In the realm of interconnected devices and the Internet of Things (IoT), the term “digital twin” often finds its way into discussions. You might have come across this concept in your favorite podcast, a blog post, or the latest cutting-edge technology update. It’s apparent from the various sources you’ve encountered that a digital twin holds the promise of addressing many challenges in our hyper-connected world and is viewed as a key element of the Metaverse.
So, what exactly is a digital twin? Is it a 3D representation of a factory? Is it a virtual embodiment of your digital self within a virtual world? Or is it a digital representation of a network of interconnected devices functioning seamlessly and almost autonomously? Perhaps it’s a complete replica of a physical environment, utilized for purposes such as analysis, design, operations, and simulation. According to Michael Grieves, the Executive Director of the Digital Twin Institute, a digital twin comprises three essential elements: a physical component, a virtual counterpart, and a communication link that bridges the
two. The digital twin’s primary role is to support the physical product throughout its lifecycle, providing insights, simulations, and knowledge about the physical element. It is most effective when used in complex physical systems with numerous interconnected elements.
To illustrate the concept of a digital twin, let’s delve into a real-world example. Imagine a wind farm comprising 80 individual wind turbines. For each wind turbine, a digital model is created, including their spatial positions on a map. These turbines are equipped with sensors on various components, such as the generator, gearbox, tower, pivoting system, and foundation. The system continuously measures the energy production of each turbine and records external conditions like wind speed, weather, and temperature. By amalgamating the physical model, sensor data, and weather information, a comprehensive model of the entire wind farm is developed. This model allows for various calculations and simulations. Consequently, the wind farm operator can anticipate wear and tear on components, predict service degradation, and assess the influence of wind based on the relative positions of the turbines. They can even optimize production by utilizing the wind shear of each turbine to concentrate the remaining wind on the following turbines, enhancing efficiency and overall performance.
An Enchanting Solution for Complex Scenarios
The digital twin is often touted as a magical remedy for the challenges posed by the exponential growth of IoT data and its diverse applications. It might seem tempting to simply proclaim, “Create a digital twin based on this data, then employ artificial intelligence and machine learning to optimize your model.” However, a word of caution is warranted for ventures that attempt to construct an all-encompassing model of the entire physical world, such as an entire factory, energy network, or an entire building, with the intention of accommodating as much data as possible. These endeavors tend to be exorbitantly costly and, in most instances, unfeasible. In essence, such models often serve no purpose beyond existing in a PowerPoint presentation.
A Digital Twin is just a mimic of a rudimentary set of network elements helping to solve a problem. They work together with some basic unchangeable rules that mimic unchangeable characteristics of the real world like the laws of physics. On top of this, there is a designed model consisting of nodes and relationships. This model is able to change but requires a solid change in the real world. This can be the design topology of the power network, a chemical plant, or a water network. On top of this, the model is infused with a feed of real-time data that is derived from sensors measuring the real world. From this, the interaction between the different nodes is measured and can be used to feed a mathematical model of the real world. Finally, the operator or a process can ask questions to the digital twin model in terms of what-if scenarios or extrapolate a current situation to assess future possible deviations (accidents). The model can generate steering actions from the digital simulation, that are being sent down to actuators to influence the real world. This closes the loop. A digital twin is always used with (near) real-time data to give insight, assess consequences, and derive conclusions from the real world. Each model has a specific goal and purpose (like energy savings, cost optimization of customer comfort)
According to a 2019 report by Gartner, 66 percent of organizations will be using digital twins in 2023. I have noticed a plethora of definitions and applications of digital twins in the real world. Some are beginning to deliver value.
I have categorized the following variants of digital twins:
1) The Digital Twin as an Imaginary Friend:
However, a substantial gap still exists between what is written about digital twins and their real-world implementation. In practice, the value of a digital twin is somewhat more modest and practical. I refer to the PowerPoint digital twin as the imaginary friend digital twin, as it only exists on paper. It is a good idea that has not yet been realized.
2) The Digital Twin as an Empty Dumb Model (Not Necessarily Blond):
This version of the digital twin emphasizes the user experience. It is often a 3D model of something spectacular, like a rocket, a large chemical plant, or an oil drilling platform. But it lacks the data of the real world. While the model may look impressive, it is essentially the 3D artist’s impression of the real world, an empty and “dumb” model.
3) The Digital Twin as a Single-Purpose Math Specialist:
This is the most common and practical form of a digital twin. It involves a model that has been tailored to address a specific problem. The digital twin serves as a complex real-time decision engine with the purpose of automatically making decisions aimed at a particular issue. For example, it may optimize the balance of an energy network by real-time management of production and consumption across a vast array of energy assets.
4) The Conjoint Digital Twin (Siamese Twin) Sharing at Least One Vital Component:
Certain physical assets already incorporate complex software to optimize their performance. For instance, a pumping and valve system utilizes artificial intelligence to optimize pump performance, maintenance, energy usage, and sustainability. The pumping system generates real-time data, which is fed into the digital twin model. However, decisions are primarily made by the embedded software of the pumping systems. Both the physical and digital models share the same decision-making logic, which is ideal for insights but makes simulations impossible, as the decision-making process is intricately tied to the physical world
5) The Stillborn Digital Twin
These digital twins are essentially dormant, often supplied with a singular dataset, and sometimes even historical time-series data for limited simulation capabilities. They lack the flexibility to accommodate new data, typically requiring manual updates, often done on a monthly basis. These digital twins might appear digital but remain inactive.
6) The Parasitic Digital Twin:
This type is closely linked to the physical model and often overcomplicates the digital twin’s purpose. As designers strive for an excessively complex model, the parasitic digital twin gradually consumes more project resources, including time and budget, leaving fewer resources to enhance the actual physical twin
In Conclusion: be focused and practical
A digital twin is most valuable when practically applied to solve real-world challenges. Begin with a specific issue that demands a sophisticated tool like a digital twin, rather than attempting to force-fit it into a problem. Many problems can be addressed with digital technology without resorting to digital twins as a trendy solution in search of a problem.
Here are some practical tips for maximizing the value of your digital twin:
- Initiate a well-defined business case that addresses a real problem.
- Ensure your solution leverages a relational model fueled by real-time data.
- Utilize the digital twin for decision-making and simulations.
- Automate data feeds and model updates to maintain synchronization with the real world.
- Avoid striving for a one-size-fits-all omni-model; create separate models tailored to specific problems.”