Return
Accelerating Time-Evolving Multiphysics Uncertainty Quantification of Millimeter Wave Components Using Reusable Preconditioning
DOI:10.1109/tmtt.2026.3694111.png)
Abstract
En 中文
Transient electromagnetic (EM)–thermal– mechanical uncertainty quantification (UQ) is computationally demanding because it requires repeated solutions of many closely related EM systems across time steps, coupling iterations, and random samples. This article presents an efficient finite-element multiphysics framework that couples the frequency-domain Helmholtz equation with time-domain heat conduction and evaluates thermomechanical responses through small-deformation thermoelasticity. To avoid repeated remeshing in large-sample studies, thin metallization layers are modeled by a reduced-dimensional thermal formulation on a fixed mesh. To accelerate the repeated EM solving process, a reusable LU-based preconditioning strategy is incorporated into the coupling loop. UQ is then performed by nonintrusive polynomial chaos expansion (PCE), from which confidence bands and Sobol sensitivity indices are extracted for both time- and frequency-domain quantities of interest (QoI). Numerical results on a substrate-integrated waveguide (SIW) filter and a silicon-based D-band diplexer confirm that reusable preconditioning and fixed-mesh thin-layer modeling can substantially reduce the computational cost of large-sample transient UQ. This makes the proposed framework suitable for reliability-oriented analysis and design of millimeter-wave components operating under high-power and high-temperature conditions.
Keywords:
Electromagnetic-thermal coupling
finite element method (FEM)
polynomial chaos expansion (PCE)
preconditioning
Sobol sensitivity indices
thermomechanical analysis
thin-layer thermal model
transient multiphysics simulation
uncertainty quantification (UQ)
Journal
IF:
4.5
Papers:
593
Citations:
3.5W

