Description du projet
Une solution révolutionnaire pour optimiser les systèmes de production
Afin de préserver la compétitivité de la production manufacturière, les gestionnaires doivent concevoir et gérer des systèmes de production complexes, collaboratifs et reconfigurables qui tirent le meilleur parti possible des nouvelles technologies. Des recherches supplémentaires sont toutefois nécessaires pour pouvoir profiter pleinement des outils numériques destinés à l’industrie manufacturière. Le projet ASSISTANT, financé par l’UE, vise à développer des solutions révolutionnaires destinées à l’industrie manufacturière, qui exploitent l’intelligence artificielle pour optimiser les systèmes de production. L’une des pierres angulaires du projet ASSISTANT est la création de jumeaux numériques intelligents. En combinant l’apprentissage automatique, l’optimisation, la simulation et les modèles de domaine, ASSISTANT développe des outils et des solutions qui fournissent toutes les informations nécessaires pour aider les directeurs de production à concevoir des lignes de production, à planifier la production et à améliorer les paramètres des machines afin que ces dernières prennent des décisions efficaces et durables qui garantissent la qualité et la sécurité des produits.
Objectif
With a multidisciplinary consortium combining key skills in AI, manufacturing, edge computing and robotics, ASSISTANT aims to create intelligent digital twins through the joint use of machine learning (ML), optimization, simulation and domain models. The resulting tools permit to design and operate complex collaborative and reconfigurable production systems based on data collected from various sources such as IoT devices. ASSISTANT targets a significant increase in flexibility and reactivity, products/processes quality, and in robustness of manufacturing systems, by integrating human and machine intelligence in a sustainable learning relationship.
ASSISTANT provides decision makers with generative design based software for all manufacturing decisions. Rather than writing ad hoc code for each manufacturing sector, it provides a set of intelligent digital twins that self adapt to the manufacturing environment. It promote a methodology that enhances generative design with learning aspects of AI thanks to the data available in manufacturing. ASSISTANT aims to synthesize predictive/prescriptive models adjusted to the shop floor for each decision levels. Digital twins will be used as oracles by ML in order to converge towards models in phase with reality. This means that rather than writing specific code to cover a restricted set of goals/scenarios/hypotheses for a manufacturing system and a decision level, ASSISTANT will aim at learning models that can be used by standard optimization libraries. In this context, ML is used to predict parameter values, characterize parameters uncertainty, and acquire physical constraints. ASSISTANT will experiment this methodology on a significant panel of use cases selected for their relevance in the current context of the digital transformation of production in major manufacturing sectors undergoing rapid transformations like the energy, the industrial equipment, and automotive sectors which already make extensive use of digital twins.
Champ scientifique
- natural sciencescomputer and information sciencesinternetinternet of things
- natural sciencescomputer and information sciencessoftware
- engineering and technologymechanical engineeringvehicle engineeringautomotive engineering
- natural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learning
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringrobotics
Mots‑clés
Programme(s)
Régime de financement
RIA - Research and Innovation actionCoordinateur
91120 Palaiseau
France