LLNL Home About Engineering Site Index Contact Us |
Modeling and Simulation |
|
References Contact Us |
Broadband Radiation and Scattering Electromagnetic phenomena are a central thread through much of modern engineering. There are two fundamental classes of applications: open region problems (where the energy propagates in unbounded space) and closed region problems (where the energy is guided by a waveguide structure or cavity). This effort strives to enhance our computational electromagnetics (CEM) capability in broadband radiation and scattering in open regions. Broadband fields consist of energy with a robust spectrum, and include applications such as electromagnetic interference and electromagnetic compatibility noise analysis, broadband radar, and accelerator wakefield calculations. LLNL analysis codes are limited by the accuracy of radiation boundary conditions (RBCs), which truncate space. We have developed improved RBCs by extending the perfectly matched layer (PML) approach to non-Cartesian meshes, and by developing discrete-time-domain, boundary-integral techniques that are compatible with high-accuracy, finite element methods and capable of arbitrary accuracy. These approaches have been compared to the traditional first-order absorbing boundary condition for a variety of radiation and scattering problems. The ultimate deliverable is an enhanced CEM capability that can provide accurate and efficient computational solutions to broadband radiation and scattering problems. The algorithms for improved RBCs will be incorporated into LLNL's existing EMSolve code. The result will be a 10- to 1000-fold improvement in the accuracy of simulations. Improved algorithms and our existing high-performance computer hardware will place LLNL's CEM activity among the top capabilities in the world. This research and the resulting capability will be documented in appropriate peer-reviewed publications. Electromagnetics is a truly ubiquitous discipline that touches virtually every major LLNL program. Our work supports the national security mission by reducing the time and money spent in building and testing existing programs. It will enable computer simulations for new devices and systems, performance analysis of systems critical to nonproliferation efforts, and the design of micropower impulse radar and other microwave systems.
FY2006 Accomplishments and Results Figures 1 to 5 are samples of our FY2006 CEM work. We have completed the development of the parallel hybrid finite element boundary element code. Several different formulations for RBCs are now available, including boundary conditions based on the electric field, magnetic field, or both at the boundary. A formulation using both the electric and magnetic fields was found to eliminate late-time stability issues due to interior resonances. We have collaborated with a professor at the University of Washington who is an expert on time-domain integral equations. It was found that the boundary element time-step could be different than the finite element time-step. By sub-cycling the boundary element computations at some multiple of the finite element time-step, large improvements in speed were observed, especially for very large meshes. This sub-cycling allows for a trade-off between speed and accuracy. For a z-oriented pulsed dipole problem (with a known solution), we have achieved almost a 1000-fold increase in accuracy using our hybrid code (Fig. 3). The conventional Absorbing Boundary Condition (ABC) method fails to converge when applied on a spherical grid with a maximum radius of 1. In contrast, the hybrid solution shows good accuracy. Figures 4 and 5 show the behavior of our computation costs and error when sub-cycling of the hybrid boundary element calculation is applied.
|
Lawrence Livermore National Laboratory
7000 East Avenue • Livermore, CA 94550 |
Operated by Lawrence Livermore National Security, LLC, for the Department of Energy's National Nuclear Security Administration |