Everyone’s talking about digital twins these days. Typically defined as a dynamic virtual replica of a real object or system, digital twins are built from data collected via sensors and models of this system. They can simulate, analyze and predict system behavior in real time, and are used in a multitude of sectors, including healthcare, energy, construction, aeronautics and nuclear energy. Their popularity stems from their ability to help pilot systems, anticipate failures, and test and validate scenarios without risk. For industry, this makes it possible to optimize production processes, improve maintenance, increase the efficiency and reliability of operations, and train operators.
Nevertheless, there are still major challenges around their design, maintenance and use. Digital twins are not yet widely deployed, especially for industrial systems, due to their high set-up costs and technical complexity. They require the collection and analysis of massive volumes of data, which raises questions about their carbon footprint and cybersecurity requirements linked to their use. Specialized expertise is also needed to integrate them, which can limit their adoption for companies lacking the necessary resources or skills.
A shared vision for research and industry
Officially launched on September 20, 2023, the three-year chair project dates back to late 2019. It shows the desire of various IMT schools to work together on a shared scientific program on the theme of digital twins. Put on hold in 2020 due to the global Covid-19 epidemic, the initiative was eventually piloted by three schools in the IMT network: Mines Saint-Étienne, IMT Mines Albi and IMT Mines Alès. It focuses on industrial systems and how they are organized.
Siemens Digital Industries Software, a longstanding partner of Mines Saint-Étienne, was the first to join the consortium, followed by Laboratoires Pierre Fabre, a partner of IMT Mines Albi. The group was completed by engineering company Inoprod, giving “an overview of the full value chain – from software publisher to user – and today’s industrial needs“, says Vincent Chapurlat, systems engineering researcher at IMT Mines Alès.
“We came together around common issues and scientific obstacles, which represent crucial challenges for both industry and laboratories. We then had to formalize the relationship between the various partners,” says Frédéric Grimaud, industrial engineering researcher at Mines Saint-Étienne. “Today, our partners are committed to the project: they are closely involved in all aspects of research, and play an active role in providing use cases.” On the topic of use cases, three have been provided, which will serve to validate and evaluate the chair’s contributions through three PhDs piloted respectively by each of the schools involved.
“Simple”, “simple” and “simple”
The aim of the chair is to be able to design (build and run) digital twins for specific applications more quickly that are responsible – using less energy and IT resources – and sustainable, i.e. maintainable over time. “What interests us is how the digital twin is designed, not the industrial system, and how using the twin can make it possible to use the system to best effect“, says Jacques Lamothe, industrial engineering researcher at IMT Mines Albi.
Gone, then, are the “mega-twins” requiring several days of computing and which did not align with the needs of the actors in charge of industrial systems! “Digital twins can be connected to reality to varying extents. With the chair, we want to achieve a high level of connectivity between field data and industrial systems, for more aligned decision-making,” says the researcher. On top of these criteria, it is also important to be interoperable with other digital twins, so that you don’t have to start from scratch when a model – even a partial one – already exists. All these requirements are combined in a single rule that the research teams summarize as the three Ss: simple design, simple operation and simple maintenance.
Digital twins for efficient and effective industrial systems
To innovate around a design process that satisfies all these conditions, each team is exploring different but complementary avenues, in line with its expertise. The IMT Mines Alès team is working on a specific engineering framework for designing digital twins. It combines the principles and processes of model-based systems engineering (MBSE) and pattern-based engineering (PBE).
MBSE is an approach used to design and manage complex systems based on digital models. A design pattern is a standardized solution, based on proven knowledge and therefore easier to reuse. The design pattern approach is a formalized and comprehensible way of moving faster. “When you want to model a physical system, it’s a safe bet that others have tried before, whether in whole or in part,” says Vincent Chapurlat. “Our idea is therefore to reuse or draw inspiration from what already exists and has been tried and tested, in order to save time and improve quality. There’s no need to reinvent the wheel!’
Ultimately, the proposed framework must take into account the need for speed and reusability when developing a digital twin, interoperability, but also usage (if usage changes during its lifespan), to quickly configure or reconfigure the model, as well as maintain it.
Integration of new critical resources
For their part, scientists at Mines Saint-Étienne are studying new “critical resources” – strategic resources qualified according to their economic importance and availability. When poorly synchronized, critical resources can negatively impact production flow, which is why they are subject to management rules developed with the help of “what if” simulation scenarios.
Historically, critical resources have included machines, operators and software systems. However, with the environmental and managerial transitions of recent years, new critical resources have emerged, notably energy (water, gas, electricity, etc.), but also human resources in terms of parameters such as fatigability and ergonomics.
“These are new modeling factors which, for the moment, are not taken into account in the software chain used to create digital twins, or not sufficiently” says Frédéric Grimaud. The teams at Mines Saint-Étienne are therefore seeking to integrate these dimensions into their scenarios, by defining new indicators for steering industrial performance.
AI to boost simulation
The end result is that the steering model favors short-term, almost real-time decision-making. This means it needs to be connected to the company’s IT system, so it can rapidly design and adjust the industrial system it simulates. However, basic simulation models are insufficient in terms of speed and response times when compared to real-time applications. It is therefore the mission of the IMT Mines Albi teams to develop models capable of producing the same simulation results, but more rapidly.
To do so, they rely on AI. “Typically, many industrial bodies want to integrate their highly complex and heavy digital twins into a more instantaneous decision-making loop. So we connect these models to AI tools to learn and propose faster meta-models,” says Jacques Lamothe.
Following the first year of discovery and implementation – including the launch of three PhDs – and the initial results of this work, the chair will continue into its second year, focusing on strengthening the maturity of its research. In the final year, efforts will focus on finalizing the use cases and reflecting on the chair’s future – which is already taking shape, with a shared desire to continue exploring and innovating around digital twins for industrial systems, respond to future challenges and even collaborate with other partners.
