How AI accelerates early product design – Transcript
In this episode of AI Spectrum, guests Boris Scharinger, Senior Innovation Manager and Technology Strategist for Siemens Digital Industries, and Dr. Justin Hodges, Senior AI/ML Technical Specialist in Product Management for Siemens Digital Industries Software came together to discuss the applications of AI in the early stages of the design process, and the exciting possibilities for the future that that holds.
Check out the full episode here or keep reading for a transcript of that talk.
Spencer Acain:
Hello and welcome to the AI Spectrum podcast. I’m your host, Spencer Acain. In this series, we explore a wide range of AI topics from all across Siemens and how they’re applied to different technologies. Today, I’m joined by Boris Scharinger and Justin Hodges. Welcome, both of you.
Boris Scharinger:
Hi, great to be here. Thanks for having.
Justin Hodges:
Likewise. Love speaking to you, Spencer and Boris, this will be a very stimulating conversation.
Spencer Acain:
Great. So why don’t we kick things off by having a brief self-introduction from both of you, so our audiences know a bit about your background and your work at Siemens. Boris, do you want to kick us off with this?
Boris Scharinger:
Yes, thank you. Well, I kind of started my career at the intersection of business administration and computer science studying at university. Maybe information management would be a good term to use for that type of studies. I worked a couple of years for a consulting company, then joined my customer at those times, Siemens, in 2004, and worked in various roles basically in IT management. And I moved on to the portfolio side of life in 2018 when I joined Siemens Digital Industries CTO office as a technology and startup scout. And my favorite topic in that job is industrial AI.
Spencer Acain:
Wow, it sounds like you’ve had quite the career up to now. So Justin, do you want to give us a little bit of your background now?
Justin Hodges:
Yeah, gladly. It’ll pale in comparison so prepare to be underwhelmed, but I did my bachelor’s, master’s and PhD in what you could consider a traditional simulation turbo machinery, aerospace focus, BFD, heat transfer, turbulence, that sort of thing. And then I took an internship in Princeton in 2017 that was combining AI to exactly that sort of numerical simulation work. And from that point forward, I realized that my career vector would be the overlap of AI to the simulation world, augmenting and supplementing what we can do rather than just taking things over with AI. And so I’ve been so passionate about it from then until now that that is my full-time role essentially for simulation products for AI as a technical specialist.
And I absolutely love talking about this subject and working with renowned experts that are just so brilliant in their core disciplines and then introducing AI and how it can further be leveraged as a tool in our toolbox. So much so that I wrote a book earlier this year on this exact topic, so it’s a very lovely time for me to see the CAE and simulation world transform, adapting more and more of these AI ML-based things. So that’s me.
Spencer Acain:
Great. It sounds like we’re in for a exciting conversation today. All right, so then to kick things off, I think we should start by exploring AI across all the stages of manufacturing since it’s really kind of come up across the entire breadth of the industry over the last few years, five years, 10 years even. And there are just a ton of applications of it now. So I’d like to start off with you, Justin, maybe you could start us on our AI and manufacturing and simulation design journey by giving us a look at the AI in the early stages of the design process, maybe design simulation, that kind of stuff.
Justin Hodges:
Yeah. And to try to have a little limbic soft introduction to warm up, I think history repeats itself in a lot of ways, and this might resemble once we get into the conversation how it looked when physical measurement and physical test was by and large the mainstream and primary modality by which engineers gathered conclusions and insights on their design to guide their design process in which time simulation was the new kid on the block and the new emerging method. And what you saw after years of adoption and progression is that then it would make things quite a lot faster. There is this, in the public example I always think about from Airbus when they’re doing their ground testing, how in the matter of 15 years or so, that process went from a month to three days or something.
And basically simulation has helped in a lot of ways save time, learn things more quickly about their design space and really help make it crucial to get there quickly. And I kind of view it as a paradigm between a metaphor very much in a similar vein to AI and simulation. You’d never get on an airplane that hasn’t been physically tested of course, but there’s a lot of ways that simulation could have helped make it go faster and arrive at design conclusions faster. And in a similar vein, simulation is king, but AI is very, very useful and in the right capacity can be used to get to an insight much faster and much cheaper in the sense of hardware and compute time and such. And so it’s not to replace simulation, but in general to set the stage I think the primary way is to be used early in the design process, like you said, to divulge a ton of information about the design space, the behavior, the performance, and then from there use more traditional methods.
But I would love to turn it to the other speaker here, to Boris to give your opinion because I think it’s a nice analogy and maybe with your experience you could compare this advent of AI to other sort of emerging technologies that have taken place and thus we could look at the technology adoption cycles and see if there’s similarities.
Boris Scharinger:
Yeah, thanks, Justin. The thought that immediately came to my mind is when you explained the role of AI in simulation, that we also as a company, Siemens as a company, we really look at the mega trends and we look at the sustainability challenge of mankind. And we believe that we could really make a difference for the resource consumption footprint of industry by not wasting physical resources but before we need to, so we should be simulating as much as we can because ultimately in 10 to 15 years from now, we all should be producing whatever gets produced today for mankind with say, 10 or 20% of the resource input that we are using today. And how can we do this?
We can only do this by optimizing anything that we can optimize and to do this optimization process everywhere, whether it’s material or the way how products are designed and the way how products are produced. All of that can benefit from simulation/optimization. And if we would try to do this, all of this new stuff like multiply a million, sorry, simulate a million times more stuff than we are simulating today, we couldn’t do this on a numerical base. We couldn’t do this with simulation without AI. We need AI to accelerate that process of simulation. So those are my two cents on the significance of simulation and AI as a dream team-
Justin Hodges:
Yeah. When you said multiply the number of simulations to some much bigger number, and then you look at that challenge and say, how’s that possible, then you can’t adapt a method that improves the simulation by 10% or 20%. You’re talking in orders of magnitude. And so you do need to look for more game-changing sort of technologies. It makes me think of this thing I read in an article. I’m not a marine person, but I read this article that was out there saying that there are more strict regulations in place for the next 20 years or something that’ll take place in the next 20 years for marine where now these companies that produce these mega ships and they design the hulls and the full vessel, they need to now do much, much stricter performance for emissions and much, much, much less emissions that are now allowed.
And they have to do many more simulations in order to quantify the performance under a wide variety of scenarios. I mean the sea condition, the flight, or not the flight, the vessel speed condition and all these things. For every design you can imagine that this is a very transient process. It’s very, very expensive. That’s what it reminded me when you said that. You’re looking at this problem where you have to simulate many, many, many more things and like you said, in a more efficient, resourceful way, frugal way. And I think that is the dream team if you can supplement with this data-driven type group of approaches, and then that may help. You have to leverage. Anyway, sorry, Spencer, I think I’m already on a bit of a tangent to answer your question, but I definitely wanted Boris’ input, but like you said, how is design and simulation in early design phase really making an impact?
I think probably the use case most people think of is faster decision making. I think most people by now have seen this idea that I can make an interactive change to my design and get real-time feedback from a machine learning model because we know that simulations or other techniques just cannot give you an answer in fractions of a second. And that is really good for the early design process because it shifts this information left. These models learn from simulations and historical data that companies have, which they gather throughout the life of the product and the full design process, not just the beginning, but as other disciplines take up and do design work. And when you consolidate that into a model, it’s very good to have that information upfront. And like I said, these models are very quick, so you can get real-time sort of feedback on how performances would be for different designs. And I would say, to answer your question, that’s probably the main pathway I see to how AI will help in the early design phase.
Boris Scharinger:
And it’s the agility that is attractive to companies. But then I’d like to add this because we see so much criticism of AI being so resource intensive itself consuming so much energy and the data centers and whatnot. But just imagine how much resources simulation and AI enabled and accelerated simulation saves in terms of doing stuff virtually and not building a prototype or 10 prototypes until you can come to that conclusion. That’s both. It’s a speed meta, but it’s also a resource consumption meta that can both be solved at the same time when you really drive simulation with the help of AI into design processes.
Justin Hodges:
Yeah, I think it’s a good point. It may seem like there is some nuance, but we want the data requirement to be low. So we are going to keep making in the scientific ML community better and better machine learning models that require less data, that generalize better, that train faster, that are not huge and gigantic. So the footprint of that trend is naturally going down and down in the future. I think there’s some AI headlines in the last year or even more where people are… Innovators in the field are saying the era of giant machine learning models is over, and now it is shifting to small models that are equally as performant and the need is to drive down to smaller and smaller models. So in terms of being resourceful and frugal, I think we’ll keep seeing that trend as well. It’s not very attractive to say you need 1,000 simulations in order to train this model. It’s much more attractive to say you need 50. So I think the need for the market will also help compel the models to be smaller and smaller footprint.
Spencer Acain:
The idea that smaller models might actually be the way that the market goes in the future is certainly an interesting one. And I think it’s a great place to end our discussion for today. So once again, I have been your host, Spencer Acain, joined by Boris Scharinger and Justin Hodges. Tune in again next time as we continue to explore the exciting world of AI.
Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.
