Discovering Digital Twin: Experiences and Resources

[Hraabo]

TL;DR

Digital twin

So I was wondering what experience you guys have in working with Digital twins?

I am currently researching for my graduation project, which is about digital twin. When looking for litterature about digital twin, the amount available is somewhat limited. So what I'm basically asking you guys about is the following:

1. Can you recommend any litterature(books,thesis etc.) of digital twin
2. do you guys have handson experience with any specific software, when working on digital twin, that you might recommend?

 

[DaveC426913]

I've never heard of the term. There are lots of Goggles for that term, but I'm going to guess which one you mean, and post it here so that others don't have to look as dumb as me.

A digital twin is a digital replica of a living or non-living physical entity. Digital twin refers to a digital replica of potential and actual physical assets, processes, people, places, systems and devices that can be used for various purposes. The digital representation provides both the elements and the dynamics of how an Internet of things (IoT) device operates and lives throughout its life cycle.
https://en.wikipedia.org/wiki/Digital_twin

 

[Hraabo]

I am also new to it all, so I didn't even know there was several terms for it.

The link you added, is the term I know of, so thank you sir for clearing possible misunderstandings up.

 

[anorlunda]

Like a flight simulator?

 

[member 656954]

Summary:: Digital twin

Can you recommend any litterature(books,thesis etc.) of digital twin

It is a little hard to know what to propose without an idea of what you've already looked at, @Hraabo , but most of the DT work I've been involved in is with utility asset management and planning teams. They are either looking at extending their SCADA systems into a DT representation for better real-time operational decision making, or wanting to improve the accuracy of their multi-decades capital allocation process via scenario analysis.

This paper seems typical of the asset-centric view of DT, which is increasingly of interest as IoT sensors are rolled out in volume. One of my clients is currently working through the architecture for over 1M sensors throughout their network and expect the DT aspect to be handled by various software components, primarily Maximo and Cummulocity. The software products will be SaaS and the IoT comms handled by a telco partner, so from their perspective, the devices represent event and time series data sources that an enterprise historian collates for use within their asset management and customer billing ecosystem.

To be fair, most of the DT work we're involved in still in exploratory stage. It is a terrific-sounding concept, but finding concrete ROI for the implementation can be a challenge. Some of the ideas we've come across - like DT coupled with VR/AR so maintenance and installation teams can "see through" walls to spot equipment in a pumping station or look "underground" at network infrastructure - has a lot of benefits but the cost to develop the model from what is usually inaccurate or not high enough resolution data means it remains one of those 'good ideas, but...'

In terms of other software, https://www.g2.com/categories/digital-twin highlights that there are many products out there. I can't personally recommend any of them from extensive use, unfortunately, but what when you ask for a recommendation, do you have a set of uses cases in mind to narrow down the field? Or were you just looking to try some software out to get a feel for how they operate?

 

[Hraabo]

Thank you for you reply, both of you.

Well to be a bit more on point, I have to make a DT of the university Power lab, which include units such as Wind turbines, solar panels etc. I am very well aware that I won't be able to make DT of everything, but I was thinking of doing of the wind turbine(residentiel size type).

I am no way near doing it with VR og AR, but I am supposed to build a virtual model which completely represents the physical one.

Which also leads to my follow up question of the software. What is used? So fare i only know of Matlab and PTC, which can be used to some degree. What is used in the field and not at "school"?

 

[member 656954]

Well to be a bit more on point, I have to make a DT of the university Power lab, which include units such as Wind turbines, solar panels etc. I am very well aware that I won't be able to make DT of everything, but I was thinking of doing of the wind turbine(residentiel size type).

Depending on how the Lab is run, you should be able to readily source asset hierarchies and common data models for such devices from the asset management system used within the Lab, but otherwise vendors often provide details for this (GE has https://www.ge.com/digital/documentation/predix-apm/INjA5YmFiYzAtM2IzZi00MDhjLWJjY2EtNWFhM2JmYzY3NTM3.html#r_apm_gaa_wind_data_model as an example of a wind turbine). The sensor stream to inform the DT in real-time / near real-time will be complex so emulating that will be a challenge (https://www.ge.com/digital/documentation/predix-apm/INjA5YmFiYzAtM2IzZi00MDhjLWJjY2EtNWFhM2JmYzY3NTM3.html#reference_lvt_fjj_y3b GE wind turbine resource illustrating the types of fields being collected as sensor data). Are you able to carve off a current sensor stream from devices in the Lab? If there is an enterprise bus (think Mulesoft or Azure Service Bus or webMethods) that should be achievable, if not straightforward.

Have you seen the imodeljs.org open source DT software stack? That is an excellent example of what's in a DT, including details of the ancillary systems and data models.

But the most important aspect to start with is the business problem your DT is going to solve. Building the necessary models and connecting to sensor streams is time consuming so it helps to visualize the end point objective and work your way back. Otherwise, you'll end up with a nice model that can't actually answer anything useful about the assets it is emulating.

 

[anorlunda]

It is a little hard to know what to propose without an idea of what you've already looked at, @Hraabo , but most of the DT work I've been involved in is with utility asset management and planning teams. They are either looking at extending their SCADA systems into a DT representation for better real-time operational decision making, or wanting to improve the accuracy of their multi-decades capital allocation process via scenario analysis.

Please tell me that you are not adding IOT devices to the power grid and utility SCADA. You may think that the applications are strictly non critical, but the demagogues who scare the public will make it sound like end of the world.

 

[member 656954]

Please tell me that you are not adding IOT devices to the power grid and utility SCADA. You may think that the applications are strictly non critical, but the demagogues who scare the public will make it sound like end of the world.

It's off topic to the OP, but customer messaging depends on the the nature of the IoT rollout and we don't usually account for demagogues in the journey maps

But just between us, these IoT projects are secured more precisely and explicitly than many SCADA networks, some of which adopt a 'security by obscurity' approach. Sign of the times of course, older kit is harder to secure, but security is a required design criteria and customers are funding it.

 

[rbelli1]

some of which adopt a 'security by obscurity' approach.

Just slap a robots.txt file on your public facing web panel and call it a day. That makes completely invisible to all web crawlers... right?

Better yet use port 81. No one will think of that!

Unfortunately these suggestions are more sophisticated than some devices use.

BoB