Graphics to visualize machine learning normalizing flows

New revolutionary technology to discover problems in industrial production

Vinnova recently granted research funds for a project with the aim to improve industrial quality monitoring through normalising flows, a new technology that is used within machine learning (artificial intelligence) to understand and model complex relationships and patterns. Together with experts from the industry, researchers from the research environment Data Intensive Sciences and Application (DISA) at Linnaeus University, will develop new methods to increase efficiency and accuracy in anomaly detection.

Traditionally, industrial quality monitoring has meant manual inspection, which is both costly and error sensitive. However, thanks to new technology based on machine learning (AI), a process to automatically detect anomalies has become possible.

“Our goal is to increase the flexibility and scalability of the quality monitoring system by reducing the need for manual handling. The new technology is called normalising flows and it helps find and predict anomalies in data, which makes it possible to identify potential problems at an early stage and take measures to prevent them”, explains Diana Unander, project manager.

The project gathers industry experts within the areas of manufacturing, automation, and quality-assurance and together with the researchers they will test and evaluate the performance of the technology in real-life environments. The companies Gimic, SKF, and Gunnebo Industries, which is part of The Crosby Group, contribute with their knowledge and production data. The aim is to make it possible to use the method in the industry and to improve the detection of anomalies.

"The consortium has initiated the work and there is great commitment and drive to continue working with the challenges we are facing. We see the combination of knowledge and skills as a success factor and look forward to being able to deliver results that will benefit not only the project parties but the industry at large”, Unander concludes.

The project is part of Vinnova’s investment in Advanced and innovative digitalisation and has, in tough competition, been granted roughly SEK 4 million from Vinnova. The total budget is roughly SEK 8 million and the project will run for 3 years.

Link to the project's website:
In-line visual inspection using unsupervised learning

Contact
Diana Unander, project manager, email: diana.unander@lnu.se, phone: +4673-057 70 64