MATISSE or the rise of a new industrial revolution

Optimizing all production systems

Is this the start of a new industrial revolution? The vast European research project known as MATISSE (1), in which a team from IMT Atlantique is participating, aims to optimize all production systems, making them more efficient, more reliable and more effective. MATISSE is thus in line with the “industry of the future” (or “4.0”), which seeks to improve production systems by relying in particular on AI and digital technologies - but taking the approach even further. Ultimately, this should enable industry on the Old Continent to make a major technological leap forward. And gain a competitive edge.

MATISSE, which benefits from the European Horizon KDT- JU label (see box), was launched in autumn 2024, for a 3-year period. The project, coordinated by the University of Mälardalen (Sweden), brings together some thirty partners (major manufacturers, SMEs, research centers, academic institutions) from seven countries: Austria, Finland, France, Italy, Portugal, Sweden and Turkey. It has an overall budget of 18.2 million euros, part of which is financed by Europe, with the remainder coming from national sources. Seven national clusters have been set up, each focusing on a specific sector of activity and a specific industrial issue: bridge construction/maintenance, energy networks, rail transport, even banking services... The IMT Atlantique team is collaborating with the University of Rennes and Softeam, a subsidiary of Docaposte (La Poste's IT branch), on modeling.

Hugo Bruneliere, coordinator of the French consortium for the project

Widespread use of digital twins

The cornerstone of the project is the widespread use of digital twins, or “virtual doubles”, to reproduce and anticipate the operation of production systems. They can be used to simulate operations, modify an assembly, predict the behavior of an assembly and thus plan its maintenance... Thanks to them, many costly or risky operations can be avoided," explains Hugo Bruneliere, coordinator of the project's French consortium and a researcher in the Department of Automation, Production Engineering and Computer Science (DAPI) at IMT Atlantique.
However, these digital twins are generally designed for a single industrial system - an assembly line, for example," continues the researcher. So every time we build a system, we also have to design another ‘customized’ twin. And that's expensive...". There are a few “generic” tools on the market - but their integration is often long and complex. And their scope is generally limited to one or two areas of activity.

Hence the idea behind the MATISSE project: to investigate ways of “federating” digital twins. “We're looking to define common models, to make them easier to integrate and more flexible to use,” says Hugo Bruneliere. In the longer term, we're aiming for model interoperability, with the ultimate goal of achieving standards." In other words, the aim is to lay the foundations for a “second generation” of these digital twins. To achieve this goal, the project will make extensive use of modeling, big data and cloud technologies. For its part, the IMT Atlantique team is contributing its expertise in model-based engineering. “Our approach is to go up in abstraction, capitalize on the knowledge thus created and reuse it,” notes Hugo Bruneliere.

A host of obstacles to overcome

MATISSE's approach is highly ambitious, but it comes up against some formidable difficulties.
Firstly, the fields of application for digital twins are extremely diverse. Between energy networks and bridge maintenance, the similarities are not immediately obvious... Not only that, but some digital twin models are dedicated to a specific trade or sector of activity, while others are dedicated to processes or system design... How can such heterogeneous models be “federated”? 
Not to mention the fact that not all business sectors are at the same stage of industrial development. The automotive industry, for example, is already largely robotized and uses a large number of sensors. This is not the case in other sectors, where the use of twins and digital technologists is still in its infancy.
Another challenge is that of the different sensor technologies used, which need to be made compatible.
Added to this are legal constraints, as well as confidentiality and intellectual property issues, which will inevitably arise between sometimes competing manufacturers...
 

AI to play a key role

To overcome these multiple difficulties, AI should be of great help. The MATISSE teams plan to use large language models (LLMs) to facilitate the engineering of the various components of digital twins. This will avoid having to do this work completely “by hand”. A PhD student at the school has already worked on using LLMs to automatically find possible links between two models. Another PhD student is currently working on the design of “agents” - in other words, autonomous software components capable of interacting with each other and carrying out different tasks. Once again, these agents are based on LLMs.
The MATISSE project is part of a context marked by the growing use of digital twins worldwide. Chinese manufacturers, in particular, are taking a keen interest in the issue. The United States, for its part, uses the MBSE (Model-Based System Engineering) approach, which is quite widespread across the Atlantic. With MATISSE, we're trying to promote a European model," stresses the researcher. There is already a scientific community working on these issues in Germany and France, for example.
 

In all, MATISSE mobilizes some sixty researchers and experts. To collaborate effectively, each of the partners works with several colleagues - a way of leveraging their own work and feedback. "We've also been working on defining use cases and developing scenarios, before turning our attention to the solutions themselves. We're still only at the very beginning of the project." Intermediate stages could also be envisaged, given the scale of the challenge.
For Hugo Bruneliere, the MATISSE project could lead to a real paradigm shift for industry and business - some are already talking about Industry 5.0... " AI will undoubtedly enable us to go even further than expected," he says. For example, it could help us optimize engineering models. It's a very exciting time for us researchers.

(1) Model-Based Engineering of Digital Twins for Early Verification and Validation of Industrial Systems
 

Horizon Europe program and KDT-JU

Horizon Europe is the EU's major framework program for research and innovation for the period 2021-2027. It has a budget of around €95 billion. Its main pillar, “Global Issues and Industrial Competitiveness”, aims to support work related to societal issues: health, “inclusive, creative and safe” society, digital, industry, food, bio-economy, natural resources. 
The KDT-JU (Key Digital Technologies-Joint Undertaking) program, dedicated to digital technologies, is attached to Horizon Europe. Its aim is to design advanced applications and components, and to work on their production and integration. It is organized into several projects - including MATISSE. It was recently renamed “Chips JU”.