Interview: A Digital Image of the Real World – Digital Twins

Smart Industry sat down with Prith Banerjee, CTO of Ansys, a world leader in engineering simulation and product modeling, to discuss digital-twins and the future of manufacturing.

Digital twins, a kind of digital replica of physical machines and devices, are one of the hottest topics in the world of product development today, but few really understand what they are and what they can do.

All of our customers who are building large physical assets are excited about digital-twins. Before joining Ansys I worked as CTO at ABB and Schneider Electric. These companies have assets, like transformers, and they would like to have a virtual equivalent of that asset. Or, if you build airplanes, you would like to have a digital model of that airplane. If you make aircraft engines, you want to have a digital twin of that.

All our customers want to have digital simulation models of their assets but it is important to stress that digital twins are more than that. They are tied to the physical asset through an IoT connection which allows you to continuously collect actual data from the real asset 24/7, and to up-date your digital model so that it is a completely accurate image of what is happening in the real world.

Could you give us a few examples of digital twins at work?

We work, for example, with a large company in France which builds and operates nuclear power plants and huge power generators – multimillion-dollar assets. We help them build digital-twins of all their equipment. Or take the motor sports division of a major car manufacturer, who we have helped build electric vehicles using a six-step digital-twins model. A third example would be a large industrial manufacturer which uses digital-twins to design different types of compressors to help their salespeople show the value of their compressors to the customer.

Digital twins are tied to the physical asset through an iotconnection, which allows you to collect actual data 24/7.
_{Prith Banerjee, Chief Strategy Officer, Tata Communications}

At least in theory, digital twins could span the entire life cycle of a product from creation through operation to disposal, couldn’t they?

We’re already there. For instance, we are working closely with our many customers in the oil and gas industry to help them to design, build, and operate their assets, all the way to the end-of-life of those assets. When you design an asset, be it a pump or a valve, a transformer, a piece of equipment, or a data center, you usually use a CAD [computer-aided design] system, either a mechanical CAD or an electrical CAD. You need to analyze how good that asset is and that’s where digital twins help you in the design phase. To build the asset, supposing you are using 3D additive manufacturing [3D printing], a digital twin model comes in handy. During operations, using an IoT-connected digital twin means you can exactly monitor what is happening at any given moment. We at Ansys like to call this “pervasive simulation”, namely simulation based on digital twins that is being used in every phase of a product’s life cycle.

Where do I as a manufacturer start if I want to employ the digital twin concept?

A good starting point would be re-mote monitoring. Suppose you have an asset like a transformer or an engine, or what have you, you first connect it through an IoT platform and you monitor exactly what is happening. This is something called data-based analytics: you measure the vibration, the temperature, or the pressure and try to observe any anomalies that occur and, as soon as you see one, you know the asset is about to fail. This is how you do predictive maintenance but the accuracy of such a data-based analytics system is usually only about 60 percent at most. You may predict that this particular transformer will fail next Thursday but you will be wrong 40 percent of the time. If it is a million-dollar asset, you will make a $400,000 mistake. This is where the simulation-based digital twin comes in because, through time-based accurate physical simulation, you can increase that accuracy to 99 percent. That means if you predict that your million-dollar asset will fail, it really will!

Gartner says organizations will implement digital twins simply at first, then evolve them over time, improving their ability to collect and visualize the right data, apply the right analytics, and respond effectively to business objectives. Do you agree?

This is exactly what I mentioned. The first step in digital twinning is to connect up your assets and do very simple, data-based analytics. Over time, you tie that simple, data-based analytics to physics-based simulation, which gives you a much better and more accurate digital-twin-model. It’s a journey – you start simple and add more and more capabilities, for instance by tying in to your CAD or your simulation which, in the end, gives you a highly accurate digital twin model.

Of course, nobody’s perfect. How about digital twins – do they have limitations, too?

When you build a digital twin, you need a model that is accurate an fast. The trouble is that there is a trade-off. You can have a very accurate model but, if it takes 10,000 hours to come up with a model, that is of no use. We have a method we call the “reduced-order model” which is an approximation of a high-fidelity model. If you try to do it really fast, accuracy sometimes suffers. There are situations where the complexity of the multiphysical interactions – say, between the fluid flow, the structural flow, and the electromagnetic flow – may be so big that it becomes hard to get an accurate model. That would be a limitation.

How do you make sure your digital twin will fit your business needs to-day and tomorrow?

Great question! Today, most of our customers are talking about using digital-twins to improve their predictive capabilities in an operational setting. In the future, we will see more service value. Today, a compressor builder sells its customers a compressor unit which is maybe worth a couple of hundred thousand dollars. In the future – if they can guarantee they can provide compressed air wherever and whenever it’s needed, and if their compressors are always up – they can provide the compressor as an asset to their customers and monetize it through “compressed air as a service”. Air as a service – why not energy as a service, gas as a service? Digital twins will make such transitions possible.

READ MORE ARTICLES