Skip to main content
2023 AOCS Annual Meeting & Expo Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back
Video

Surfactants and Detergents

Session:: Surfactants in Agricultural Applications

Predicting emulsion stability using machine learning in agrochemical formulations

Monday, May 1, 2023
12:40 PM – 1:00 PM MDT
Room: 406
Abstract: Adjuvants for agrichemical formulations are a multi-billion-dollar industry with surfactants playing a key role in the formulation and application of these formulations in agriculture. An important formulation type is emulsifiable concentrate (EC), which deliver the biologically active ingredient through dissolved in an oil concentrate along with surfactants and other adjuvants. ECs provide ease of transport & storage but must readily form emulsions upon dilution at the grower’s field under uncontrolled and highly variable conditions (water hardness, pH, temperature, and mixing). These emulsions must exhibit minimal separation over the extended use period of the emulsions in the field. Increasingly complex active ingredients along with shifts in the oils and other adjuvants towards sustainable materials creates a perennial challenge for the formulation of ECs.

Here we describe a joint computation and high throughput experimental study undertaken within Dow to predict the stability of a given emulsion useful for agrichemical formulations. Utilizing thousands of high throughput formulations taken from across multiple surfactants, oils, and water hardness, we created an experimental data set of emulsions with and without separation over 24 hours. This experimental data set relied upon an industry standard multichannel liquid handler, a custom imaging station, and a custom image analysis algorithm to accurately identify emulsions with minimal separation. Combining the experimental data set with physical and chemical descriptors of the formulation components enabled training of an ensemble of cluster-based machine learning models (e.g., Decision Trees, k-means, etc.) to predict whether a given formulation would be stable, some of which demonstrated cross-validation accuracies as high as 75%. Finally, we report our progress on the development of structure/property relationships between the descriptors and emulsion stability and discuss opportunities for future research.