Last month ANSYS announced the release of its ANSYS 19.1 multiphysics simulation suite along with a new product named Twin Builder.
We touched on this release briefly in this article. To jog your memory, Twin Builder is a new product that builds on previous ANSYS products and consolidates them into a convenient workflow exclusively dedicated to helping engineers design digital twins.
A digital twin, in case you missed it, is a 3D virtual representation of a component, assembly or system that is connected to its real-world counterpart by means of sensors, networks and Internet of Things (IoT) devices. The digital twin behaves just as the real-life twin behaves, and can be used for diagnosis, prognosis, and what-if scenarios of its real-world variant. The data that the digital twin provides can be used for process optimization and maintenance scheduling, and can even be used to increase profitability by reducing downtime. You can see a summary of the benefits of digital twins in the video that follows.
Engineering.com spoke to Sameer Kher, director, Systems Product Line at ANSYS, to get the details on what Twin Builder is, what it does, and how customers are using it for their digital twin needs.
A Consolidated Workflow
“Twin Builder is built on existing technology such as Simplorer, which is an existing ANSYS product,” said Kher. “It’s also based on SCADE. And also, we’ve invested in reduced order modeling (ROM), which is something we’ve been involved with for a few years, so Twin Builder is an amalgamation of work that we have done in the past.
“Reduced order modeling typically takes data from 3D (FEA/CFD) simulation and converts it into a faster simulation,” explained Kher. “[ROM] techniques vary—from matrix reduction, equivalent model extraction, regression, polynomial fitting, transfer function/system identification, ML/DL, etc. We also have techniques that can render the field back from the reduced model.
“One extra thing that we did add [to Twin Builder], which has taken a couple of years, was to add connectivity to some popular IoT platforms. That involved phasing out some specific APIs, and actually making the connections with the new IoT platforms, so that’s sort of the new piece,” continued Kher.
How to Build a Digital Twin
So that’s how Twin Builder came into existence. It was based on previous technologies from the ANSYS stable. But how did engineers build digital twins before these technologies were amalgamated? Was there a proper and accepted method for creating them?
“I think there wasn’t really a [clearly defined] method of building a digital twin in the past,” said Kher. “What people would do in the past would be to just build a virtual prototype. There was no real notion of connecting it to operational data. And there was no real notion of actually deploying the digital twin at scale in an operational setting. So, if you think about it, simulation is typically used at the design phase. You would build one virtual prototype for your design, validate it, and that’s it. But now what we want to do is to take that design and replicate it as you would with other virtual assets. That concept of scaling out is what’s really new here.”
Kher then went on to describe the steps involved in creating a digital twin.
“We start by telling customers to look at their top service costs and warranty costs that they may have; they can then do some kind of failure mode analysis to determine exactly what kind of digital twin model they need to build,” he continued.
“We can then look at existing simulation models that were used during the design of the system, subsystem or component that was identified as key to the failure mode. From these simulation artefacts, we can extract—via techniques like reduced order modeling, behavioural modeling, etc.—an accurate virtual replica of the physical equipment.
“Then you validate it by tuning and optimizing parameters to accurately match measured data.
“Finally, we can export out the model into an executable deployable runtime. Additionally, we have connectors for several popular IIoT platforms, allowing customers to actually connect the simulation model to data through the IoT platform and deploy the digital twin at scale,” said Kher.
Use Cases in Industry
So, who are the biggest users of digital twins at the moment?
“At the moment, the digital twins are sort of limited in deployment and limited in scale, but generally the most common uses of digital twins at the moment are for industrial equipment such as motors and pumps, and heating and cooling systems…those types of applications,” continued Kher.
“Also, we are seeing a lot of use in turbines…wind turbines…energy and oil and gas—those types of industries are where we are seeing the most activity in terms of deployments. There are other industries where we are seeing early-stage conversations going on, such as automotive and aerospace, but we are seeing industry and oil and gas as being the early adopters.”
We have heard the word “scalability” used when describing the benefits of digital twins. This can refer to the scalability of the number of instances of digital twins (see the GE case study at the end of the article to see how many digital twins GE has), or it can be used to describe the number of sensors and the scale of the analysis that a company wishes to perform.
So that begs the question: How big can a digital twin be in terms of scope and complexity?
“The scale of deployment will vary a lot. For example, in one application, we have a motor, and it has a few sensors on it. The number of sensors is very limited in that case—we are just using them to determine useful life and things like that. Then there are more complicated systems such as submersible pumps that contain multiple subsystems and multiple sensors. These are complicated systems and can contain tens to hundreds of different sensors,” explained Kher.
And those sensors aren’t restricted to physical sensors. Virtual sensors play an important role, too, especially when it comes to generating analytics and graphs outside of the typical functional data that is representative of nominal physical system operations. Yep, in short, a digital twin can be as big or as complicated as you want it to be. But then it really boils down to a question of what information do you need? What is useful and actionable?
Trials
Naturally, being the well-connected company that ANSYS is, it ensured that Twin Builder was fit for public consumption before releasing it along with ANSYS 19.1.
“We’ve done several trials,” said Kher. “Because this is based on existing technology, we have been testing at low scale in a few different industries with some of our advanced customers. There are about three of them who have gone public with their projects, and I can send you some links to those stories after we finish here.”
And you can indeed see those three use cases by clicking the following links.
The first use case shows how GE is using digital twins for predictive maintenance of wind turbines in France. The article mentions that GE currently has over 1.2 million digital twins at use in the field, which highlights Kher’s point of scalability. You can see how Twin Builder was used to build the GE wind turbine twin in the video that follows.
And the final case study shows how flow-control system manufacturer Flowserve is using digital twin technology to improve reliability in the oil and gas and processes sectors.
Thanks to Sameer Kher of ANSYS for taking time to chat with us about Twin Builder.
Twin Builder is out now as part of the ANSYS 19.1 update. You can read more about Twin Builder here.