Project Management – Work Package 1

Lead Beneficiary: ATOS IT SOLUTIONS AND SERVICES IBERIA SL (ATOS IT)

Objectives: The scope of this WP is the overall management of the activities to be performed in the project, including quality assurance and risk management, data management, and other ethical considerations.

Human-centric Design – Work Package 2

Lead Beneficiary: CARDIFF UNIVERSITY (CAR)

Objectives: The goal of this WP is to provide the requirements specification and architecture guidance to the development of the rest of the technical work packages, considering a human-centric approach.

Federated Learning and Continual Learning – Work Package 3

Lead Beneficiary: ATOS IT SOLUTIONS AND SERVICES IBERIA SL (ATOS IT)

Objectives: The implementation of the Federated Learning (FL) framework will be the main goal of WP3 in order to enable Machine Learning models to be trained in a distributed manner. This will avoid the need to use a single and central repository to collect data coming from multiple devices, machines, processes, systems and facilities. The framework will be composed of two main components: i) a set of agents with the capacity to train the models locally in each edge location or data source, and; ii) a central server that will be in charge of aggregating those local models. The FL framework will incorporate mechanisms that defend against adversarial attacks which  may try to alter the global learning model or to poison the data of the individual agents. Transfer Learning and Domain Adaptation techniques will be developed to ensure that the resulting models produced by FL will generalize properly for new industries and use-cases. Continual Learning will be used to detect domain shift and trigger a retraining process. Finally, WP3 will leverage the results obtained by the European Artificial Intelligence On Demand platform for resources and experimentation services.

Modeling and Optimization – Work Package 4

Lead Beneficiary: SCUOLA SUPERIORE DI STUDI UNIVERSITARI E DIPERFEZIONAMENTO S ANNA (SSSA)

Objectives: WP4 will focus on the creation of models using the data collected from the project use-cases and leveraging the framework provided by WP3. The models will optimize the energy consumption and the waste generation at the process level. Considering how the different process stages are interconnected and how they affect energy consumptions and waste generation. Different types of Machine-Learning based and hybrid approaches will be explored, to specifically design models that can fully exploit the capabilities of the developed Federated Learning framework..

Validation and Replication – Work Package 5

Lead Beneficiary: CELSA Group (CELSA France, CELSA Poland and CELSA Spain)

Objectives: Two pilots will be executed: steel-making in three melt shops at CELSA Group (Spain, France, Poland); and the production of automotive parts in three factories from Fonderia di Torbole. The pilots will consist of a set of iterative deployments starting from specific exercises and ending in a final complex demonstrator.

Exploitation, Dissemination and Environmental impact – Work Package 6

Lead Beneficiary: MANDAT INTERNATIONAL ALIAS FONDATION POURLA COOPERATION INTERNATIONALE (MI)

Objectives: The objective of this work package is to successfully create and carry out outreach and capacity building strategies for ALCHIMIA.