Achieving Scalability with Digital Twins

Tom Clark
5 min readMar 11, 2020

The engineering industry is alight with the phrase “Digital Twin” — a tool in the armoury of someone tasked with making decisions based on data. But what about when the system you are modelling gets really complicated?

First, why is there such a buzz? Applications are widely varied, but let’s look at one specific case: operation of a large wind farm. In a typical 600MW Offshore Wind Farm, figures suggest that digitalisation can unlock a 2% increase in yield, a 10% increase in asset lifetime and a staggering 15% reduction in operations and maintenance cost, combined worth c.£26m/yr.

Generating all this energy for £26m/yr less sure would help us tackle climate change.

This isn’t atypical of case studies where data-driven decision making is used to design, build and operate plant. These kinds of figures are pretty attractive to the top- and bottom- lines of a business, so it’s pretty clear why all the fuss.

Making a commercial argument like the above is a painful-but-necessary part of developing a digitalisation strategy for any company: this is what justifies an investment.

Next, how do we deliver that? Invariably, there are tons of working parts — economic models, optimisers, aerodynamic simulations, predictive ML/AI processes munching sensor data, fatigue analyses, etc etc. Typically, each of these working parts comes from a different team of specialists.

At Octue, when building early automated systems for data analysis (that later evolved into Digital Twins) we quickly became swamped by the number of stakeholders, analysis codes and data sources in even a simple R&D project. Herding cats is not my favourite pastime.

So, we started building a modular system: by wrapping each of these aspects into its own Digital Twin, we can have a specialist team/stakeholder providing a sub-twin then assemble all those services into a larger twin.

This is just like creating a CAD model — except with data services instead of components and assemblies.

  • Each stakeholder focuses on their specialty (high quality)
  • Each Digital Twin can run separately (limited data exposure), communicating only with its parent/children over a secured network
  • Twins are modular, so you can replace/update as a system evolves (flexible)
Tom Clark

Fluid Dynamicist at the core, lover of chaos theory. Experienced scientist, developer and team lead working in wind energy — from startups to heavy industry.

Recommended from Medium