Digital twins in R&D: Bridging the GAP between simulation and experimentation

Author: Saskia Hoving

Editor-in-Chief

Over 50 years ago, Apollo 13’s safe return relied on simulators, a precursor to what we now call digital twins. Today, these virtual replicas are transforming R&D by bridging simulation and real-world experimentation. In this post, we explore how digital twins accelerate innovation, improve reproducibility and support decision-making. We’ll hear from Victor Richet, Nuclear Engineer at Assystem, on how this technology is shaping processes in one of the most complex and regulated industries.

The Apollo 13 mission faced a life-threatening crisis when its main engine was damaged. Engineers relied on simulators to model scenarios and guide the astronauts safely back to Earth, a pioneering glimpse into what we now call digital twins. Fast forward to today, and this concept has evolved into a powerful tool reshaping research and development across industries. But what makes digital twins powerful?

Digital twins are virtual representations of physical systems, processes or objects, continuously updated with real-time data. They act as a dynamic bridge between simulation and experimentation, enabling researchers and engineers to predict outcomes, optimise designs, and make informed decisions before committing to costly physical trials. By integrating AI and automation, digital twins create standardised environments that improve reproducibility, accelerate innovation and reduce time-to-market.

Their impact is particularly significant in sectors where precision, safety and compliance are non-negotiable, such as nuclear engineering. Here, the stakes are high, and experimentation is expensive and heavily regulated. Digital twins can help minimize risk, streamline workflows and support compliance with stringent safety standards while improving efficiency.

To understand how this technology is being applied in one of the most challenging engineering environments, we spoke with Victor Richet, Nuclear Engineer and Head of Digital at Assystem. In the conversation that follows, Victor shares how digital twins are transforming R&D in the nuclear industry, the benefits and limitations of this approach, and what the future holds for this technology.

How does Assystem define a digital twin in the context of its work, and what prompted you to adopt this technology?

Digital twins are a model which is fed with data in order to provide added value to processes. Our processes in the nuclear industry aren’t as up-to-date as those in, say, the automotive sector, which has fewer regulatory barriers and faster iteration. Our industry is heavily regulated for reasons of safety, and this makes us less agile in the adoption of new methods. Assystem were one of the industry’s first movers in the digital twin space as we recognised the huge scope this technology has to offer. Our mission is to accelerate the energy transition throughout the world and digital twin technologies play a major part in this, they advance digital transformation by addressing challenges throughout the full life-cycle of complex systems.

Digital twins are often described as the bridge between simulation and experimentation. How do you see that playing out in practice at Assystem?

At Assystem, digital twins aren’t just virtual simulations. They are digital ecosystems that integrate simulation tools, real-world data, and life-cycle management into one collaborative environment. In the specific case of nuclear engineering, the cost of experimenting is huge compared with other industries. Digital twin technologies aren’t a replacement for experiments, but they do reduce trial and error, they’re a way of increasing the likelihood of getting things right the first time.

How do you validate your digital twin models to ensure confidence in their predictions?

We run extensive test case processes to ensure the data provided by digital twin models is relevant and accurate. A good illustration of this is a project we carried out in the South of France to redesign a decommissioned nuclear facility. It involved moving potentially contaminated flows and irradiated materials out of the facility to be treated elsewhere and establishing temporary barriers, all in a seamless process. There are specific regulatory constraints which the safety authority issues for rooms in a nuclear facility, depending on the amount of radioactivity, red, orange or green zones. To complete this project, a 2D digital twin of the facility was built so that we could quickly, efficiently and reliably know where a barrier should be established, and what the radioactive zones of the rooms we then created would be. We validated the digital twin early in its development by comparing its results with the engineers’ calculations.

How do AI and machine learning enhance the predictive power and adaptability of your digital twin models?

The use of AI in the nuclear industry is still limited due to its probabilistic nature, precision and explicability are strict requirements for the safety authorities. Nevertheless, putting safety-related applications aside, there are several ways to use AI within the framework of digital twins. It’s mainly used to generate synthetic datasets and produce expected outputs, these don’t of course constitute proofs, but they are key insights for the early stages of the twin’s development. In a nutshell, we can generate synthetic data for a test case, use AI to simulate a wide range of results, and identify any key flaws in the twin.

When we’re developing and refining our digital twin models, access to trusted research and scientific data is absolutely crucial. The nuclear industry is not the most advanced, so there’s a strong consensus that it would benefit heavily from research conducted in other industries.

What measurable impact have digital twins had on your R&D process?

Digital twins have had a significant impact on our R&D process. They’ve contributed to reduced time to market and improved reproducibility by making processes more consistent and traceable. They’re also enabling us to achieve traceability and improve the quality of our experiments, which are long and extremely expensive.

Can you share an example of how collaboration with academia or industry partners has helped refine your digital twin models?

At Assystem we collaborate a lot with academia. Our partnership with the prestigious French School of Engineering IMT Mines Alès is a good example of this. In 2019 we set up a five-year R&D programme together dedicated to Model-Based Systems Engineering (MBSE), to refine and establish common methodologies (a ‘grammar’) for nuclear modelling systems. This model-based approach facilitates the handling of complex and stringent requirements in critical infrastructure projects, ultimately forming the basis for the use of digital twins industry-wide.

What are the biggest challenges in integrating digital twins in the R&D process, and what have you learned from this?

Surprisingly, the biggest challenges we’ve seen are not related to technology. Regulation, complexity, high cost and long timelines means there aren’t as many projects per year in nuclear as there are in other industries. Change management and the ability to integrate existing engineering processes is therefore the biggest challenge.

Secondly, nuclear is still a heavily siloed industry by discipline. You need seamless access to quality data, but as soon as you bridge silos is where difficulties arise.

There are three things we learned about the successful adoption of digital twins in our processes. Sponsorship and endorsement are critical; you need strong support to navigate the high-level safety requirements and regulatory barriers we face. It’s also crucial to have a clear vision of what you want to achieve, and what this would unlock in quantifiable terms.

Lastly, there’s no magic wand. With digital twins there’s no ‘plug-and-play’. Every project has different requirements, so you need to align your vision, resources (such as data and subject matter experts) and sponsorship, not to mention regulatory pathways and change management plans.

Where do you see digital twin technology heading in the next 3–5 years for R&D?

Digital twins are likely to become prevalent, regardless of industry. In nuclear the experiments are long, inefficient and expensive, so there will be a gradual increase in their relevance in the design process. There won’t be a digital twin revolution, this will happen gradually, all journeys start with a single step.

For those looking to adopt digital twin technologies, my advice would be to start small and ensure you have your sponsorship in place, being overly ambitious can increase the risk of failure and lead to significant cost. There are also a wide range of actors involved in the digital twin journey. The most successful projects I’ve seen have resulted from the efficient collaboration of stakeholders all working in their own area of expertise, whether that’s engineers, facility operators, solution providers or tools integrators. Digital twins are like a puzzle, everything and everyone, has to be in the right place.

Digital twins bridging simulation and experimentation

As Victor emphasized, adopting digital twin technology is a progressive journey that thrives on clear objectives, strong sponsorship and collaboration across disciplines. In highly regulated industries such as nuclear engineering, these virtual models can deliver substantial value by helping enhance reproducibility, accelerate design cycles and support informed decision-making.

The benefits for R&D are significant: streamlined timelines, improved traceability, and greater confidence in experimental outcomes. Assystem’s experience shows that success comes from starting with focused goals, aligning resources effectively and building partnerships with academia and industry experts to realise the full potential of this technology.

For those ready to explore digital twins further and strengthen research workflows, access to trusted methods and protocols is essential. Springer Nature Experiments offers a comprehensive portfolio of reliable resources to support innovation and reproducibility. Text and data mining (TDM) can also help by revealing patterns and connections across large research datasets, enabling deeper insights, identifying relationships and supporting semantic analysis. You can also discover more about the rise of digital twins and their growing role in sectors from healthcare to advanced engineering.

Related content

• Reproducibility as a competitive advantage in life science R&D
• Improving research reproducibility with open method sharing
• How research access drives public health innovation
• Trends and practicalities of digital transformation in life sciences R&D 
• Top 5 benefits of digital labs in life sciences R&D

Don't miss the latest news & blogs, subscribe to The Link Alerts