研究目的
To introduce a generalized Pareto ranking bisection algorithm (GPRBA) for low-cost multi-objective design optimization of antenna structures, allowing for handling any number of design goals.
研究成果
The GPRBA algorithm allows for handling an arbitrary number of design objectives, demonstrated using a monopole antenna optimized w.r.t. three goals: minimization of in-band reflection, the structure footprint, and realized gain variation. A uniform Pareto front representation has been obtained at the cost corresponding to only 116 evaluations of the high-fidelity antenna model.
研究不足
The original bisection algorithm was limited to only two objectives. The computational complexity of population-based methods may be very high, which is prohibitive if the device performance is evaluated using high-fidelity EM simulations.
1：Experimental Design and Method Selection:
The GPRBA algorithm performs bisections of n-dimensional simplexes in the feature space and is equipped with a better poll-search procedure. It starts from the initial set of designs optimized with respect to individual objectives using a coarse-discretization EM model.
2：Sample Selection and Data Sources:
The algorithm is demonstrated using a UWB monopole antenna with three figures of interest: structure size, reflection response, and gain variability.
3：List of Experimental Equipment and Materials:
The antenna is implemented on Rogers RO4350 substrate. EM antenna models are implemented in CST Microwave Studio.
4：Experimental Procedures and Operational Workflow:
The algorithm involves iterative partitioning of the intervals connecting previously found designs and executing a Pareto-ranking-based poll search. The initial Pareto set representation is refined to the level of the high-fidelity EM model.
5：Data Analysis Methods:
Pareto ranking is calculated for all perturbed designs, and the best design is selected based on the lowest ranking value and proximity to the Pareto front normal vector.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
