial and parallel results,and give the error of the cold cavity field,the error of the hot cavity field,and the movement error of the hot cavity particles.The maximum error is less than 0.2% and the error sources are analyzed in detail,ensuring the parallel program.Correctness5.The physical analysis of the electron magnetic confinement problem in the ion thruster is performed using a parallel program to investigate the constrained phenomenon of the applied magnetic field to the electron motion.