Solving the complexity of bio-based building blocks for surfactant production through Machine Learning
For the surfactant industry, the shift towards bio-based economy implies sourcing more building blocks from natural resources. Even though a wide variety of bio-based raw materials is already available, their utilization is very limited in surfactant production as their chemical characterization is too tedious, owing to the complexity of the mixtures they are made of. Such an example is the molecular family of tannins which, despite having attractive features, presents a too important molecular heterogeneity for being employed as such.Our proposal aims at solving this issue by utilizing Machine Learning in the characterization process of the raw materials as well as in the synthesis of the surfactant. We will produce a database of tannin-based surfactants that will be low-level analysed by spectroscopic techniques (e.g. NMR, IR, UV spectroscopies). The spectra along with the physicochemical properties (e.g. critical micelle concentration) and the modification path will feed a convolutional neural network. In the end, the trained network, by integrating spectroscopic data, physicochemical properties, chemical modifications, will predict, from the spectroscopic fingerprint of the raw materials the chemical path to achieve the desired properties.This technology could then be integrated in production to continuously adapt the chemistry based on online spectroscopy (e.g. IR) to achieve product specifications but could also be generalized to other bio-based building blocks.
Romain Bordes (contact)
Researcher at Chalmers, Chemistry and Chemical Engineering, Applied Chemistry
Doktor at Chalmers, Chemistry and Chemical Engineering, Applied Chemistry, Applied Surface Chemistry
Professor at Chalmers, Chemistry and Chemical Engineering, Applied Chemistry, Lars Evenäs Group
Swedish Research Council (VR)
Funding Chalmers participation during 2020–2023