What is a digital twin?

COSMO stands for Complex Systems Modelling. So, we are doing what we call 360-degree simulation digital twins to predict all possible futures of an organisation. We start at the problems customers are facing and provide software for their three big challenges:

The first one is they have to make decisions in a more complex and uncertain environment. Complexity comes in many forms, but in particular the fact that with everything being connected to everything, you have a lot of cascading effects.

The second thing that they are facing is that more than ever they have to balance between objectives that seem contradictory. Let's say you have to increase your revenue, reduce your cost, increase your efficiency but also resilience – and of course reduce CO2 emissions. So, all these things you have to take into account and at the same time try to be able to solve them, despite them seeming contradictory.

And the third is that things are changing very fast. So all industries have to make decisions, but also to be agile and to adapt to what is happening in real time. So, these are the three very big challenges we find with any customers and when you want to address this and you want to use technology to help you in this context.

Building a world-leading future

At Cosmo Tech, we have a technology that we have developed that can be combined with other technologies to allow us to learn from data from the past and data from the present – and data from the future. What we are able to do is to look at a system and create what we call a digital twin, so a replica, a 360 degree digital twin, meaning that we are taking the system as a whole. It's not a digital twin of a physical machine or space, it’s something bigger and more heterogeneous – like a supply chain with people, or finance or risk. We do work with people building 3D simulations, but that’s not our business.

Whether at an operational, tactical or strategic level, Cosmo Tech technology brings 360° simulation to digital twins, simulating over time the dynamics (or behaviour) and evolution of an industrial system such as a supply chain, in the context of its environment (regulatory or sanitary constraints, shortage of spare parts or raw materials, volatility of demand, weather conditions...). This allows the accurate prediction of a decision’s, a disruption or an unexpected event’s impact on the entire organisation. This capability enables better control and even capitalise on cascading effects and recommends optimal action plans.

We have partnerships with very big players like Microsoft. We are the preferred partner for simulation digital twins for Microsoft working with the Azure suite’s digital twin. Cosmo Tech has developed this unique platform that allows us or our partners to create these twins very rapidly because the language is dedicated to that.

Sometimes you don't need to be very fine-grained but you use the twin and the simulation, you build it from the top and you test, you do what we call sensitivity analysis. You check. You change a little bit of the twin and you look at whether it impacts your KPI. If it impacts the KPI then you say, “oh I have to refine, I need more data, I need more rules”, which will be dynamic rules that describe the system and the way it evolves more precisely.

The twin is geared towards the user

What is important for the business user is to have the right representation, the visualisation that corresponds to what he or she needs to make the decision. So, the 3D digital twin is specifically used for the cases where you need to visualise physically in 3D. Very often, you're using that because you're focusing on physical objects you want to represent, to have a twin of your physical production line.

You want to visualise it because what you're interested in is the physical aspect. But what if you’re more interested in the impact of the action on CO2? That wouldn’t need to translate to a 3D model. You will rather need dashboards to monitor and see the future behaviour of your system and the resulting KPIs, and you could also need to visualise the impact of your decisions on a map.

How can you forecast multiple futures?

You may know that in the field of AI, one of the big limits of data-based AI is accessibility to data. If you want to learn about one machine in your factory, you might have a lot of equivalent machines with lots of past data. But if you have one machine, it’s not enough for the machine learning algorithm to learn.

For you

Be part of something bigger, join BCS, The Chartered Institute for IT.

To counteract the data limitations, we create synthetic data – or data from the future. Gartner predicts that in a couple of years the majority of AI tools will be trained on synthetic data. When we say we can predict multiple futures, we can predict an unlimited number of futures, with unlimited synthetic data. When historical data is impossible or expensive to get, digital twin simulations bring capabilities to better react in any future situation taking into account infrequent or unknown scenarios; illuminating edge cases that would otherwise never be found.

Is digital twinning power hungry?

Michelin, reduced its transportation by 60%, I can tell you that the amount of CO2 emission that it saved was many orders of magnitude bigger than the CO2 emissions made by using the simulation. Digital twin simulation is computation to create knowledge, to make better decisions. It's a good question to wonder about the consumption and the CO2 emission and the energy consumption, but at the end of the day you have to look at what do you use it for?

Nexans is one of the world’s top cable manufacturers and a global player in energy transition. It’s committed to its investors and to its board to become carbon neutral by 2030 while maintaining financial profitability. Carbon reduction is very hard and getting gain on both sides is even harder.

You have to do that in all the factories all over the world, taking into account Financial data, Energy cost, Emission factor, transport mode… it's extremely complex. So, Nexans uses a digital twin of its whole infrastructure on a day-to-day basis. It reduces CO2 and each year saves and plans saving around the equivalent of 50,000 tons equivalent of carbon every year based on its twin simulation.

Why do people use digital twins?

People want to get rid of complexity in their systems. We don't want to get rid of complexity, we want to use the complexity to create value – whether that’s increasing profitability, reducing CO2 or both. There is no way for the world to become less complex in the future. So, the only way to address the problems we face is to use complexity. There is an incredible richness in complexity if you are able to use the cascading effects. What the simulation digital twin allows you to do is to use the cascading effect for the good. That’s very important for value creation.

What makes it worth the investment?

Michelin wanted to adapt to volatility in the Chinese market and needed to know the best way to organise the supply chain to respond to demand. After implementation, the result was 60% less transportation costs, 60% less customs costs, 5% additional profit per year. It's 10 million profit every year so the cost of the software is not the problem. The thing we first had to overcome is to get them to trust the software from a start-up company.

What emerging technologies are making digital twinning a reality?

We are using AI; we are using big data, cloud and globally I would say we are connecting with 3D visualisation with virtual reality tools.

This complex system is part of an ecosystem that is already very complex itself. It has to have different timescales, space scale – people need to make decisions in a factory, others have to make decisions about the world and they need a lot of technologies to help them to manage that.

The IoT is gathering data. AI is learning from data from the past, virtual reality or 3D representation when it's a physical object – and not when the ability to see the physical world isn’t needed. In a simulation digital twin, you experiment possible futures and use synthetic data to train AI algorithms.

What I have described is an industrial metaverse – the creation of meaning. Linking one model with another, with another, that allows someone to make decisions, to know, to have information to forecast, to see the future, or many possible futures. In the future, it will not be possible to make decisions without looking at the impact.

The future of digital twins

What is happening in particular with simulation digital twins and this ability to visualise the future in particular for industrial systems, it's going to be 100 times more important than the impact of AI we have seen in the last decade. Imagine you have the possibility to see the future. Now you can.