Efficient Mesh Space Mapping Optimization for Tunable Filters Incorporating Structurally Simplified Coarse Mesh Model Without Tunable Elements

Mesh space mapping (MSM) is widely recognized as a popular surrogate-based optimization approach for expediting electromagnetic (EM) design, particularly in cases where traditional equivalent-circuit coarse models are not readily available for standard space mapping (SM). For practical microwave tunable filter design, the EM optimization becomes very time-consuming even using standard MSM, since the tunable elements usually require a much denser mesh than the pure cavity. This article introduces a novel MSM technique to address the challenges associated with optimizing tunable filters in EM simulations. This approach incorporates a coarse model that not only uses a coarse mesh but also simplifies the structure by removing all tunable elements within and between resonance cavities. The structurally simplified coarse mesh model without tunable elements can have much sparser mesh and subsequently can achieve much faster coarse mesh model optimization. Since it is challenging to find the mapping relationship of the design variables between the structural simplified coarse mesh model and the fine mesh model, a systematic model simplification algorithm, design variable mapping method, and parallel multipoint surrogate training technique are proposed. During the coarse mesh model optimization, the mesh deformation technique is used to guarantee the continuity of the coarse model EM response with respect to geometrical value change and to increase the coarse model optimization convergence. Consequently, the proposed MSM optimization for tunable filters incorporating a structurally simplified coarse mesh model without tunable elements can be realized to speed up overall EM design optimization.