DIGITAL INNOVATION | ARTIFICIAL INTELLIGENCE
Above: AI has shown positive results” in
development of PU prepolymers at the Lanxess R&D facility at Naugatuck in the US
According to Dr Markus Eckart, Senior Vice
President and the Head of Urethanes Systems BU at Lanxess, the prepolymer project was initially thought to be ideal for the application of AI. The company had total control of the ingredients in the system and, what they considered to be, a large amount of relevant data. Citrine determined otherwise. “They said we did not have enough data. And
we learned we would never have enough data to be able to just use an algorithm. But we could combine our data with expert knowledge – domain knowledge — to make a new machine,” Eckart said. The first attempts revealed many gaps in the company’s data. Eckart said for any formulation it would be typical to have around 20 potential data sets but only five or so would be complete. The only way to speed up the decision making process was to fill those gaps. “Filling the gaps does not mean going back and measuring but using calculation tools to get working data,” he explained. The gap filling work entailed linking existing empirical measurement data with knowledge from the company’s process experts and then applying a “chemistry-aware” algorithm to calculate additional measurement values. Only a few real-life measure- ments were required to verify this new working data. “In just two months, positive results were being
seen, Eckart said. The next step is to see how reliably optimal formulations can be predicted to meet customer-specific product requirements. “If the next tests are successful, we will be able to fulfill customer requests even more quickly and effectively. Our existing knowledge of formulations will be enhanced by AI-assisted formulation design – in other words, systems that are not yet part of our portfolio but for which artificial intelligence will
46 COMPOUNDING WORLD | March 2020
enable us to know instantly whether we can manufacture them and how.”
Sizing up
Similar potential benefits have also been seen in glass fibre sizing development. The fibre sizing is critical in achieving optimal interaction between the glass fibre and the resin matrix, so determining the ultimate performance of a reinforced com- pound. “We expect AI to cut the development time for optimised formulations by more than half,” according to Dr Axel Tuchlenski, Head of Global Product and Application in the Lanxess High Performance Materials business unit. Tuchlenski describes the fibre sizing project as
“The Champions League of AI.” However, he says it is just a starting point in what it hopes to achieve using the technology — the long-term aim is to use the company’s 30 years of compound develop- ment and its collection of more than 100,000 individual recipes to meet the future performance and lead-time needs of customers. “For example, new mobility is one of our big
growth areas for the future. This – Evs – needs a change in material properties. Using AI we can make use of our large material database and make these products faster,” he said. “We will use Citrine to find the right recipe to deliver, for example, good fire retardance or electrical properties. Product development and compound development will always need experience — AI will not change that — but the number of experiments will be reduced.”
CLICK ON THE LINKS FOR MORE INFORMATION: �
www.lanxess.com �
www.citrine.io
www.compoundingworld.com
IMAGE: LANXESS
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68
