Inverse design of material, structure, and process for dielectric properties of additively manufactured PLA/BaTiO3 polymer composites

Abstract

Additive manufacturing holds excellent potential for engineering highly integrated and miniaturized microwave components with targeted dielectric constant and loss tangent for the desired operating frequency bands. However, searching for optimal combinations of material, structure, and process that produce components with desired dielectric properties remains a nontrivial task due to the large design space and complex relationship. This paper proposes an inverse design method for tailoring the dielectric properties of additively manufactured PLA/BaTiO₃ polymer composites by exploring their multidisciplinary design space, including material composition, infill structure, and process parameters. This approach utilizes the theoretical Yamada and first-order models to predict the material-property and structure-property relationships, respectively. For given material composition in PLA/BaTiO₃ polymer composite, the few-shot learning (FSL) model with model-agnostic meta-learning (MAML) algorithm is trained to predict the targeted dielectric properties on the integrated structure-process-property relationship of additively manufactured PLA/BaTiO₃ polymer composites. The inverse design approach of the theoretical Yamada model provided a relative error below 3.25 % on the dielectric constant and 6.15 % on loss tangent in material composition ranging from 5 wt% to 80 wt% of BaTiO₃ filler, achieving 13.913 ± 0.028 and (4.373 ± 0.023) × 10⁻². The inverse design approach of FSL with the MAML algorithm had a relative error of less than 4.93 % and 4.73 % on the dielectric constant of linear and gradient patterns in the structure-process-property function. This work successfully proposes the FSL model with the MAML algorithm to inversely predict targeted dielectric properties on structure-process-property integration in PLA/BaTiO₃ polymer composites, which can accelerate the inverse design cycle and enlarge the controllable property range of additively manufactured microwave components.