Stronger, lighter, and faster: multi-objective Bayesian optimization for fused filament fabrication

The aim of this work is to reduce the experimental trials required to optimize the parameters of fused filament fabrication (FFF) to meet specific product requirements. Balancing conflicting objectives within budget is a challenge for functional prototyping and direct manufacturing. Herein, we present and validate a multi-objective optimization approach that identifies process parameters to minimize manufacturing costs without compromising mechanical performance. The hypothesis is that Bayesian optimization (BO), which leverages data from previous experiments to inform future experiments, can significantly reduce the number of required experimental trials and improve part performance compared to design of experiments. The approach involves building a surrogate model based on empirical data and maximizing the expected improvement against the current Pareto front for balancing three conflicting objectives: (i) fracture energy, (ii) component weight, and (iii) build time. The efficacy of the m-BO is demonstrated by converging on Pareto-optimal parameter within 12 experimental iterations. Our results show that selecting parameters from the Pareto front identified by the m-BO can reduce build times by up to 74% and simultaneously increase the strength-to-weight ratio of the FFF-fabricated parts up to 127%. This approach reduces costs and lead times and has potential for use in other materials and processes.