Digital twin evolution: The path to simulation-based twins

Imagine if you will, you are a manufacturer that deploys remote assets. For example, an aircraft manufacturer or a satellite company. You just launched an IoT-enabled satellite to monitor and gather weather data. The data from your satellite sensors starts coming in, but it is not quite what you expected.

You need to change your approach, logic or parameters, but it is too late; the satellite is already in space.

You have done basic design simulation models, but this is only telling you what may happen in the specific physical scenario that was simulated. It may not tell you how the design will behave in the real-world, when virtual and physical environments interact with multiple physical dimensions like material strength, fluid flow or electromagnetic fields collide. This might warrant a need for real-time redesign or configuration of how the remote systems work.

You might have IoT sensors in place to gather telemetry about the asset. However, before that first IoT signal comes in, you must map out the influential factors. These could include things like external environmental factors, load conditions etc., around the devices. You need to be able to understand the causality and how these factors interact with these devices and behave in the real world where those devices live and operate.

Very often manufacturers need this type of validation before they even build the very first product. They also need an ability to monitor the behavior and performance of their assets on a serialized basis and then make changes to them in real time as the asset continues to be used. Think about how Tesla updated the braking performance of their cars with a software update.

Do you like the concept but find it difficult or impossible to implement it in your world?

The answer could be simulation-based digital twins. It is the next big thing. It is what the manufacturing of tomorrow will look like with the multi-physics-based simulation capabilities that are being built today. Although analytical twins have existed for a while, this evolution of simulation-based models that lead to the creation of digital twins is really transforming manufacturing, as product designs become more customized, unique and intricate by nature.

In this blog series, I will look at where we are today with digital twin capabilities and how we will get to that future state.

A growing digital twins landscape

When we take a step back, we know that digital twins have evolved rapidly over the last decade. Back in the 1980s, having a CAD model of a paper design was considered a digital twin of that paper design. As computing power increased, engineers tried to apply simulation techniques to interact with the digital designs. For many years, the creation of a digital twin has been centered solely on design product twins, most notably around optimizing the lifecycle of a specific piece of equipment. Then, we moved to process twins, which allowed us to work with multiple digital twin models. This helps us, for example, optimize the entire manufacturing process—ensuring that multiple assets and production processes are all working together. A capability to interact with these twin models also began to evolve using mixed reality and tools like Microsoft HoloLens. Companies including Rolls-Royce, Schneider Electric, Tetra Pak and Unilever are early leaders of this innovation.

Today, the potential for a digital representation via a twin has evolved even more dramatically. Beyond products and processes, we now have many different types of twins.

Let us look at some examples: 

Supply chain digital twins go across organizational boundaries across the value chain to enable collaborative scenarios with multiple suppliers and organizations. One of the best examples is Microsoft’s own supply chain for hardware devices. Microsoft runs one of the most intelligent supply chains in the industry. We have developed a track-and-trace capability that allows us to stay connected to every shipping container in transit anywhere in the world, in any mode of shipment, reaching every warehouse and every customer location. This digital twin of the in-transit material across the supply chain provides unprecedented visibility and traceability of our products around the world, enables on-time deliveries and unleashes hundreds of millions of dollars in working capital and savings.

You then have IoT and Analytical modeling scenarios. With IoT twins, you have the telemetry, knowledge and operational feedback from sensors on a real asset. You can do analytics against it to get insights into things like asset usage, availability and performance. However, the challenge with these types of twins is that they tend to provide insights from telemetry that actually takes place within the asset. Therefore, by nature, the insights they provide are after-the-fact and reactive. Although these insights could be useful from an analytics or design and engineering standpoint, they could pose a challenge for certain real-time monitoring use cases, such as flight or engine monitoring in aerospace, where failure is not an option. This is the reason there is a tremendous interest in the industry around predictive analytics using statistical techniques and leveraging historical time series data from these sensors to predict the future health of the asset.

Digital thread traceability across the lifecycle

Another class of digital twins, typically referred to as configuration twins, manage the asset bill of material (BOM) through its entire lifecycle—from ideation and design, to manufacturing, all the way to sales, service and even disposal. Typically, the CAD design models provide the “As Designed” bill of materials. Once manufactured, the “As Manufactured” BOM information typically needs to be serialized for that particular asset. Eventually, as the asset is configured and sold to the end customer and maintained, there could be other versions of this BOM commonly referred to “As Sold” and “As Maintained.” These BOMs need to include not only the mechanical components on the asset, but also the electrical and electronic parts as well as other hardware, software and firmware information for version control.

To add to the complexity, over the lifecycle of this asset, data generated from it might be owned by multiple organizations, such as the OEM, operator, customer, etc. For a complete traceability across the lifecycle of this asset, it is important to have this data housed in a digital model and framework such that we can track down every single component and firmware to the finest level of detail, tracing the genesis and evolution in the product’s lifecycle.

This is where a digital thread comes in, enabling full lifecycle traceability that can track a product and its digital assets all the way down to the serial number through its various stages and evolution. This is especially important for certain industries, for example automotive, where assets are tracked by VIN number, and aerospace, where they are serialized by the aircraft tail number. Because every asset can now be tracked at this very detailed level, manufacturers can manage things like asset configuration, software and firmware version control, over-the-air software updates, and much more.

Our good partner Aras offers the ability to manage the digital twin configuration with this level of information across the lifecycle of the asset, thereby enabling full digital thread traceability. “The ability to interpret and act upon the data streaming back from the field and factories often requires traceability to prior information from related revisions, what’s called the Digital Thread,” notes Marc Lind, SVP of Strategy at Aras. “The accuracy of Digital Thread traceability from the Digital Twin configuration back through the lifecycle is a necessary foundation for many IoT-enabled product scenarios that involve Machine Learning.”

Azure Digital Twins: The next wave of IoT innovation

At Microsoft, we are committed to supporting these digital twins by leveraging the dimensionality of the operational data that can be captured and analyzed through digital twins. We do this through the Azure Digital Twins platform. This exciting breakthrough allows manufacturers to track, optimize, simulate and predict the future by creating a comprehensive digital model of any physical environment, including the people, and places, things, and the relationships that bind them.

The result? Improved operations, efficiencies, productivity, and the list goes on.

Microsoft also supports the OPC Twin framework which consists of microservices that use Azure IoT Edge and IoT Hub to connect the cloud and the factory network. OPC Twin provides discovery, registration, and remote control of industrial devices through REST APIs. You can try it in our Connected Factory Solution Accelerator. Microsoft’s Erich Barnstedt has a video where you can learn more.

Market-maker thyssenkrupp Elevator is partnering with Willow to build a digitalized virtual model of their Innovation Test Tower in Rottweil, Germany to revolutionize the way buildings are maintained and to enhance the experience of tenants and visitors. Willow Twin is built on Microsoft Azure, using a wide range of Azure services, including the Azure Digital Twins platform.

Six-time Microsoft Partner of the Year award winner ICONICS is doing impressive work with the Azure Digital Twins platform. They have built a new mixed reality layer, extending Azure Digital Twins to marry mixed reality and human interaction within physical environments.

The future of digital twins is simulation-based

But what is most exciting on the horizon for digital twin innovation is the newest class of twins that leverage multi-physics-based simulation. These allow you to have a twin of an asset before manufacturing them so you can experiment with various multi-physics models. You can potentially integrate these models using a common taxonomy and simulate over a High- Performance Computing platform like Microsoft Azure HPC. This will allow you to generate answers for various what-if scenarios. You can then test and assess asset performance prior to physical prototyping in very specific situations under very specific circumstances before the very first launch of the product, such as an aircraft or a machine. The need for physical prototyping, associated costs of testing (including destructive techniques like crash testing), and the time required from concept to launch of a product, are greatly reduced by leveraging these capabilities.

It is groundbreaking!

ANSYS is an industry leader in the world of multi-physics simulation and they are doing significant work in this area. We are pleased to collaborate with them to extend digital twin capabilities with the ANSYS Twin Builder tool, leveraging the performance, scalability and security of the Azure platform.

Microsoft’s Tony Shakib just wrote a comprehensive blog summarizing how simulation increases digital twins’ effectiveness and sharing more depth around our work with ANSYS.

In our next blog, we will delve more into how simulation-based twins can empower manufacturers with new insights to optimize asset production and operations, increase revenue, and outpace the competition even faster using the power of HPC.