Key outcomes
- Ease-of-use of Alchemite™ machine learning makes it ideal for chemists
- Clean air project identified improved formulations and designed new experiments, doubling efficiency
- Life science technologies project delivered a significant yield improvements and a 5x experimental speedup
Summary
At an Intellegens webinar, Claire Hatfield, Digital Analyst at Johnson Matthey, presented case studies in the application of Alchemite™ machine learning to chemicals R&D. Johnson Matthey are specialists in sustainable technologies based on catalysts and related products and services that deliver cleaner air and healthier environments. The company recognised, in the words of R&D Director Liz Rowsell, that “JM is a company rich in data but not always rich in the insights we could gain from our valuable data resources”. Part of the solution was that “the application of quality digital systems could unlock this data mine.” Alchemite™ is one such system.
In one project, focused on clean air applications, around 600 complete datasets which included over 70 different variables were studied, leading to improved formulation designs for catalytic converters. The machine learning model developed was shown to be highly predictive of experimental results in a low temperature regime when provided with high temperature results. This can now support design of experiments to focus future testing programmes on high temperatures, halving the amount of physical testing. Work from this project has been published – see details.
A second project, with Johnson Matthey’s Life Science Technology unit, was able to identify changes that resulted in a 4% increase in yield for a key reaction for which no further improvement was thought possible. When scaled-up, this corresponds to significant cost and energy savings. In another experiment, machine learning insights enabled the JM team to reach target performance five times faster.
Claire Hatfield also explained how Alchemite™ has facilitated deployment of machine learning through its easy-to-use web user interface, which enables chemists to engage with the technology with no need to code or develop a deep data science background – “the biggest success that we’ve seen is in how easy the platform is to use. We have scientists now using it who previously didn’t have that much interest in analysing data.”
