Having a digital twin that acts like a proverbial genie from the lamp may sound like a fairy tale, but the subject is actually pretty technical.
Now that solutions within the domain of the Internet of Things (IoT) are finally maturing beyond device connectivity and management to real solutions, there is need for analysis to support advanced use cases like predictive maintenance. In order to do that, a proper understanding of the context of the connected assets becomes more pressing. This is where the digital twin comes in, which is a computerized companion mirroring the state of a physical asset.
A digital twin has to determine three things about its physical twin in order to remain relevant:
- How were you doing?
- How are you doing?
- How will you be doing?
Although that might sound simple, the challenge is immense because your digital twin has to cover hindsight, insight and foresight – something that no single technology can do at the moment. It has to be your virtual genie in the lamp.
In order to create a digital twin that can answer those three questions in real time, a number of technologies will probably have to be forged. I think it will be by combining graph database technology with time series database technology and complex event processing capabilities.
The graph will basically mimic the asset by representing it in the form of nodes and relationships, as graphs are really good in quickly adopting to unstructured data. Then new insights (read nodes and relations) can be added on the fly, extending the context continuously.
CEP filters, analyzes and enriches the data that needs to be stored in the graph. The graph stores the most recent values on the relationship and offloads the previous values into the time series database. This way, a digital twin is created that can not only continuously adopt itself to the changing behavior of a physical thing, but can also serve anyone with information about the past, the present and the future.
Because the digital twin was needed in the space of IoT, common thinking links digital twins to physical assets, or Things. That is a misperception that will quickly dissolve as companies in other industries adopt twins as an ideal way to administer not just the behavior of Things but also of their customers. Where it was the fashion to talk about Customer 360 views a few years ago, the real-time implementation could quickly be replaced by the “customer digital twin.”
Finally, I want to discuss context. While people think that the digital twin exactly represents a physical asset, there is nothing stopping companies from enriching the digital twin with features that the physical asset doesn’t have. For example, think about a switch in a railroad system; wouldn’t it be great if the digital twin giving an exact representation of its state could also have an agenda? Then it could participate in re-planning exercises, i.e. if a train is late and needs a new time to cross the switch, the digital twin of the train could call out to the switch if it is late: “Is it still safe for me to pass?” In the quest for autonomous behavior of things like cars, Things must be able to ask such questions.
To conclude, the role of the digital twin is first and foremost to enrich decision systems. It then can become an ideal interrogation candidate if it can collect as much as data as possible on any asset, whether Thing or human; answer the three questions and enrich itself with nice features that the original object doesn’t even know about. Let your digital twin genie out of the lamp!