Digital Twins Demystified
If you’ve heard the term Industry 4.0, you’ve also probably heard of a digital twin. There is a lot of hype surrounding digital twins, and we have built a number of them using various methods for the data component. In this post, we’ll explain what Digital Twins are along with how they can be built and interacted with in manufacturing scenarios. We’ll also help you understand why you may—or may not—want to consider a digital twin in your facility.
What is a Digital Twin?
At the most basic level, a digital twin is a representation of a process, and/or the individual assets/equipment you can interact with—independent of the physical hardware. A digital twin is like a flight simulator for a factory instead of a plane. You can set up various training scenarios to help operators experience interacting with the process without having to waste raw materials or incur utility costs. A digital twin can also help you gain a deeper understanding of how your people and equipment will interact in the real world.
A Digital Twin consists of two major components:
The data used to run everything inside of the digital twin.
The interface for interacting with the digital twin.
Since the data is usually the more complex component, we’ll dive into it first.
Getting Data In and Out of a Digital Twin
For a Digital Twin to be of any value, it needs the ability to use data to operate. The data may be as complex as a fully fledged physics simulation—similar to making a video game version of a factory—or as simple as a way to see how your plant performed at any given point in time. The entire spectrum of complexity can provide value, although costs exponentially increase as more granular detail is added to the digital twin.
Historical Playback
The simplest approach to building a digital twin is to use data from your existing Process Historian, make a copy of your existing SCADA system to look at it, and build out a historical playback tool to replay the data as the process. By using a SCADA platform like Inductive Automation’s Ignition, building the historical playback functionality is not especially complicated. There’s even a resource on the Ignition Exchange enabling historical playback in Ignition Perspective.
Using Historical Playback is by far the easiest and cheapest solution for trying a limited form of Digital Twin functionality. But, you will only be able to see how your process acted during a period of time—and it will be limited to what your operators did in that scenario. While this can be useful as a root cause analysis tool or for overall training, the insights into your process will be limited.
Simulator Based Designs
To go one level deeper than historical playback, you can build out a process simulator for your digital twin initiative. The benefit of a simulator over Historical Playback is that you are not limited to what actually happened in your process—and you can build out various scenarios to use as training tools.
A basic example is a pump is filling a tank from the top, with a valve that opens to release liquid from the bottom of the tank. You can simulate the process with a few variables for a fairly realistic result you can use for training. Set the level in the tank to rise by an amount based on how fast the pump is running, and to lower based on the valve’s percent open. You can then simulate what happens in various scenarios (e.g. if the valve is stuck closed while the pump is running) and see how operators react.
Your simulation can also be very complex, using many variables to represent all of your process conditions. The simulation can include equipment like conveyors with simulated work-in-progress materials moving along, photo eyes to track locations, and even barcode scans to trigger various downstream processes.
The hurdle with simulators is that you need to understand your overall process enough to simulate it effectively. Then, you need to build out the simulator in such a way that it can interact with your SCADA system as a useful training tool.
Simulators are also limited since they can only do what you build into the simulation. Unfortunately, the results can quickly become stagnant from a training perspective if you can’t spend the time and effort to build out more involved simulations.
Model-Based Control Implementations
Model-based control implementations move beyond process simulators—and begin to tap into the real power of digital twins. By using model based control tools like Simulink from Mathworks, you can build out a very detailed representation of your process with physics in mind. You can automatically output a simulator from your model and design interesting scenarios by tweaking setpoints—without the need to build entirely new scenarios each time.
Another benefit to model-based control is that you can output PLC code directly from Simulink. This makes moving from the digital twin to the real world a very easy process. As you use the actual system, you can continually update the model with more granularity. Eventually, you can develop a Digital Twin with a very close approximation to your process.
Interaction with Digital Twins
After the data is in place, it’s time to decide how to interact with the Digital Twin. Just like building the engine itself, you have many possibilities: from a basic view into the process, to a fully virtual environment you can “play” like a videogame.
Dashboards and Reporting
Dashboards and reports are the easiest visualization tools for a window into the digital twin and results of various simulated scenarios. A simple example would be an Andon Screen for monitoring the simulator while it is running, and a summary report after the simulation has concluded.
This example could be useful as an introductory training tool for operators see how the process operates. But, it isn’t very robust for problem solving. Normally this type of approach would be used in conjunction with a SCADA or a virtual system to provide real-time analytics and after-action reports to review with your team.
SCADA Interfaces
A Corso Systems customer (who is heavily invested in a Simulink model to generate PLC code for their physical process) has a Digital Twin version of their SCADA system for operator training. It is identical to their production SCADA system, except for the background color and text callouts that indicate that it is not the live version and should only be used for training.
They have set up a PLC to run a simulator for the model, and can deliver various scenarios for operators to troubleshoot before they move into the real control room.
This is a very powerful training tool, since it has the same interface the operators will use day to day in production. It will give operators the experience of running the process without having to worry about anything in the physical world. Operators can learn how to spot anomalies and handle them without harming people, equipment, or the product. Any issues in production can be recreated in the simulator while everyone figures out how to avoid the problem in the future. Finally, this Corso Systems customer can test out various changes to their SCADA screens and get feedback from operators before making changes to production.
Using a SCADA interface identical to the production environment is a low-cost high-value approach that many companies can use without much additional hardware or software. It might not be the most marketable concept ever—and it certainly isn’t the sexiest technology—but it is extremely effective for the job it is designed to do.
Virtual Reality
The last option for Digital Twin interfaces is the full virtual reality route. This interface could be implemented with the Unreal Engine or the Unity Engine which are fully featured video game development systems. These systems can tie into the same data streams you find in a manufacturing facility including OPC-UA and MQTT.
These tools allow you to literally create a Digital Twin at any level of detail you are willing to pay for. Anything from this simple box represents this machine all the way to incredible details such as here are the wires inside the cables running to this control panel—and everything in between.
Some companies use this type of technology for tasks like robot arm programming, building out complex packing and shipping systems, tracking real-time status of products on the line—all with the option to tie into physics engines for managing details like the flow of fluids throughout a process.
A Virtual Reality interface can also be leveraged on top of your SCADA system and dashboards/reports. You can build in digital Andon screens visible in the virtual realm to monitor what the process looks like while it is running.
Some companies are using equipment like Oculus VR headsets or Microsoft Hololens glasses to give their operators a fully virtual experience where they can walk around inside the digital twin and interact with equipment, or use augmented reality technology for a heads up display view of equipment when they look around the process on site.
The cost and resources involved in building the necessary digital assets and adding them into the Digital Twin Engine of choice are the main hurdle for these VR/AR approaches. Some companies are offering off-the-shelf digital assets for manufacturing equipment. Similarly, the popular CAD programs that most companies are already using to design equipment can export assets as well—but there are usually many more steps to take before the assets are usable for a Digital Twin.
Including animations may seem like a good idea, but they can get into “uncanny valley” territory—and a very deep rabbit hole of tweaking assets in the virtual environment. That said, one approach to lower the project’s cost is to give it a “video game” look and feel instead of attempting realism. This allows you to focus on the data instead of attempting a 1:1 match with the facility down to every last nut and bolt.
Why Use a Digital Twin?
Digital Twins are great for a few things:
Training
Wow Factor
Simulating Changes
The training benefits of a digital twin should be pretty obvious: you can give your operators free reign to learn how the system functions with no real-world consequences. You can set up various scenarios with no impact to production and train your people as new products or process changes happen. You can also use the digital twin to train new people before they go onto the plant floor.
The more effort you put into the engine behind the Digital Twin, the better the results are from a training perspective. You could even be very clever with your modeling and build in visible “wear and tear” on equipment to simulate how regular maintenance or a lack thereof may impact production.
The “Wow Factor” is harder to quantify and seems to be where much of the digital twin marketing is focused. Holographic displays straight out of a science fiction movie or advanced interfaces where people are projecting screens up on walls like they are Tony Stark are examples of this “Wow Factor”. But, while these implementations may be cool to look at, they might not yet exist in reality. More importantly, they may not actually provide your operators with additional value that’s beyond a more streamlined, simplified digital twin system. However, there is a lot to be said for the ability to show off cool technology to your customers and even your competitors.
An emerging idea for digital twin use is to simulate changes on the digital twin—then run historical data to verify a satisfactory result—before making changes on the plant floor. The changes may be as simple as a small equipment or setpoint change, but they can be tested out with real historical data from the plant before committing to the changes in reality (with real world consequences such as downtime).
Wrapping Up
With the convergence of a few technological advancements over the past few years, Digital Twins are gaining ground in the manufacturing industry
Implementing innovation in this space will require simplifying the work of creating complex art and assets. By using tools like Unity and Unreal Engine combined with MQTT, it is relatively easy to get data into a Digital Twin. If you can accept using assets that aren’t necessarily a true 1:1 to your equipment, you can even get a digital production line pulled together in a matter of hours or days, not weeks.
Many companies are working on building extremely robust models of their plants with very granular levels of detail. This not only helps from an operations and training perspective, it also gives them an amazing showpiece for customers, investors, or news outlets touring their facilities.
