Digital Economy Dispatch #086

1st May 2022

Not Subscribed Yet? Sign up Here!

A Model Future for Digital Twins

As an external examiner for a PhD thesis recently, I was reminded of my own PhD viva over 30 years ago. It was a lengthy meeting in which the 2 external examiners took me through the work I had carried out over the previous 4 years and written up in my thesis. Not surprisingly, it was a nerve-racking and painful process. In my case, this was made worse because over 100 pages of the thesis consisted of a complex formal model of a large software system written in first-order predicate logic using a formal modelling notation called Z. No sooner had I sat down at the viva and the first examiner opened the thesis, turned to page 127, and pointing to a specific model element asked if I could justify the statements. So began my 3-hour ordeal.

Ever since, models and modelling have been a major part of my career. From those PhD roots in formal modelling languages to specify complex systems, I have spent many years using more informal approaches to describe the software architecture of deployed systems, helped to design and build visual modelling tools for documenting shared understanding of system behaviour, participated in standards committees seeking to define common approaches to modelling for specific domains, taught modelling and design to students using a wide variety of notations, and much more. Given this background, it is a surprise to me how little really understood about where and how models are used. And most importantly, why they are used.

Why Models Matter

Lately there has been a resurgence of interest in models and modelling driven by the widescale deployment of IoT sensors in factories, on streets, in offices, in homes, and many other locations. Using these devices, we now have the ability to generate and share vast amounts of data in near real-time. In addition, this data is generated at scale across many devices, from many locations simultaneously, and correlated over long periods of time.

Consequently, as they become cheaper, higher quality, and more robust, the challenge has shifted from acquiring data to manging these rich data streams to enable meaningful insights to be obtained from large quantities of data. How do we use all this data to improve our understanding of what is happening inside an aircraft engine during take-off? What are factors that influence energy use in a specific home, or across neighbourhoods? What happens in your body when you take a prescribed medication and how does that vary in different patients? And so on.

The key is to use the data to construct a digital representation of the part of the real world from which this data is being captured. In many cases, the data can be used to construct a digital model of the physical system that is being monitored with a focus on the accuracy and alignment between the physical and digital. The resulting Digital Twin connects the virtual and the real worlds to form a cyber-physical system. Crucially, there is 2-way feedback between the virtual and real world system to ensure alignment and synchronization.

Toward Digital Twins

As digital devices for data gathering become more prevalent, interest in Digital Twins has grown significantly. Their use and impact has been widely cited, with application in sectors as diverse as manufacturing, utilities, construction, environmental science, medicine, and many more.

Undoubtedly, the adoption of Digital Twins offers many new opportunities. However, based on experience over many years I would suggest that some caution is required in considering the future of Digital Twins. In particular, I believe that anyone interested in Digital Twins needs to be reminded of 3 fundamental issues.

All Models are Wrong; Some Models are Useful

A Digital Twin is a model. All models adopt a specific point-of-view on the system being described. Therefore, it is by its nature a representation of a real system where choices have been made about what to represent and (crucially) what to leave out. That results in the model being incomplete, partial, and abstract. This may seem to be a failing of the model. However, perhaps rather paradoxically, the power of a model is precisely because of these characteristics. The viewpoint you take highlights areas of concern, removes distractions, and brings insights related to the perspective adopted.

This is important because a Digital Twin is not the system is represents and never will be. Keeping this in mind is critical. It allows you to consider where and when the model you are creating is useful, and also to understand that there are times it is not. Multiple models of the same system may be required, each making different choices for what is in focus. They all may have value in helping understand a system and its operation. Imagine, for instance, that you have 2 models of the same physical system and are asked which one is “better”. How would you respond?

Furthermore, because of these choices, all models represent the intentions, experiences, biases, and failings of the modellers who created them. Most real-world systems of interest are sufficiently complex that the modellers’ knowledge influences the model and leaves its fingerprints in the way the model is represented, structured, and reacts to certain stimuli.

Lesson: Focus on the choices of what is included and excluded in the Digital Twin. Use representations that support those choices to highlight areas of the real world that matter to you. Don’t expect a Digital Twin to be complete, consistent, or objective.

A Model has a Purpose; Stick to it

A Digital Twin has a purpose. From personal experience, much of the success (and failure) in modelling is a result of the (mis)alignment between the intent at the model’s creation and its subsequent use. The context in which the model was developed is particularly critical. For example, in one situation where I was working a failed system maintenance programme was found to be relying extensively on a requirements model hastily put together by inexperienced engineers to gain early insight into the data flow in a system. It was never constructed with the depth of understanding that is needed to support long-term maintenance decisions.

Most discussions of Digital Twins assume a rather rigid context in which the scope is large (a city, a factory, the human body, etc.) and the purpose for the model is broad (understanding, prediction, maintenance, etc.). This should be challenged. In practice, models have a variety of distinct purposes. Typically, they include:

  • Sketch. A high-level representation used to explore, illustrate, educate, and communicate some aspect of a system.
  • Specification. A formal description acting as a contract between different parties involved in the design, delivery, or use of a system.
  • Support. A collection of information used to deepen understanding that may document choices, alternatives, rationale, and other useful contextual aspects of a system.
  • Simulation. A description sufficiently robust to enable execution of the dynamic behaviour of some part of a system under some defined circumstances.
  • Substitution. A complete description of some part of a system with behavioural characteristics identical to the system it represents.

These examples illustrate the breadth of possibilities for a model’s purpose. Each has significant implications for how a model is created, how it is managed, and how its value is assessed.

Lesson: Understand the main purposes for the Digital Twin. Adopt modelling approaches and processes that support these needs. Be wary of requests beyond these areas.

Modelling Requires Tools and Processes; Choose Wisely

A Digital Twin is a significant, managed investment. Over time, models represent a significant investment in time and energy. Furthermore, they become vital assets in understanding the system they represent as elements of its documentation, history, operation, and future evolution. Hence, their construction, management, governance, and maintenance require tooling and processes that ensure they are accessible, available, and accurate. Using models in other domains has brought experiences on how this is best achieved.

Some years ago, while working with the Unified Modelling Language (UML) for modelling large software systems, I led a team with Grady Booch, Jim Rumbaugh, Sridhar Iyengar, and Bran Selic to propose a set of principles for managing complex models. The subsequent “Manifesto for Model Driven Architecture (MDA)” defined 3 key areas for success that I believe are essential for progress in Digital Twins. Slightly updated from the original descriptions related to MDA, these are:

  • Direct representation. Shift the focus away from the technology domain toward the ideas and concepts of the problem domain. Reducing the semantic distance between problem domain and representation allows a more direct coupling of solutions to problems, leading to more accurate designs and increased productivity.
  • Automation. Use digital tools to mechanize those facets that do not depend on human ingenuity. One of the primary purposes of automation is to bridge the semantic gap between domain concepts and implementation technology by explicitly modelling both domain and technology choices in frameworks and then exploiting the knowledge built into a particular application framework.
  • Open standards. Standards have been one of the most effective boosters of progress throughout the history of technology. Industry standards not only help eliminate gratuitous diversity, but they also encourage an ecosystem of vendors producing tools for general purposes as well as all kinds of specialized niches, greatly increasing the attractiveness of the whole endeavour to users. Open source development ensures that standards are implemented consistently and encourages the adoption of standards by vendors.

From my modelling experience, I cannot over estimate the importance of a consistent, rigorous, and standardized set of tools for the creation and management of Digital Twins. Currently, there is no well-established basis for this. Without that, progress to scale the adoption of Digital Twins will be severely hampered.

Lesson: Use tools and processes to define a discipline around the creation and management of Digital Twins. Adopt standard approaches where they Exist. Encourage their creation where they don’t.

Our Digital Future

As digital transformation increases, better understanding of complex cyber-physical systems is critical to our future. Digital Twins are an important way to describe those systems and define their key characteristics. Yet, we must not forget that a Digital Twin is a model. Models have always played a significant role in helping us to understand and reason about complex systems. There are important lessons from this experience that Digital Twins must adopt. These include recognizing the impact of the choices made in modelling, defining the usage context in model creation and management, and supporting an open ecosystem of tools ensuring models can be shared.

Digital Economy Tidbits

Microsoft 2022 Annual Work Tend Index Report. Link.

Here is the latest workplace trends survey and analysis from Microsoft. Makes very important reading for anyone interested in the future of the workplace.

One thing is clear: We’re not the same people that went home to work in early 2020. The collective experience of the past two years has left a lasting imprint, fundamentally changing how we define the role of work in our lives. The data shows the Great Reshuffle is far from over. Employees everywhere are rethinking their “worth it” equation and are voting with their feet. And as more people experience the upsides of flexible work, the more heavily it factors into the equation. For Gen Z and Millennials, there’s no going back. And with other generations not far behind, companies must meet employees where they are.

Key Findings

Five urgent trends business leaders need to know in 2022:

  1. Employees have a new “worth it” equation.
  2. Managers feel wedged between leadership and employee expectations.
  3. Leaders need to make the office worth the commute.
  4. Flexible work doesn’t have to mean “always on.”
  5. Rebuilding social capital looks different in a hybrid world.

Not yet subscribed?

Sign up now to received Digital Economy Dispatches in your mailbox…..