Using Fdtd Rcs Computation Of Geometry Processing And Parallel Computing Research

The counterwork between stealth and anti-stealth makes an important contribution to modern electric warfare. And the knowledge of radar targets scattering characteristics is the foundation for stealth weapons and anti-stealth radar systems design. Using FDTD (Finite-difference time-domain method) method to calculate RCS (Radar Cross Section) has the advantage of high precision, but also suffers from computation complexity and high memory consumption. In this thesis, we proposed an object oriented design for the FDTD based RCS calculation system, which alleviated the difficulties associated with original method.This thesis set up the parallel system framework, which has the function of model loading, model display, voxelization and result contrast. The system has the advantage of easy to control and update, as well as high efficiency. The prospect for the application of our system is broad.This work is introduced two parts: voxelization of complex models and cluster based computation. The voxelization algorithm is migrated to adapt to cluster. Then we made a modification to the FDTD calculating module so it can run on the cluster. For the voxelization module, a 3D scan line algorithm is adopted, BSP (Binary Separate Partition) is also introduced to accelerate voxelization, and then we take the strategy of out-of-core to reduce memory consumption. By the above means, we are able to voxelize complex models. For the FDTD based RCS calculation, we developed a cache system to assure that the memory needs can be met.Several parallel computing topics are discussed in this thesis, such as: design of parallel program, assignment partition and data communication, we also analyzed the module of model voxelization. The general serial program is migrated to the cluster. Utilizing the characteristics of RCS, we made minor modifications to the FDTD module to transplant it to the cluster, and we got very good results.