AI-TWILIGHT: AI powered Digital Twin for lighting infrastructure in the context of front-end Industry 4.0
Over the last decades, LEDs have taken over the lighting industry. Their design is a multiphysical problem, involving mechanical, optical, electrical, and thermal components and considerations. All of these domains influence the operational behavior of LEDs, with the temperature having the largest impact on the lifetime of a given product. Due to the complexity arising from the interplay of these factors, LED luminaires require accurate multi-domain models to be designed efficiently. The previous Delphi4LED project (coordinated by Genevieve Martin of Signify) established combined design and simulation workflows, which were proven useful by industrial applications.
The new AI-TWILIGHT project (coordinated by Genevieve Martin of Signify) aims to develop Self-Learning Digital Twins of LED-based luminaires. Two major problems will be tackled. The first goal is to reduce the amount of data necessary to create high quality models, since an extremely large cost is associated with obtaining lifetime test data of LEDs. The second aim is to improve the prediction power of the simulations. This includes novel aging and LED driver models, that will be used for design purposes and to help create a Digital Twin during operation. The luminaires will then be able to learn their lifetime history, monitor their health, and enable predictive maintenance.
For more information see the overview article (pp. 14-15) about AI-TWILIGHT written by Wil Schilders, who is one of the project leaders from Eindhoven University of Technology. The team at TU Darmstadt consists of Sebastian Schöps, Idoia Cortes Garcia and Peter Förster.
The project AI-TWILIGHT runs from 1 June 2021 to 31 May 2024 and is funded under H2020-EU.220.127.116.11. with grant agreement ID 101007319. More details are on the CORDIS website of the European union: