Improved Dual-Primal Domain Decomposition Method for the Angular Radar Stealth Analysis

Angular monostatic radar cross-section (RCS) analysis plays a critical role in assessing the stealth performance of aircraft. However, it becomes computationally intensive when applied to large-scale finite-element (FE) models. This challenge stems from the need to solve linear systems for multiple incident waves, or multiple right-hand sides (RHSs). This paper proposes two improved variants of the electromagnetic dual-primal FE tearing and interconnecting (FETI-DPEM) method to accelerate the analysis by exploiting the linear dependency among the RHS vectors. The first variant, recycling FETI-DPEM (rFETI-DPEM), is a sequential method that reuses Krylov subspace information from previous solutions. The second variant, block FETI-DPEM (bFETI-DPEM), initially employs rank-revealing QR decomposition to reduce the dimensionality of the RHS block and subsequently solves the resulting reduced system using a block iterative solver. The numerical results demonstrate that both rFETI-DPEM and bFETI-DPEM preserve the accuracy of the original FETI-DPEM while substantially reducing the computational time. Among the proposed variants, the bFETI-DPEM method exhibits superior robustness, achieving a speed-up factor of up to 19.6\((\times)\) in large-scale missile analyses. This framework thus provides a highly efficient and robust solution for large-scale radar stealth analyses in complex aerospace structures.