第386回化学システム工学専攻公開セミナー Data-driven models for advanced process engineering

講演題目	Data-driven models for advanced process engineering
講演者	Dr. Bernardo Castro-Dominguez Assistant Professor University of Bath, UK
概要	The use of data-driven models for chemical applications has accelerated due to its success stories in materials discovery, healthcare, and smart manufacturing. In this seminar, we will describe how chemoinformatics translates the properties of molecules into a digital format for data analysis. In particular, how two-dimensional images of chemical structures can be used as molecular representations. These methods were employed in both classification and regression tasks, and used to predict the solubility of pharmaceuticals, co-crystallization and amorphisation events in mechanochemical settings, and the prediction of crystal morphologies. Beyond images, we will briefly discuss the use of chemometrics used for in-line characterization and process optimization. As case study, Density Functional Theory was used to generate theoretical infrared spectra of various polymers, which then were correlated to a property (e.g. melting point). These spectra-to-property datasets were used to create data-driven models and predict the property of untested ones. All this work follows the vision of developing "agnostic-to-feedstock" processes.
世話人	杉山弘和（内線27227）

東京大学化学システム工学科／専攻