AkzoNobel is a well-known Dutch multinational company in chemical industry. They provide high quality paints and performance coatings for thousands of paint distributors across the globe and large retail outlets such as B&Q, Leroy Merlin and OBI.
For creating new paintings and updating the components in their existing productions, they use chemical and physical models to predict important characteristics of candidate formulas before performing expensive tests on them which requires lengthy manual iterations on ingredient concentrations to meet the specifications desired.
In 2018, AkzoNobel got the help of Machine2Learn to upgrade their model into a deep learning model, so it becomes more flexible regarding the level of details in input parameters and more future-proof regarding the introduction of new ingredients in the market.
Machine2Learn task included:
- Designing, developing and deploying deep learning and optimisation models;
- Developing a customised software for integration on-site;
- Adding self-learning capabilities to the deep learning model;
- Transferring of knowledge to the in-house team to perform the maintenance and extensions locally.
Using the deployed application, AkzoNobel reduced the need for manual operation work in their ingredient and concentration selection by 15%.
In order to make the application available to the AkzoNobel engineers in different offices, Machine2Learn deployed the application on Microsoft Azure cloud and integrated it with their cloud infrastructure. You can read more about the integration part on the Microsoft page dedicated to this use case.