An Overview of Large Range Modeling Using Decomposition Method

When modeling electromagnetic parameters, the range of geometric parameters varies greatly and modeling with previous techniques can be very difficult. In order to solve this problem, this paper reviews a systematic and automated algorithm based on second-order derivative information. The algorithm divides a flat region into a few large sub-regions and a nonlinear region into a large number of small sub-regions. When the geometric parameters of the sub-region change from one sub-region to another, the electromagnetic response discontinuity will be found on the boundary of adjacent models, which also leads to the problem of multidimensional discontinuity. This paper reviews a modified method to solve this problem to obtain a continuous global model and in this paper, three methods of parallel data generation, parallel sub-model training and parallel sub-model modification are used to accelerate the model development process and finally get the desired model quickly and accurately.