this article is a repost from l'usine nouvelle : https://www.usinenouvelle.com/article/ia-generative-bots-apprentissage-automatique-comment-renault-augmente-son-ingenierie-de-conception-avec-la-start-up-dessia.N2071097
The generative AI and “bots” developed by the start-up Dessia significantly accelerate the exploratory design phases and facilitate the work of engineers. The example of the electrical wiring of a hybrid powertrain, presented by Renault at the last NAFEMS conference, speaks volumes in this respect, with a design time cut by thirty. Later, supervised and reinforcement learning techniques should make it possible to formalize the implicit and empirical practices of engineers.
A day instead of a month: here is the significant time saving within reach of the engineers of the Renault group , when they strive to determine the best concept of electrical architecture feeding a hybrid powertrain. The key is design software using generative algorithms and developed by the start-up Dessia. This convincing result was presented at the regional conference NAFEMS (organization focused on the technological aspects of digital simulation), the last edition of which was held in Senlis on November 23 and 24, 2022.
“ Our software automatically traces the wiring paths during the preliminary design phase of a motor ,” explains Pierre-Emmanuel Dumouchel, CEO and co-founder of Dessia. It is in a way the GPS of the wiring. Renault reduces design time and can test more things. »
This start-up, born in 2017, endorses the principle of generative design – an iterative technique that serves to explore multiple design possibilities based on parameters set by the engineer – while modernizing it: the engineer creates, using a development kit where he describes the product, the rules, etc., a robot-companion (or "bot") who will assist him in his task.