Constrained composite Bayesian optimization for rational synthesis of polymeric particles
*ab
Maryam
Parhizkar,
b
Abstract
Polymeric nanoparticles have critical roles in tackling healthcare and energy challenges with miniature characteristics. However, tailoring synthesis processes to meet design targets has traditionally depended on domain expertise and trial-and-error. Modeling strategies, particularly Bayesian optimization, facilitate the discovery of materials with maximized/minimized properties. Based on practical demands, this study integrates constrained composite Bayesian optimization (CCBO) to perform target-value optimization under black-box feasibility constraints for by-design nanoparticle production. In a synthetic problem that simulates electrospraying, a representative nanomanufacturing process, CCBO avoided infeasible conditions and efficiently optimized towards predefined size targets, surpassing the baseline methods and state-of-the-art optimization pipelines. CCBO was also observed to provide decisions comparable to those of experienced experts in a human vs. BO campaign. Furthermore, laboratory experiments validated the use of CCBO for the guided synthesis of poly(lactic-co-glycolic acid) particles with diameters of 300 nm and 3.0 μm via electrospraying under minimal initial data. Overall, the CCBO approach represents a versatile and holistic optimization paradigm for next-generation target-driven particle synthesis empowered by artificial intelligence (AI).
Please wait while we load your content...
