Tag Archives: digital twin

  • -

Playing and planning with a digital twin

Category:EA,UK Blog Tags : 

We continuously elaborate on the direction that Enterprise Architecture is taking. Recently we found the inspiring presentation from Gartner  and Marc Kerremans, talking about the construct of a ‘digital twin’ as a new and emerging concept:

“A digital twin is a dynamic software model of a thing or system that relies on sensor and/or other data to understand its state, respond to changes, improve operations and add value. It includes a combination of meta-tada, condition or state and event data”,
Marc Kerremans, Research Director.

This is interesting, as it puts focus on the ingredients of meta-data to provide a better future set of operations. So a ‘thing’ is no longer a computer…  A ‘thing’ can be an organisation, a performance, a situation or a value. It simply becomes  borderless, see also Gareth Morgan’s 8 metaphors. And there is certainly a potential to apply the known practice from adaptive control theory, where there is a blend of digital models and physical measurements which can be compared, optimised and used for improved performance.

The basic idea of a digital twin is hence back to classic cybernetics, it is to let “play” and “observe” at the same time. The real system plays it traditional way with lots of data, but behind the scene, the digital twin is “observing” using a simpler model; and eventually, this will create new controls. Some of the core material by Karl Johan Åström and Björn Wittenmark go back fifty years, describing how dynamic models relying on external sensor and/or other data can be modeled  with a known distinction to

  1. System identification and adaptation (getting the model parameters),
  2. Filtering (getting the states), and
  3. Control (adjusting the input) based on different predictive horizons and criteria.

Would it be possible to apply Kalman filtering to improve the response of a digital twin? why not! The idea of a digital twin is to apply the systems thinking to the organisation so that the organisational performance is considered a system, which can be modeled with states and parameters, and which is likely a stochastic and non-linear system with many input and many outputs. Is it controllable? Is it observable? What type of model would we end up with?

Clearly it will by a hybrid model where finance or sales is easier to measure and forecast than a lot of other performances – and likely as well, where no digital twin is one-size-fits-all, they will be tailored to specific use-cases. Where the digital twin is mostly focused on the prediction, it will benefit from the updated set of data which is also part of the living architecture.

So, the trajectory is to move the digital twin more into the direction of information management, where information models are established, and where statistical properties of the core behavior is turning measurable, so that with data-lakes and big-data that we can start to model and identify the dynamic models of the enterprise – rather than to see a bunch of tables. Not necessarily an easy task, but as stipulated by Gartner, the purpose is to provide better planning to lower risk of fact-based decisions. With a digital twin of an organisation, we can measure the input/output of an organisation, and we can forecast how it would develop. The adoption of digital twins will likely lead to models that can be used for simulation and forecasting.

The Internet of things (IoT) provides a bigger and larger system where the business operating model at heart to us should include performance measures and measurements of many variables. And to apply this in a business context, the target is to find an algorithm to forecast and optimise the predictions given different types of input.

There are however, different grey zones in this concept to get it fully operational.

  1. One is the cultural shift that one needs to consider.
  2. Another one is the statistical models to be applied, how and what,
  3. A third and not the least, the choice of ‘states’ and ‘parameters’ of the model, a model typically known as the business operating system (BOS).

We advocate to start the journey towards enabling a digital twin by three simple steps:

  1. Identifying the ‘meta-data’ initially to provide a logical layer of how an organisation is connected form suppliers to clients,
  2. In the second phase to apply states and parameters to consume measurements to connect the dots of the BOS.
  3. With focus on dashboards, outcomes and KPI’s this brings the third phase to compare targeted and measured values. This is what essentially is the fuel for the digital model to optimise your fact-based input to the strategy-to-execution.

The scenario-based planning may be accelerated by these kinds of models. Use a digital twin of an organisation to help your business leaders to minimize the risk of new growth. We advise to begin with the logical layer. This is where fly-by-wire dashboards can be produced for guiding your business journey. This is where the digital twin should be used to maximise business outcomes. The importance is the act of digital planning as well as the precision of the models. If you interested in technology supporting a digital twin of your business, we can help you! Don’t build it yourself, there is emerging technologies supporting big data and digital twins.

We power your digital MooD.