Approximate Direct Inverse For Embedded Domain Decomposition Method

The embedded domain decomposition method (DDM) is a powerful tool for analyzing complex electromagnetic applications. It can decompose a computational domain into non-conformal subdomains and provide great flexibility in modeling and optimizing add-on components. However, convergence issues are found for embedded DDM, particularly at low frequencies. To overcome the problem, we propose approximate direct inverse (ADI) based on the principle component analysis (PCA) of the system equations. The ADI is utilized as an effective preconditioner ensuring robust convergence of the embedded DDM system.