Title: AI in action – rapid reformulation of ink products

Authors: Tom Whitehead and Gareth Conduit (Intellegens); Phillip Woolston, Dominic Gunstone, and Josie Harries (Domino Printing Sciences)

The traditional ink development process requires the formulation and testing of a large number of designs to find an optimal formulation. This empirical analysis is resource-intensive, and so using artificial intelligence to accelerate the process by identifying the most efficient development route delivers significant real-world advantages. Domino Printing Sciences and Intellegens have used artificial intelligence to inform the replacement of ink ingredients that were no longer suitable or available. The investigation leveraged historical experimental data from multiple projects, each of which had measured a variety of different performance properties, requiring an artificial intelligence approach capable of capitalising on this sparse data. With limited lab access during the COVID-19 pandemic hampering activities that rely on empirical analysis, maximising the value from each experiment using artificial intelligence is building towards a data-driven future of accelerated chemical formulation development.

This presentation will demonstrate the value that artificial intelligence can bring to active industrial chemistry projects. A multi-target artificial intelligence model [1], [2] generated accurate predictions of 28 properties of commercial significance (including viscosity, adhesion, and surface tension), with 75% of the modelled properties achieving coefficient of determination (R2) above 0.5 and 36% having R2 above 0.7. The artificial intelligence model was then used to propose new ink formulations, which were experimentally validated showing high agreement with the predicted values. The model is being extended to include wider families of ink formulations, and the presentation will also cover how this artificial intelligence model is being deployed to bench chemist users to deliver the benefits of improved predictions and formulation design without requiring the reskilling of users in artificial intelligence approaches.

[1] B. D. Conduit, N. G. Jones, H. J. Stone, and G. J. Conduit, “Design of a nickel-base superalloy using a neural network,” Mater. Des., vol. 131, pp. 358–365, 2017, doi: 10.1016/j.matdes.2017.06.007.

[2] T. M. Whitehead, B. W. J. Irwin, P. Hunt, M. D. Segall, and G. J. Conduit, “Imputation of Assay Bioactivity Data Using Deep Learning,” J. Chem. Inf. Model., vol. 59, no. 3, pp. 1197–1204, 2019, doi: 10.1021/acs.jcim.8b00768.

• Alchemite™ for Formulation Design
• Ink formulation case study with Domino Printing Sciences
• An overview of the Alchemite™ Technology
• Case studies

Subscribe to our newsletter to get the latest news on our other upcoming events.