Birthing the "Fedetwin"

It's project review season. Over the last couple of weeks I've been assessing a slew of projects for academia, accelerators and industry associations that are considering developing standards, funding innovative projects or accelerating Industry 4 adoption in manufacturing. Yum. Brain candy.

Although the details of these projects must remain under wraps, what was exciting was that virtually all of these efforts incorporated one form of AI or another. What was unexpected however, was the number of these efforts that suggested using AI in combination with digital twins, a concept which may seem somewhat redundant but actually isn't.

Okay so refresh my memory. What's a digital twin again?

An up-to-date representation, a model, of an actual physical asset in operation—think: pumps, engines, power plants, manufacturing lines, or a fleet of vehicles. The digital twin can be a model of a component, a system of components, a system of systems, or even a process.

The real value of the digital twin is that it is active, not passive and usually the data is streamed in real time to a tuning algorithm. So it can be used to evaluate the current condition of the asset (e.g. a pump is running hot) against its history (e.g. it's usually luke-warm), and more importantly, predict it's future behavior (e.g. yikes, the pump will break down next Friday at 3pm when the temperature exceeds safety limits, if all things remain the same). In our digital world, not only would the data be readily available so action can be taken, but a shut down to fix the pump could be scheduled during off-hours long before it's a safety hazard. Moreover since the digital twin can also be used to refine control (e.g., slow the speed to lower the temperature) or optimize an asset's operation (e.g.,change a part to lower temperatures) the result may also be lower power consumption for cost savings and/or in support of an ESG initiative.

Among the most common types of digital twins for manufacturing, are:

1.????Equipment digital twin: Represents a physical machine or piece of equipment in a digital form. It can be used for monitoring, simulation, and optimization.

2.????Production line digital twin: Represents an entire production line, including multiple pieces of equipment and material flow. It can be used to optimize overall production processes.

3.????Factory digital twin: Represents an entire factory, including multiple production lines and support systems. It can be used to optimize the overall operations of a factory.

4.????Product digital twin: Represents a specific product, including its design, production, and usage history. It can be used for tracking and traceability, as well as for performance optimization.

5.????Supply chain digital twin: Represents the entire supply chain, including suppliers, transportation networks, and distribution channels. It can be used to optimize supply chain operations and reduce waste.

Each type of digital twin provides a different level of detail and insight into different aspects of manufacturing operations. But what if you could use a combination of these digital twin types? Manufacturers could gain a comprehensive understanding of their operations and make data-driven decisions to optimize their processes from product design to recycle or disposal.

If only it were that simple. Like everything else in tech the devil is in the details.

Modeling methods can be grouped into two basic types: first-principles or physics-based methods (e.g., mechanical modeling) and data-driven methods (e.g., deep learning). Composites do exist, but they are technically very challenging and usually evolve over time as new or expanded use cases emerge.

For most organization's digital twins are either too engineering centric so funding these is difficult, too narrow in scope, or are challenged by legacy systems and data silos that make getting the data and doing the modeling quite difficult.

But what if there was another way to skin the cat? To overcome the issue of siloed systems and be able to expand the scope of the twin? What is we could make the digital twin cohesive and complete? Solve multiple challenges instead of one by one? Remembering that dreams of cobots still dance in my head like sugarplum fairies should, enter the "Fedetwin", ta dah.

The term "FedeTwin" is a portmanteau combining the words "federated" and "digital twin" and refers to the concept of multiple digital twins working together in a connected, decentralized environment.

In this context, the term "federated" means that the digital twins are connected and coordinated, allowing for greater collaboration and information sharing among different systems and organizations.

How could one accomplish that? Hint: Metadata might work, semantic modeling is another alternative.

But the idea of FedeTwin is to create a comprehensive and integrated digital representation of the physical assets or systems that can be used for various purposes, such as monitoring, analysis, and decision-making.

By having multiple digital twins work together in a federated environment, manufacturers could gain a more complete and accurate view of their operations, as well as improved efficiency, flexibility, and resiliency.

How can one harmonize different digital twin types?

1.????Define common data structures and ontologies: Ensure that all digital twins share a common understanding of key concepts and data elements. Where to begin, perhaps with a Bill of Materials...

2.????Establish a common data exchange format: Choose a standardized format, such as JSON or XML, to ensure that digital twins can easily exchange information.

3.????Use APIs to enable communication: Implement APIs to enable digital twins to exchange data and interact with each other in real-time.

4.????Define common governance and security policies: Ensure that all digital twins comply with the same security and data privacy policies. Identity might be a start point.

5.????Implement data reconciliation and synchronization processes: Regularly compare and reconcile data across digital twins to ensure that they are in sync and accurate. Metadata?

6.????Monitor and adjust: Continuously monitor the digital twin ecosystem and make necessary adjustments to ensure that all digital twins are harmonized and working together effectively.

While I have no doubt this is an overly simplified example the Fedetwin concept is one that might help manufacturers not only adopt digital twins more readily but make far better use of their data. After all, tomorrow is Friday and that overheating pump could spoil more than a few people's weekend.

#AI, #Digitaltwin, #Federated #Metadata #Fedetwin