Digital twins - demystified

In order to understand this raw data, companies hire data scientists to put together predictive models. Yet this falls short of the value it brings to an enterprise. Further the operational real time data provides only partial truth to solving business challenges.

Our objective is to demystify the term digital twin along with the approach, which is often wrongly compared to topics like Internet of Things (IoT), AI/ML or terms like predictive models, CAX, augmented reality (AR) / virtual reality (VR) models.

While digital twin isn’t getting the same level of attention as the trending topics mentioned above, there is little doubt that it will become an integral part of all businesses (not just manufacturing), optimizing products, processes and communication in an enterprise.

Demystifying the digital twin

The term Digital Twin was originally coined to define physics-based simulation of products or assets. It is a duplicated model that represents a physical object or process which can be analyzed using computer-aided engineering (CAE). This was also termed as virtual prototyping and did not necessarily need a physical counterpart in the real world. For example, space agencies referred to this view of “digital twin technology” for next-generation launch vehicles, as model-based twin.
The last decade has seen physical products getting smarter and process dynamics being captured, thanks to cheaper communication and sensor costs. The objective was to try and understand how products and process work by building a software-based model, with rules, algorithms and math, that define the working of an asset/process. Such models are called data-driven twins. These are empirical representations, which capture the data describing how an asset or process works without asking the experts who designed, built or ran them. These models are particularly interesting when we analyze a whole system by empirically understanding all the assets that create it, all the lines in a plant, and all the plants in an enterprise or supply chain.

The true digital twin

When we refer to digital twin, we are neither talking about a control model nor a simulation model nor a CAD, AR/VR model. We refer to digital twin by building a holistic model of assets, systems, and lines to understand what is happening and to establish a closed feedback loop with the physics-based model using data feeds in real-time.

The holistic digital twin model contextually binds together diverse digital data threads – enterprise IT data, field OT data and semantic graph data, and uses trending technologies (IoT, AI, ML, AR) – so that they can be used to look at challenges and business performance objectives. In this case digital twins meta model is created using data from more than one of the five main categories:

1. Design data (PLM, CAX, FEM, Thermodynamic, Geological)
2. Visual data (CAD, AR/VR, GIS/GEO)
3. Transactional data (ERP, MES, SLM, EPM)
4. Time series data (IoT) & historians
5. Semantic data (BPM, KM)

What challenges are we solving?

Through digital twin we solve business issues by addressing the data challenge. Most customers are looking at vast amounts of IT and OT data and figuring out how to use them to continuously bring about business improvements. Most companies end up spending valuable time and money in figuring out their value in isolation. Companies seldom consider the value of expert knowledge about their assets, systems and processes, which is captured through semantics knowledge graphs. Digital twin meta models are built through contextualization of IT, OT and semantics data to support applications by providing comprehensive decision-making insights.

In the pursuit of realizing a successful digital transformation and enabling new business models, the industry focus has been around the hype topics of the IoT, AI / ML, AR, additive manufacturing (AM) and more. But digital twin has truly proven to be the real game changer.

Topics

• Digital Twin
• Tech Insights