Artificial Intelligence #59 - An Enterprise Digital Twin architecture

Firstly, this week, we crossed 50K subs in just over a year. Thanks for your support!

In the previous post, we covered Digital twins vs traditional IoT applications

In this post, we will explore digital twins architectures

One of the challenges for designing a digital twin architecture is: the idea of digital twins is mostly manufacturing (and building / infrastructure ) focussed

But that does not necessarily cater to the enterprise

I found a source which covered a digital twins architecture from the standpoint of the enterprise.

The architecture is based mainly on Azure Digital Twins - ?

Physically, the twin is a combination of knowledge graph, AI Models, simulation models ?and visualization

The architecture components are as follows:

Data Platform: concerned with tasks such as ingestion, Persistence, Management, Integration, AI/ ML/ Analytics, Microservices

Digital Twin Framework: comprised of the Digital Twin Engine and Digital Thread. First a knowledge graph is used to represent a collection of interlinked descriptions of entities corresponding to real-world things, places, business processes and people. In Azure Digital Twins, the knowledge graph is created using an open modelling JSON like language called Digital Twins Definition Language (DTDL).

The Live execution environment instantiates (brings digital twins to life) in a live graph representation. The inputs are from IoT and business systems including IoT devices. Output is to Time Series Insights, storage and analytics.

Digital Twin Solutions: Visualisation, Consumption, Intelligent Workflows, Intelligent Operations, Assets and Simulation, AR/VR

Governance & Operations: DataOps

Operational Processes: MLOps, DevOps,

Edge Services

?This is a good representation and I like the approach and the depth. It also contrasts with the range of IoT / Manufacturing / Building architectures for twins

Some more comments

1) Live execution environment is an interesting concept to instantiate twins using a graph. Digital models in Azure Digital Twins are live, up-to-date representations of the real world. Using the relationships in your custom DTDL models, you'll connect twins into a live graph representing your environment. You can visualize your Azure Digital Twins graph in?Azure Digital Twins Explorer. I suspect instantiation would be by use case i.e. you have a number of persistent objects and you have a problem statement?not all objects need to be in memory

2) Simulation models are complex covering at least Discrete event simulation, System dynamics and Agent based modelling

3)????We can also contrast with other established simulation driven products like ?goldsim which are basically monte carlo simulations – however, they are not twins (as in system of systems) and they do not have full data pipeline as we see here.

4)????State modelling and replication for digital twins is a complex topic

5)????Finally, there could be a lot more work done on the data lakes for digital twins

Image source and reference: Sujay Nandi - https://medium.com/@sujnandi/building-the-digital-representation-with-digital-twin-using-microsoft-stack-86f339981871