Project B1

The aim of B1 is the implementation of simulation approaches for training and augmentation of artificial intelligence methods to predict the polymer properties in a precise manner. The approach will lead to an understanding of the essential measurements yielding a design of experiment method (DoE) and a reduction of unnecessary experimental labor. For this purpose, systematic copolymer libraries will be designed to study the properties of copolymers, such as glass transition or melting temperature, switching temperatures of reversible polymers as well as phase transition in aqueous media. Since the experimental effort to design all possible combinations will be too high, only selected candidates can be synthesized, resulting in an underrepresentation of the parameter space. Thus, DoE approaches will be utilized and will be combined with active learning for predicting properties of copolymers. In addition, sample size estimation (SSP) methods for the construction of polymer libraries will be developed.