Digital Twins, a Pillar of Digital Transformation

Industry 4.0

,

Digital Advisory

Digital Twins, a Pillar of Digital Transformation

Francesco Belloni | Oct 09, 2020

Digital twins are a truly representative - and one of the most interesting - examples of digital transformation, where objects in the physical world are given virtual replicas fitted out with artificial intelligence and fed with real-time data. However, they are also subject to multiple interpretations, which tend to obscure their real practical application.

In this article, we will try to show you our personal way of looking at digital twins and provide some real examples that have come out of design thinking sessions with our clients and have been implemented by them.

 

Definition and Features of a Digital Twin

When it comes to digital twins, there are many definitions out there (a wide range of definitions can be found on Wikipedia alone).

When we look at the dates, it’s actually surprising to realize just how long this concept has been around: Michael Grieves of the Florida Institute of Technology had already applied the concept of digital twins to the manufacturing sector as the conceptual model at the base of Product Lifecycle Management as far back as 2002.

All the definitions of digital twins have three distinct elements in common:

  • That digital twins are a virtual representation of something physical, whether an asset, a process, person, place, system or device, whether currently existing (physical twin) or still hypothetical
  • That the virtual representation replicates the same dynamics and key elements that affect the physical object as it operates and develops throughout its life, based on the purpose the digital twin was created for
  • A connection is established between the physical object and its virtual representation via real-time data exchange, collected from sensors on the physical object and sent to its virtual counterpart.

It can be summarized like this:

Digital twins integrate IoT, artificial intelligence, machine learning and analytics with a graphic and/or spatial representation to make animate digital simulation models that can change and develop as their physical counterpart does.

A digital twin learns and updates itself continuously using real-time data from sensors on the physical asset in order to represent operating, environmental and working conditions in near real-time.

Digital twins learn:

  • by themselves, using the data acquired
  • from experts with in-depth domain knowledge
  • from other similar assets

What’s more, they fine-tune their simulations by integrating and using historical data.

 

The structure of a Digital Twin

Digital twins are typically made up of 3 layers:

  1. A Connectivity layer using the IoT, SCADA or historians (if you’re interested in knowing the difference read our article Comparing historians and the IoT)
  2. A Modelling and Simulation layer, which can consist of a whole range of solutions: industry simulators (thermodynamic, fluid-dynamic, chemical, and more), artificial intelligence, etc.
  3. An Insight and Visualization layer, which can be built online, with analytics tools, or even mixed reality

These 3 layers are topped off by “learning feedback”, which allows past data and expert feedback to be used to improve the behaviour of digital twins, and the reliability of the physical twin.

3 layer structure of a Digital Twin

What are the practical applications of a Digital Twin?

 

Virtual Metering: evaluating non-measurable quantities

Using field data makes it possible to gauge measurements that can’t be directly measured on the physical element due to economic or positioning factors, poor accuracy of the instruments available, or other reasons. 

Let’s imagine an old heat exchanger in a fluid preheating line. Due to its size and the difficulty of accessing that section, it’s difficult to install temperature sensors on the exchanger’s internal coil. With the Digital Twin, it is possible to determine the temperature even without direct measurements, by reconstructing it using a mathematical model of the exchanger, fed with other measurements such as flow rates and the other stream’s temperature.
By measuring this temperature, we can gain valuable information about the heat exchanger’s efficiency levels, and highlight any hot spots that are detrimental to energy efficiency - information that is otherwise unobtainable.

 

3D Model: Integrating Real-Time Data Into a 3D Model

3D plant models are becoming increasingly popular: greenfield plant designs now appear in 3D from the offset, while laser scanning can be used to create a 3D model of a brownfield plant with reduced time and costs. 

3D models can give us precise information about the obstacles facing plants and machinery, and allow us to plan with maximum precision and without plant inspections and works. 

Our vision goes beyond the simple 3D model with the Digital Twin for Maintenance concept, a combination of the latest technologies with a new perspective and a working approach that is fully integrated into digital asset lifecycle management. The goal is to expand the available information through new solutions and technologies, center it on the 3D model, and analyze this information in an integrated way by creating dashboards that provide fresh insight and facilitate the evaluation of the most appropriate maintenance and investment strategies to be adopted.

 

Parameter Estimation: Measuring the Physical Parameters

This function allows the user to estimate the value of a geometric, physical or chemical parameter which cannot be directly measured but is fundamental to ensuring the equipment works correctly and efficiently. 

Once again, let’s take our heat exchanger as an example. By knowing its dimensions and running a model of it through a thermodynamic simulator, it is possible to estimate the fouling of the exchanger day by day and obtain fundamental information to schedule the best cleaning operation time.

 

Performance Monitoring: Comparing Real-Life Performance With Model Performance

This feature allows users to detect how the real plant deviates from its digital twin in terms of yields, quality and other essential KPIs in real-time.  

This information may be indicative of impurities in the fluid, which could later lead to failure and unrecoverable products, performance problems or incipient failures, the detection of which can be anticipated through continuous and constant monitoring and an alert system that gives warnings when an inspection is needed.

 

What if? Simulating Possible Scenarios in Predictive Terms

By using historical plant data and mathematical models, it is possible to run simulations to predict how the plant may react to unusual conditions. 

Let’s imagine we have a model of a utility distribution network and want to examine how the network behaves when changes occur in the plant, monitoring, or operative conditions. The digital twin will be able to evaluate how the physical twin will behave in a predictive scenario, allowing more informed and automated decisions to be made.

 

Predictive Maintenance: Identifying and Anticipating Faults

Predictive maintenance has become widely known and increasingly popular in recent years. 

A digital twin can be a predictive maintenance solution: by starting from historical data and fault history, it is possible to anticipate when faults will occur, leaving time to plan maintenance work before a critical safety, environment, or product failure is reached. Predictive maintenance systems typically learn from feedback provided by the operator over time about the outcome of predictions made, creating a cycle of continuous improvement.

 

Digital Twin: what's next?

The advantages of using single technologies are becoming increasingly obvious, and all the more so when combined with other technologies. We have already talked about the advantages of combining digital twins with artificial intelligence and machine learning, with 3D modelling systems and mixed reality, but every day this technology offers new opportunities that can bring scenarios to life that were hitherto impossible.

One example is edge computing, a distributed computing model in which data processing takes place as close possible to where the data is produced. Edge computing is just one model to have evolved from digital twins: here, the model and the simulation are run directly on the device, and equipped with increasingly powerful and miniaturized computing capabilities, while the cloud connection is no longer necessary, but useful for sharing logic learned through an approach called “learn one, learn all” with other devices.

 

Do you want to discover the potential offered by digital twin for your business? Thanks to its technology onboarding approach, Techedge can give you a better understanding of these technologies, and identify solutions for you to implement in just a short time.

 


Other articles of this series on Digital Advisory:

 

New call-to-action

Subscribe!