Developing products that meet every design and cost requirement can be an extremely expensive and time-consuming process. And nowhere is this truer than the automotive industry.
It is estimated that 99% of valuable engineering data is never recycled or re-used because of complex, often incompatible data formats – such as CAD, CAE and CAM. But it doesn’t have to be this way. Past design engineering data can be utilised and monetised. For example, the automotive industry could cut 6 costly weeks off track testing programmes simply by combining historic track data with contemporary wind tunnel trials. And by optimising past test campaigns, the car industry could cut future track tests by a staggering 70%. (Source: Monolith AI)
But don’t stop there. Just imagine that you could develop a perfect new car design and comprehensively test it for performance without cutting any metal…
Cue digital twins.
This is one of the most talked-about concepts in technology. The notion that a virtual representation can serve as the real-time counterpart of a physical object – or process – is hugely liberating. So, not surprisingly, the impact of digital twinning has been instant and far-reaching. But, amid all the hype and excitement, it’s easy to forget that this is a very new concept. It is barely ten years since NASA started using AI to improve the design of their rockets. And one of the first people to be head-hunted to NASA’s formative digital twinning programme was Richard Ahlfeld – the founder and CEO of Monolith AI, a member of the Clustre ecosystem.
Combining a PhD background at Imperial College, immersive NASA experience and Founders Factory support, Richard leads an eclectic bunch of engineers, data scientists and software developers. Together they pursue one simple, but very clear vision: ‘The firm belief that traditional engineering will soon change dramatically as it is reimagined by Artificial Intelligence.”
Since, digital twinning is the master key to unlocking this vision, I recruited Richard’s support to answer the most basic question of all…
“Basically, there are three ingredients that usually comprise a digital twin – the first two are compulsory; the last one is ‘nice to have’ but not absolutely necessary…
First, there must be a digital model of something that actually exists in reality. This is an absolute imperative. Second, I need data – ideally data I have measured on something real and in real time. Third – and certainly not least – I always look for clever insights created through analytics, pattern recognition, physical calculations and physical simulations that tell me what is going on”.
Well, for a start, it’s not just any model you may have created. And it’s also not just any data that you have collected. For Richard, the most critical part of any digital twin is the ‘twin’ element and not the ‘digital’ component. Let me explain with some examples…
A Rolls Royce aero engine with sophisticated performance monitors is a definite candidate for digital twinning. In fact, aero engines are the stereotypical digital twins. Some while ago, Rolls Royce changed their entire business model so they could digitally monitor engines around the clock. It’s remarkable to see how the company can simultaneously track all of their engines, in the air, anywhere in the world. To Rolls Royce and its global customers, this uniquely tangible benefit is nothing short of spectacular.
Now imagine a very different scenario. Picture a theatre of war with ships, aircraft, drones, tanks and infantry deployed on multiple fronts. This is another arena where digital twins are making a dramatic impact. The MoD now uses this technology to ‘war game’ tactics for future conflicts. Every weapon system and human deployment is meticulously represented and can communicate crucial data to ‘headquarters’ in real time. Analytics can then strategically stress-test every decision, every order and every shift in battlefield dynamics. If computers predict a move, the consequences can be mapped against any possible response from the enemy. In terms of benefits, this has to be the ultimate real-life, real-time, scenario planning tool,
Similarly, the US government is using digital twins to evaluate the performance and utilisation of key assets – such as the police and hospitals. Covid has taught us that even well-ordered and heavily resourced emergency services are vulnerable in a fluid, very unpredictable world. Obtaining fast, accurate insights on vital services is an immense benefit. Digital twinning defines these priorities, shapes plans and optimises investment.
But let’s look at a practical example of digital twins in action. An example that is critical to our planet’s climate-challenged survival…
The North Sea is one of the most dangerous stretches of water in the world. A graveyard for more than 50,000 wrecked ships…
130 kilometres off the east coast of England lies the notorious Dogger Bank. A shifting, submerged sandbar feared for its unpredictable currents and vicious swells. But these treacherous shallows are now the unlikely location for the largest offshore windfarm ever built. When completed, this £9 billion development will generate enough electricity to power 5 million households – 5% of the UK’s energy needs.
The scale and complexity of this project is mind-boggling. A new generation of turbines is being created to withstand everything that the North Sea can throw at them. Standing over 850 feet tall and with blades longer than a football pitch, these are the true giants of power generation. A single rotation of the blade is enough to power the average home for 2 days!
The challenge of designing and stress-testing such leviathans has proved daunting. But this is where digital twins excel. And, because Richard has been closely involved in such projects, I took the chance to probe his deep experience…
“To understand the practical science, I and my engineers create virtual scale models. We experiment with different heights of tower… different lengths of blade… different numbers of blades… different angles of wind-to-blade attack – from zero to 90 degrees… different turbine gear ratios to optimise power generation in every weather scenario. We test every conceivable option and parameter in our search for the ultimate performance”.
The lessons learnt provide the baseline data for digital twinning. And one of the great unseen virtues of machine learning is that an engineer with little or no coding experience can confidently use the Monolith platform to build AI models, create intuitive notebooks, design workflows and automatically generate dashboards.
Using multiple sources of data, the team can then start to attack a host of critical design and engineering questions. Will such slender towers withstand the shockwaves of a Force 10 storm? Will a windfarm of 190 close-proximity towers create its own destructive turbulence? Will the blades – each one twice as tall as Nelson’s column – become dangerously destabilised in extreme gales? Perhaps most critically of all, how do you maintain and repair these massive structures in such a hostile, inaccessible environment?
Hundreds of ‘what-if’ questions are asked in the search for vital answers. And, as Richard can testify, the solutions soon become clear. The astonishing accuracy of ‘twinning’ shaves massive chunks of time and project cost. It makes the engineering process significantly more efficient. It tests and proves alternative turbine designs without ever leaving the lab. It can trial techniques for changing turbine components at sea, from the safety of dry land. And crucially, it gives engineers the tools to avoid costly mistakes and delays by forecasting the effects of their decisions months ahead.
I started this short article by describing digital twins as a virtualised representation of physical things. And this still holds true. ‘Twins’ are being enthusiastically adopted by everyone from carmakers to aircraft manufacturers, town planners to nuclear power station designers. But I also believe that we have only started to scratch the surface of the future revolution.
In a recently published white paper, Nokia made this observation:
More and more, digital twins are also being applied to systems, processes, behaviours and even other digital phenomena. A digital twin can be essentially an abstraction of any “thing” used to achieve a goal or outcome based on the analysis of a purpose-defined data set.
When virtual simulation morphs from the physical to the abstract, the scope for innovation is limitless. The power of 2 then becomes an unstoppable force for change. Bring it on.
If you would like to learn more about digital twins or informally meet Richard for a deeper discussion, please get in touch. I would love to introduce you to the true subject experts. Just email firstname.lastname@example.org