Digital twin and virtual twin, or physics-based simulation model, are both valuable in the world of engineering. They help us identify novel design solutions and take proactive action to mitigate risks. These are two separate, disconnected entities. But, what would happen if they were used together, reinforcing each other’s work? Let’s find out.

The Hybrid Twin: Marrying Simulation and the Digital Twin

Let’s start with a little background. At the ESI Live 2020 event, ESI Group Scientific Director Francisco Chinesta introduced a new concept: the hybrid twin. The hybrid twin is a simulation model working at the intersection of virtual and digital twins. 

It’s a clever idea. Physics-based simulation (which Chinesta also called the “virtual twin”) is often used to study the behavior of products and systems. But simulation has its limits as an emulator of reality. It represents an idealized version of an item. This is where the digital twin comes into play. 

Sensor data from our smart, connected products is gathered and analyzed for anomalies, offering up-to-date insight into behaviors. But can simulation learn from this real-world data to move closer towards reality? Maybe—and the answer lies in the hybrid twin.

This post investigates the hybrid twin. To start, we examine physics-based simulation and its constraints when mimicking reality. We also explore digital twin and the ever-increasing data volumes they store. If we could link the digital twin’s information bank to the world of simulation, as Chinesta suggests, we can overcome some of the limitations of today’s virtual twin. And that’s the basic concept behind the hybrid twin.

Physics-Based Simulation (the Virtual Twin)

A virtual twin is an ideal model of the product based on simulation. Here, physics-based simulation uses analytical and numerical methods to model the behavior of products and systems. Finite Element Analysis (FEA) for structural simulation is one example. 

But there are limits to physics-based simulation. It requires computer resources proportional to the size and complexity of the problem. Large systems (or a system of systems) cannot be practically simulated. This is because the simulations take too long and the required IT infrastructure costs too much. Furthermore, the simulation setup may not exactly match the operating conditions of the real world, introducing a source of divergence between the simulation and the real, operating product. 

Recent advances in numerical techniques, cloud computing, and GPU processing have helped address this issue to some extent. But limitations still exist. Solving large systems (or a system of systems) is still impractical for many physics-based simulations. 

The Digital Twin

A digital twin is a digital replica of the real-world system or product. There are many examples of digital twin across multiple industries. For example, you could have a digital twin of an operational aircraft engine sitting in the Pratt and Whitney data center. Here, the engine is the “product.” Sensors in the real engine send data to the digital twin. This data is stored and processed, creating a digital replica of the physical system. 

The sheer volume of transmitted data can be overwhelming. The aircraft engine in our example sends data every second from multiple sensors, but the practical use of this deluge is not that obvious. Is there a way to put this data to good use? Can limited datasets help us improve physics-based simulation? Chinesta and his team think so.

The Advantages of the Hybrid Twin 

There are two types of twins: virtual and digital. So, how can we use them together and to what advantage? 

A virtual twin has limitations, deviating from reality with margins of error that can be significant. When you create a hybrid twin by using the data from a digital twin as an input to a virtual twin, you can drastically reduce or even eliminate these errors. According to Chinesta, even a small sample of the digital twin’s data can reduce errors significantly. 

A hybrid twin increases the accuracy of simulations by reducing the errors to near-zero. As a result, it enables you to study large systems or a system of systems. This would be completely impractical using physics-based simulation models alone.

A hybrid twin model is a powerful design and product development asset. You can further improve and expand it by integrating artificial intelligence and machine learning algorithms. Such an approach will facilitate the development of improved, reliable, and resilient products of the future.

Who Can Benefit?

The technical advantages are clear. But who actually benefits from hybrid twins?

First, companies building complex and connected products stand to benefit. Many of today’s products are now embedded with sensors, which collect and transmit all kinds of data in real time. Unlike a digital twin alone, a hybrid twin can make real use of this data to improve a company’s simulation models and move them closer to reality.

Second, the hybrid twin is a boon to companies adopting simulation-driven product development. They can now use their products in the field to improve their simulation models. They can use more accurate simulations to make more informed decisions, boosting innovation. They can also verify product and system behaviors more precisely, reducing the reliance on prototyping and testing.

Lastly, companies building large systems or a system of systems can now use simulation to study those systems. This helps them better understand their systems and stress-test them, virtually.

Summary and Takeaways

  • The hybrid twin concept was proposed during the ESI Live 2020 event to address the limits of the virtual twin using digital twin data.
  • Physics-based simulation model, or virtual twin, often deviates from reality, especially when dealing with complex and large systems. The computational resources and time required to model such systems is impractical from both a time and a cost perspective.
  • Using sensors, digital twin gather and store vast amounts of real-world data— information that could help reduce simulation errors in a virtual twin. A hybrid twin harnesses that digital twin data to reduce errors in a virtual twin to near zero.
  • Companies building complex and interconnected products, those adopting simulation-driven product development, and those building large systems stand to benefit from a hybrid twin approach.