Ansys Digital Twin’s Post

Transcript

- What is Ansys TwinAI and what are the key capabilities of this AI-driven solution? I'm Asmaa from Ansys and you're watching "The Twin Talks." (energetic music) To answer these questions, I'm joined today by Dr. Prith Banerjee, CTO at Ansys, and Sameer Kher, Senior Director R&D for Digital Twins at Ansys. Prith and Sameer already have been one of our special guests for "The Twin Talks" and discussed the value and real definition of digital twins and how this technology is revolutionizing the future of all industries thanks to AI and ML. First, thank you, Prith and Sameer, for being part of this special episode. - Welcome, looking forward to it. - Prith, we have exciting news to share with our audience and we're delighted to have you announce it. What is Ansys TwinAI? - Absolutely, so in the last five years, we at Ansys have been working on this amazing technology of AI/ML applied to simulation, and we have made a lot of progress. As I talk to customers, they are excited about it. So what Ansys TwinAI is, essentially we have been working on the technology of Hybrid Digital Twins that I mentioned in my previous podcast episode, where we are combining data-analytics-based digital twinning with simulation-based digital twinning. So, and this is something that in my previous experience as CTO at ABB and CTO of Schneider Electric, I used to see that all the time. Well, in those companies, and in fact, most companies use pure data-analytics-based digital twinning. The disadvantage of that is it only gives you, it is easy to use, but the accuracy is not there. By combining the two, we are able to get super accurate predicted analytics capabilities, which is what Hybrid Digital Twin is. And so Twin-AI represents the realization of that vision, a modern product that allows customers to build Hybrid Digital Twins. - Sameer, what is the core value proposition of Ansys Twin-AI? - Sure, thanks, Asmaa. So first I just want to say that we're super excited to announce the launch of Ansys TwinAI as a part of our Digital Twin product line. And I'm also really excited to be able to sort of have this tech talk episode alongside Prith. Prith's guidance has been really important, setting the vision for our Digital Twin product line. So back to TwinAI, TwinAI really seamlessly combines insights from the real world data, it allows you to bring real world data insights and then apply them in the context of the simulation models, as Prith was explaining. And we do this by applying, you know, AI/ML techniques, a variety of AI/ML techniques, so specifically the techniques that we have in TwinAI include things like calibration or Bayesian calibration that lets you use sensor information measurement data to create these fast, accurate, evolving digital twins. We also have Reduced Order Modeling capabilities that can be used to create models quickly either from simulation data and simulation results or measurement data. And then finally, TwinAI provides a really easy path to scale deployment via methods that let you automatically create apps, Python apps or web apps, as well as automatically deployed containers that are driven by REST APIs. And then finally, I'll just say TwinAI is really complementary to Twin Builder, completes out our Digital Twin product line, but it can also be used directly with our flagships to enhance simulation, any simulation with data flagships, like Fluent Or Ansys Mechanical, all of those could benefit from the use of TwinAI. - Prith, what is the impact and benefits for Ansys users of combining SimAI and TwinAI? - So this is what I got started with, right? So in the last five years, we have been looking at the use of AI technology across our simulation products. And so our customers are using the different multiphysics simulation like Mechanical Or Fluent or HFSS for electromagnetics and so on. But what SimAI has done is to take those solvers and in the design phase of a product, trying to accelerate the simulation to make it run a hundred times faster. So essentially you are using the simulation core simulation solvers from Ansys to do the product design, and now SimAI just accelerates it much farther. So it is complementary with TwinAI. TwinAI is more focused on the validation, verification of the actual part in operations. So these two are complementary technologies. SimAI is for the product development and TwinAI is for the product operation and validation. And the two tech technologies actually are very complimentary. They're leveraging a lot of shared ideas, like the concept of, when I talk about AI/ML, I always talk about top-down methods and bottom-down methods and reduced order models, and the same Reduced Order Models, the dynamic realms are used in TwinAI as well. So these are complementary technologies, but SimAI is more for the product design part and TwinAI for the product operations and the validation part. So I'm super excited to actually talk about both these product launches together. - Sameer, can you dive into the technical aspects of Ansys TwinAI? How do AI and ML enhance the capabilities of digital twins? - Certainly, and this is evolving and, you know, increasing sort of set of capabilities in terms of a roadmap. What we know is when we create models typically using simulation tools, you know, Twin Builder or any of the simulation tools for that matter, the parameters and characteristics that you use to define the model tend to represent some static point in time. And these often relate to the initial manufacturing values of specs. But in reality, when you're looking at digital twins and operations and verification, you know, the equipment changes, deteriorates and ages, you may not have captured all of the physics, certain parts may have been swapped in and out, things like that. And all of these can be modeled and predict more accurately through the use of AI/ML techniques. So in TwinAI, we've basically built out a framework and capabilities that let you do that. We have workflows for things like Bayesian calibration, optimization, gaussian process modeling or fusion modeling. And all of these help address and build towards the requirements I just outlined. - Prith, in your role, you talk to customers all the time, how do you see customers reacting to this announcement of TwinAI? - Absolutely, so as I mentioned during my previous talk, Ansys has lots and lots of customers. We have thousands of customers, and I have the privilege of talking to literally hundreds of customers every year. And we have customers in sort of five broad verticals, I talk to customers in the automotive space, I talk to customers in aerospace and defense, talk to customers in energy, in industrials, in high-tech and semiconductor and healthcare. And each of these five verticals, all these customers are excited. They're actually talking about using AI technologies for simulation, the SimAI, as well as the digital twin technologies, right? So if you're in automotive setting, you want to build a digital twin of a car, if you're in aerospace, you want to build a digital twin of an airplane or an engine. If you are in the energy space, you want to build a digital twin of a wind farm, right? Or if you're in healthcare area, try to build a digital twin of a human heart or a pacemaker. So as I talk to customers, every one of these fields are excited about the technology of digital twin, the technology of AI, and therefore TwinAI. - Sameer, in the first episode of this podcast series, you shared some insights about Tata Steel. Can you tell us more about their usage of TwinAI? - Yeah, absolutely, so as we've rolled out our capabilities, our digital twin capabilities, we've had the privilege and opportunity to really work alongside some of the best customers in the industry. And several of them have been willing to share their stories for the benefit of other customers. So one of them is Tata Steel. As a quick reminder from that previous podcast, you know, Tata Steel is a global steel manufacturer, you know, and one of the key challenges that they have in the manufacturing process is around the transportation of hot metal within the manufacturing facility. Higher the temperature that the metal is transported at, the better the yield and lower the carbon dioxide emissions. But unfortunately, on the flip side, you know, the higher temperature also causes much higher wear of the equipment that's used to transport as well as requires energy usage. And so, you know, the digital twin that we've built with them that they're using is really a very detailed model of the transportation equipment, the material and the lining, that really lets Tata Steel understand and optimize for where, you know, based on the current state of the operating facility. And you know, ultimately they've really been able to use our digital twin technology and deploy it in a few facilities and now they're expanding it to other regions and this helps them reduce maintenance costs, downtime, while still maintaining sustainability targets in terms of energy utilization and carbon dioxide emissions. So it's a really exciting use case. You know, they use TwinAI specifically to really do the data portion, calibrating the models that they create using measurement data so that it's up to date and accurate. And then also to use, you know, to create deployment, deployable apps out of the digital twin models that we've created. So they've been an early user of TwinAI, have had a lot of success with it and we're really excited to be continuing to work with them. There are also other customers and other use cases out there. We put those out on our LinkedIn channel and you know, I would definitely recommend anyone interested to continue to follow that channel. - And Sameer, what new features or capabilities can users expect from Ansys TwinAI in the near future? - Yeah, so this is just the beginning. We are really excited about the future roadmap of Twin AI. You know, coming up immediately next in the next release or two is a wizard based workflow for our Reduced Order Modeling techniques. We've had Reduced Order Modeling techniques available for a while, they're in a fairly sophisticated state already, but now we are really making it much more intuitive and much more interactive for the user. Just look at the data or the type of system they're trying to build the model for, and TwinAI will be able to handle the extraction of a appropriate accurate drum with uncertainty information and other sort of relevant information without needing too much user intervention. And so, that's what we're working on right now. This is especially useful when you think about simulation, you know, people who are not simulation experts and it'll really help, we believe lot more other people, you know, non-simulation experts embrace and adopt and benefit from the whole digital twin capability. We also have plans for additional algorithms for calibration optimization, data modeling and so on. And of course a lot of this is always driven with customer feedback in mind. And as we continue to in innovate, we talk to our customers and, you know, continue to sort of develop and grow the product. So super excited about what's next and we'll certainly, you know, keep posting on our channels. - Prith, how do you envision the future of digital twins? - So actually I would like to take a sort of follow up on what Sameer mentioned about the customer example of Tata Steel. And that is, it is in the domain of steel manufacturing, but we are seeing opportunities for manufacturing in say, semiconductor chips, right? I mean, the US government is investing so much money in bringing chip manufacturing back to the US, right? With these large plants for TSMC and Intel in Arizona and other places and IBM building plants in New York. As those new semiconductor plants are being built, right? A digital twin of such a fabrication plant can really have huge implications in terms of the yields of the chips coming out of those things. And you have in these semiconductor facilities, you have equipment, the semiconductor equipment companies, companies from say, Applied Materials or Lam Research or ASML, right? These are very, very complicated equipment. And if you could model all of those physics in the equipment, you could come up with digital twin models of the equipment, the digital twin models of the processes, like epitaxial growth and digital twin models of the entire fab, right? And if there is a failure, you can very quickly do "what if" analysis, right? The diagnosis of what is going on. So I see the potential for digital twins, very, very significant in the area of manufacturing alone and specifically for semiconductor manufacturing in the world. - Thank you so much, Prith and Sameer, for being part of this special edition of "The Twin Talks," discussing the launch of Ansys TwinAI. - Thank you for inviting us. - Thank you. (energetic music)