Optimization of Formulations Using Robotic Experiments Driven by Machine Learning DoE

Formulated products are complex mixtures of ingredients whose time to market can be difficult speed due the lack general predictable physical models for desired properties. Here, we report coupling a machine learning classification algorithm with Thompson sampling efficient multiobjective optimization (TSEMO) simultaneous continuous and discrete outputs. The methodology is successfully applied design formulated liquid product commercial interest which no available. Experiments carried out in semiautomated fashion using robotic platforms triggered by algorithms. procedure allows one find nine suitable recipes meeting customer-defined criteria within 15 working days, outperforming human intuition target performance formulations.