At the upcoming Hannover Messe in April, Altair will focus on a theme of “The Science of Possibility,” highlighting the company’s technologies for simulation, high performance computing (HPC), data analytics, and artificial intelligence (AI). Areas of interest will include:
- Simulation-driven design
- AI-powered engineering
- Digital twins, and
- AI on the factory floor
We spoke to Mirko Bromberger, Altair’s director of marketing, about the company’s products and asked his thoughts on a couple of industry trends.
“Altair has been around for almost four decades in the fields of virtual product development, simulation, and providing simulation tools to the industry,” he said. “Over the last decade, we have been investing heavily in data analytics and artificial intelligence solutions — our own developments and a couple of acquisitions — acquiring the technology and integrating it into the complete portfolio. We’re focused on high-performance computing to make the computing infrastructure more effective because simulation, as well as data analytics, are compute-intensive things to do. We’re the only company that really brings together these three pillars into comprehensive and slick solutions across all industries.”
Hannover Messe is industry agnostic, and so is Altair, according to Bromberger. He said that the company’s focus is really on enabling technologies that are applicable in a variety of industries.
“We provide solutions for artificial intelligence data analytics and work on the digital thread with the help of our computing infrastructure. We’ve helped build digital twins and make them more efficient and more quickly, and we implement AI capabilities into engineering processes. We keep on solving the key bottleneck that’s in the way of simulation-driven design,” he said.
Bromberger noted that while commercial FEA applications have been around for five decades, the model building process is still a hurdle to overcome, and it hinders simulation-driven design.
“It’s kind of crazy that, in general, in the development process, you invest a lot of time into a model building process. You have a detailed CAD design, either mechanical or electrical, and you put a lot of details into that. Then, you simplify it to make a simulation to draw conclusions if you’re going in the right direction. Here, we really take care to reduce the time to build the models and accelerate and automate model building processes. Because when you do an automation, you have a script, and the script has a version that gets old. So, what do you do? It’s not sustainable. Investing in a platform that is the backbone for automation processes, and inventing or investing in key technologies that overcome the need for model building processes, takes away the burden. A couple of years ago, we acquired a technology called SimSolid that takes away that hurdle completely. But it’s still a bit of an applicational training process to get through that hurdle and out of people’s mindsets.”
Q: Why is Altair integrating AI into design and simulation? Are there certain problems machine learning is better suited to solving?
A: I think AI is an umbrella term for a whole bunch of statistical tools. We use appropriate tools in the right environment to identify a certain behavior. A clustering algorithm based on unsupervised learning is well-suited to put the results into a certain number of buckets. And it’s doing that way better than any human. Then, a human can look at it, select the bucket with the right behavior, and connect that clustering element as a response in an optimization — or accelerating digital twins or system simulations.
You feed a bunch of simulation results into a neural network, and you can use it as a block in your system simulation, which is real-time capable. So, you replace a comprehensive co-simulation with a neural network. For example, if you think of a fluid thermal simulation that takes a long time, one part of the model always has to wait on the other loop. With that application, what we call ROM (reduced order model) AI, we have the simulation down by a factor of one thousand times quicker.
Q: How do you see AI in the next few years integrating into an engineering workflow? Where do you think it’s going next?
A: I think many people would be amazed at how much AI is used already. It’s already there. For the ones who missed it, they better start using it! Understanding which AI or machine learning element is the right one for certain applications is basic knowledge now and should almost be commonsense. But we’re still in the phase where people don’t understand the difference between a huge language model behind ChatGPT and a simple small neural network that we’re feeding data for a dedicated purpose to capture a certain behavior. For us technology providers, we make the technology accessible. We try it out, do demonstrations, and collaborate with clients. And we’re pointing at this as a good way for engineers to proceed.
That whole topic of reduced order modeling and system identification is important. It’s a mouthful of words for accelerating digital twins or replacing co-simulation with AI to speed up system simulations. People need to know that’s an option.
Filed Under: AI Engineering Collective, NEWS • PROFILES • EDITORIALS, AI • machine learning, DIGITAL TRANSFORMATION (DX)
