High-Performance Numerical Method Of Multi-Physics And Its Applications

As the developments of emerging areas in high-power microwave (HPM), deep-space exploration and plasma physics, people have much more desires to get a profound understanding on the interaction between electromagnetic field and other physical fields. This requirement sharply promotes the study of numerical algorithms of multiphysics, which covers a variety of disciplines including electromagnetism, physical electronics, quantum mechanics, biological electronics, thermal and elastic mechanics, computer science, and many other subjects. Currently, the research of numerical methods of multiphysics is still at the beginning of primary stage. Thus, it has encountered many severe challenges, such as the controllability of computing precision, the efficiency challenge in large scale simulation, the problem of interaction between multiphysics fields, and so on. Funded by two national projects, we have devoted to developing a high-performance numerical method of multiphysics and its application, and have made many innovative achievements in both the fundamental theory and practical application.Part 1:In the aspect of fundamental theory, a number of innovative researches have been presented in this dissertation to achieve a faster, smarter and more accurate numerical method of multiphysics, including:1. To develop a novel numerical method of multiphysics, we propose a time-domain finite-integral theorem (TDFIT), which has combines the merits of classic finite difference time domain (FDTD) method and finite-integral theorem (FIT). Because the proposed method can effectively eliminate the staircase error brought by the spatial discretization, the precision of TDFIT is significantly higher than that of traditional FDTD method. In addition, different from the classic FIT method, there are not complex matrix operations in the proposed TDFIT. Thus, the implementation of TDFIT can be much more concise and efficient than conventional FIT algorithm, especially in the process of parallel programming. Based on this novel and efficient numerical method, several fundamental multi-physical equations have been successfully solved to realize the numerical simulation of multiphysics. The relevant achievements have been published in IEEE Trans. Electron Devices, IEEE Trans. Plasma Science and IEEE Antennas Propag. Magazine.2. To improve the accuracy of the new numerical method in modulating multiphysics, a conformal technology of electromagnetism-particles is proposed in this dissertation. The conformal technology of electromagnetism includes the uniform conformal iteration and the extraction of electromagnetic conformal information. The conformal technology of particles includes the collision model of particles and the extraction of particle conformal information. These conformal techniques can effectively eliminate the spatial discretization error of traditional numerical methods, and thus can significantly increase the precision of multiphysical simulation. The relevant achievements have been published in IEEE Trans. Antennas Propag., and IEEE Trans. Electron Devices.3. To accelerate the simulation speed and increase the simulation ability in modulating multiphysics, a new MOG (MPI- OpenMP- GPU) hybrid parallel technology is presented. This hybrid parallel technology has made most of the natural parallel-programming feature of our TDFIT method, and has integrated the merits of existing parallel techniques. For instance, its MPI solver can efficiently utilize the distributed memory to simulate the electrically-large target; its OpenMP solver can make full use of CPUs’ multithreading resource to realize the parallel acceleration in task level; while the parallel acceleration in data level can be efficiently complete by its GPU solver. To perfectly balance the computational efficiency in each parallel part, a new adaptive load-balance scheme has also been given.4. To make the numerical method much more intelligent in simulating multiphysics, a new adaptive steady-state criterion has been presented. Compared with previous steady-state criteria, the new adaptive steady-state convergence strategy is more intelligent in determining whether the current iteration reaches the steady state or not. To realize this intelligent process, we have innovatively defined an attenuation coefficient and a stability coefficient, which can be used to overcome a series of defects existing in the traditional criteria, such as the premature convergence, non-monotonicity and non-convergence challenges. The relevant contents have been published in IEEE Trans. Microw Theory Tech.Part 2:In addition to the theoretical innovation, we also actively apply the latest research achievements into the real-world applications, including:1. In Metamaterial design, there are serious grid-dispersion error and coupling-response error for zero-volume lumped element (ZVLE) model to simulate metamaterials, which are mounted with lumped components. To figure out the error sources, we have delved into the classic filed-circuit hybrid algorithm. Based on our past research experiences, a novel finite-volume lumped element (FVLE) model is proposed to eliminate the grid-dispersion error and coupling-response error in traditional ZVLE model. Many measured results have verified the superiorities of the new filed-circuit hybrid scheme. The relevant results have being scheduled to be published in IEEE Trans. Microw Theory Tech.2. In the research area of HPM devices, a conformal hybrid algorithm combining the time-domain finite-integral theorem and particle-in-cell method (TDFIT-PIC) is presented to efficiently and accurately simulate the interaction electromagnetic field and charged particles. Based on the TDFIT-PIC hybrid method, we detail a novel approach to numerically solve the Maxwell’s equations and Newton-Lorentz equations. Meanwhile, we successfully expand the classic Vaughan secondary-electron-emission (SEE) model, and incorporate it into self-developed TDFIT-PIC code package. We also propose a new power-to-amplitude transformation (PAT) to ensure that the EM-Particle interaction can be analyzed by the numerical method and measured method in the same exciting condition. After the experimental verification, it is shown that the numerical results can match the measured results very well. The relevant contents have been published in IEEE Microw Wireless Compon Lett., and IEEE Trans. Electron Devices.3. In the research area of multi-carrier communication, we have proposed a new field-circuit hybrid solution to numerically analyze the phenomenon of passive intermodulation (PIM), which is frequently discovered in high-power multi-carrier communication. In this novel field-circuit hybrid solution, the part of electromagnetic field is simulated by TDFIT method, and the part of nonlinear circuit is analyzed with ZVLE model. The coupling between the EM field part and the circuit part is achieved through the generalization of Ampere’s circuital law. To solve the multiscale problem in PIM analysis, we obtain the gap voltage tactfully by using the surface magnetic field, and develop a new approach to incorporate the nonlinear lumped elements with the help of conformal technology. Above achievements make it possible to analyze PIM numerically.