Chemical industry leader shares experience
- Leader in lubricant technology applies Alchemite™ machine learning to accelerate development.
- Applications have included fluids for pipe coatings, hydraulics, metal working, and shock absorbers.
- In an example project to adapt a product to a new market segment, a 50% performance improvement was achieved.
- Adaptive experimental design enabled FUCHS to hit their target faster.
Summary
FUCHS, the world’s largest independent lubricant manufacturer, has shared its experience in applying Alchemite™ machine learning to accelerate lubricant development. Dr Richard Bellizzi presented recent progress at an Intellegens webinar, discussing how Alchemite™ has been applied in a project to adapt an existing product to a new market segment, achieving project goals in significantly reduced timescales.
With around 10,000 different lubricant products serving 25 industries, FUCHS faces a major innovation challenge in maintaining a leading edge and continually adapting its products to new applications. Conventional approaches work well when projects are relatively straightforward, with few iterations, a known design space, and familiar materials. But, often, the objective is complex, requiring many iterations, a large design space, and the use of new materials. The desire to speed up these more complex projects motivated FUCHS to investigate machine learning solutions.
The team selected Alchemite™ and has embraced an Adaptive Design of Experiments approach, in which expert knowledge is combined with data-driven guidance from AI to focus experimental work on the formulations that are most likely to succeed. Dr Bellizzi explained that the method has gelled with the existing FUCHS approach by guiding candidate identification and providing data-driven analytics when evaluating performance of candidates, while retaining the familiar cycle of chemist-led experimentation. Alchemite™ has been applied to lubricant formulations in areas including:
- Pipe coatings
- Hydraulic fluids
- Metal working
- Shock absorbers
One example project was to adapt an existing product formulation for use in a new market segment in the aluminium forging sector. The FUCHS team trained a machine learning model using data from 24 historic formulations. The model was used in three cycles of experimental design to generate 30 candidate formulations, with the objective of a 50% improvement in key performance metrics. A chemist reviewed these recommendations, aided by Alchemite™ analytics, and selected seven to be tested. Two of these which met the performance target were shared with the customer, one of which is now being pursued further.
“Having Alchemite™ walk us through this process, we were able to get there faster,” explains Richard Belizzi. “A tool like this really helps us tackle the plethora of different market segments, which require different material sets, and different combinations of those materials to meet requirements.”
A full recording of the webinar, including the FUCHS presentation and a demonstration of the Alchemite™ software for a lubricant application is available.
