Machine learning aided optimization of multilayer lattice for microwave absorption

In response to the increasing demand for lightweight stealth structures, this research proposed a novel optimization strategy for the microwave-absorbing performance of multilayer lattice structures based on the octet-type lattice structure. Initially, unit cells with varying densities were constructed, and their S-parameters were simulated across the 0.5–18 GHz frequency range to retrieve the corresponding equivalent electromagnetic parameters. The relationship between lattice volume fractions and equivalent electromagnetic parameters was subsequently established using the Random Forest algorithm. An optimization model was then formulated, employing the volume fraction of each layer as design variables, the total volume fraction as a constraint, absorbing bandwidth as the objective, and utilizing particle swarm optimization as the optimizer. The optimization results revealed that the optimal lattice structure exhibits a non-linear volume fraction gradient across the layers. Compared to traditional uniform and linearly graded lattice structures, the optimized configuration achieved a markedly broader microwave-absorbing bandwidth within the calculating frequency range. The proposed approach, thus, provided an effective framework for the design of advanced lightweight microwave-absorbing structures.