This video was sponsored by Altair.
Computer-aided design was developed for a simple reason: to make the design engineering process more efficient. CAD was so successful that it rapidly became a global must have, but in the process, it also changed the way all design engineering is done. By freeing the designer from the drudgery of draftsmanship, it became possible to find more creative solutions to problems of increasing complexity.
Today, highly sophisticated algorithms allow design engineers to virtually test concepts and ideas, then iterate rapidly, something impossible with physical prototypes. One consequence of simulation and the rise of AI is that design engineering itself may change in ways more significant than it did when computers replaced the T-square. Where is engineering going? Brett Chouinard, Chief Technology Officer at Altair, discusses this issue, and more, with engineering.com’s Jim Anderton.
The transcript below has been edited for clarity.
Jim Anderton: For over seven decades, computers have been instrumental in advancing the art of leading-edge engineering. Now, the era of desktop computing, well, that brought us computer aided design, which digitized rendering and made the drafting table obsolete. But the key attribute of CAD was aided. The heavy lifting, still dependent on that supercomputer between the engineer's ears. Now that's changing today with a new generation of advanced tools that allow creative minds to simulate and test designs in real time, combining better initial design with very fast iteration, a powerful combination for better solutions. The latest technology puts all this on steroids with advances in artificial intelligence, augmented and virtual reality, and model based systems engineering. So where is engineering going? Well, to answer part of that question, I'm joined today by Brett Chouinard, chief technology officer at Altair. Brett it's a 35 year industry veteran and holds MS and BS degrees in mechanical engineering from the University of Cincinnati and the Michigan Technological University. Brett, welcome to the show.
Brett Chouinard: Thank you for inviting me.
Jim Anderton: Brett, CAD was originally designed to do what engineers and draftsmen did with pencil on paper only digitally. Then it moved to digital renderings that could be manipulated and visualized in three dimensions to help designers understand their parts better. You know, the high level mathematics, the code that drove this system, that was secondary to the primary function, which was to leverage the designer's imagination. Then we had tools like CFD and today's simulation. Do you think that engineers think differently today than they did when they did the calculations offline, then press to paper to get as close to a final design as possible?
Brett Chouinard: I would say they go through the same thought processes, generally speaking. But of course, they have much more powerful tools in order to help them. In general from the CAD perspective, when CAD designers were drawing, when they draw, they're basically just documenting the idea of what the design is. So CAD, in its basic form, is really a documentation. It's no different than if you were to describe something via a Word document, if you will, only it's doing it from a visual point of view. So the tools in the past were more like you said, it was the intelligence in the engineer's brain and their experience to design products. Today, they have very sophisticated tools to do that. And they don't really think differently, but they actually think, if you will, with a lot more power. And they actually put a lot more of that physics and, if you will, advanced engineering upfront in the design to help them guide the process before they actually put it to paper. And in many cases, actually in combination with putting it to paper, so they get linked. And so as you do the simulation, it reflects in the CAD and they work together.
Jim Anderton: Now, I have to be careful. I should really refer to engineering professionals. Because today of course, people that design aren't necessarily professional engineers. When they think of power in the context of the tools they use to design things, what do they mean by power? Do they mean the same thing that say, the code world or computer people think of?
Brett Chouinard: Well, I mean, that's a good question. In terms of power and the power of design or are you talking about the power that's in the actual products themselves?
Jim Anderton: Perhaps both. But let's think to start with about the tools.
Brett Chouinard: So the power of design, from my point of view, you can do thousands of simulations actually in what you would have normally had to do, if you will, on a physical test. So this primary transformation isn't so much from CAD, from my point of view, although it's an important transformation, the primary transformation was you would simulate via physical testing in the past a great deal. And we still do a great deal of that, but you used to do it all via physical prototypes. So if you didn't make a physical prototype, you weren't simulating. Otherwise you were just doing basically calculations on paper, if you will. So today, they actually do a lot of ... we actually are supplanting all of that physical testing with simulation.
And actually when you're doing a simulation, instead of doing one or two physical tests, or maybe 100 physical tests, we can do 10,000 simulations on a vehicle. That is the power that we're talking about for simulation, where you will do millions, millions, tens of millions, if you will. One other really interesting thing with simulation, which I think is hard for people to necessarily understand, is that when you do a physical test, let's say you do a physical test on a car. Okay? And it's a very expensive test. It's a very elaborate test, it's a very expensive test. You have to build a whole prototype. You have to basically build a car and then you instrument it. And when you instrument it, you have to choose where you're going to get information. Okay. So you say, I want to have a camera angle here like we're showing in our show. You want to have accelerometers in these locations. You want to have strain gauges in these locations.
When you do a simulation, we have 15 million channels. We can get data anywhere. So everywhere that builds the model, we can get data from. In addition, we can say, take the door off. What does it look like inside the door? Make the roof transparent. What does it look like when you're looking from the top when the roof is transparent as part of our simulation, if you will? Those are sorts of things that we don't talk a lot about in the power of simulation, but it is really crucial.
As a matter of fact, one of the key things in a crash simulation is energy management. How do you manage energy? We can actually determine where the energy has been absorbed in every single part of the car after it's done. You can't do that in a physical test. You can look at it afterwards and say, "Obviously this thing absorbed a lot of energy because it got crushed." But how much energy? In the simulation, you can literally make a histogram. Here's all the parts. And this is how much energy they absorbed as part of the process. These are the kinds of things that I think are power of simulation.
Jim Anderton: Brett, that's an interesting point you mentioned about tests. I think back in the day, or I might have a fistful of type K thermocouples and I want instruments, a heat exchanger to do heat rise measurement. And usually I was limited by the number of channels available in the data acquisition system, which would typically be something like, as I recall, I think 24, 28 perhaps at the most. And really you didn't want to go much more than anyway because the data reduction problem then would rear its head. So the question then was that you had to sort of use gut feeler intuition to figure out where do I put the instrumentation and then sort of make inferences about what's happening between the thermocouples, if you will, to make that work. But there was no sort of hard, fast rule for doing that. And often, I do that and I wonder if the customer is testing it the same way, I wonder if everyone's on the same page.
Brett Chouinard: Well, I mean its repeatability from one test to the other. It's repeatability from engineer. It's repeatability from, if you will, OEMs who make things, how they do it. That's why there are a lot of standards around that kind of testing, especially in things that need to be certified, like crash simulation, to get a certified vehicle for safety, if you will. They have to guide that more. In simulation, we still have to. We have to make sure we satisfy those things in the same way. But by definition, you just get a lot more information. So that's positive and also can be negative. It's a very overwhelming flood of information. And so you have to develop methodologies to boil that down into something more manageable.
And of course, the whole industry is doing that and has been for a long time. So it's exciting. It's a huge, huge change. So in the past, it would take 120 days to build one final model of a car, let's say, one car. And it would take many, many days to solve it. And then many days, sometimes a month, and then many days to post-process it. Today, you can basically draw all the files that it takes from a CAD system, a PLM system, literally immediately. You can mesh them automatically overnight, oftentimes in one day. You can solve it in five minutes. Then you can have automated reporting that gives you the reports in, let's say, a day. So that process is compressed massively. And on top of that, you can then run it through a design of experiments or a stochastic study or something, run it on a high-performance computing environment, and manage that through software. And the scale is just sort of astronomical. It's just sort of branching out in all these directions.
Jim Anderton: Brett, the hot topics, almost the buzzwords I hear a lot today, digital twin and model-based systems engineering. And of course, now we add the cloud to this. So we've had COVID, people have been physically separated in ways that perhaps engineering teams weren't in the past. So we're working differently. Are we going to change the way engineering teams collaborate some projects? Most of the things that we're talking about, automotive aerospace, we're looking at fairly large engineering teams that are expected to work together.
Brett Chouinard: For sure. So I just met ... just yesterday I was thinking about that, I was driving here to do this interview. And one customer, if I told you their name, everybody would know them. So they're super famous. And in that meeting, the statement was made, "We're going to be 100% virtual. We plan to be 100% virtual in our prototyping by 2025." So I'm like, wow, that's impressive. So when you do that, of course, now you further enable model based systems engineering. So model based systems engineering has historically been connected mostly to the electronics world and to electrical simulation, because they have been able to be more, I don't know how to describe this ... but it's easier to make it in an automated fashion. And so that you can connect people just by the nature of the data.
In the 3D world, it's more difficult. The data is disparate, it's all over the place. It isn't naturally necessarily connected, that data. But if you're going all virtual and everything is virtual, now you can connect all of those kinds of simulations. This electronic simulation with full 3D systems, computational fluid dynamics, explicit crash analysis, static crash and static analysis for durability or so on. All those things can be put together in a model, if you will.
And then you can collaborate across all the groups that need to operate. So that's a natural cloud. That's a natural cloud extension because basically what you're going to have is systems that manage that. Here you have requirements. I have 10,000 requirements that have to be satisfied for this vehicle. For each requirement, you will have a model based process. It may not be a specific model that is an electronics kind of a model, if you will, like a partial differential equation solver. But it will be a process that defines how do you get that simulation result from requirement to validation. And then all the people who have to work in that, they'll know their part. They'll actually interface with this on the cloud. They'll upload all their data. It's got product data management, it's got simulation data management, all of that is included. It's the wave that's coming. That's the next step for us, what's coming.
Jim Anderton: Brett, I'm glad you brought that up because these days, integration of course is the other big thing that we see in engineering software. And we're looking at modular software in many cases, integrates things like ERP, PLM, quality, a lot of functions. A lot of functions, which in some ways are peripheral to the engineering design and development process, but is sort of built in. Now in the world I come from in automotive manufacturing, we would design a part and sort of throw it over the wall to the manufacturing engineers. And the manufacturing engineers, of course, would look at this and say, "Well, we can't make this the way you have designed it." I need you to relax that tolerance there. I want you to break that edge. And I need a softer radius. And the fistfight would start. And then in that process would emerge a producible, low cost, but effective part. With the software, with model based systems engineering, this ability to actually work collaboratively the way you're talking, is that going to end? Are we going to reach a point now where basically you get a manufacturable, producible, low cost part right from the get-go?
Brett Chouinard: Well, right from the get-go is an interesting way to describe it. Okay? So we will continue to get better. So manufacturability is already being introduced very, very early in the development process. Whereas as you mentioned in the past, maybe you could say it was an afterthought. I don't think that's a fair thing to say. It's not truly an afterthought. But in any case, it would often cause stress in the process at perhaps an inopportune time or a costly time. Right? But today, the ability to really integrate, if you will, not only manufacturability, but also choices between manufacturability, here's a part, it needs to do this, it needs to look like this, it needs to fit into this system. What is the best manufacturable? What is the best manufacturing process, first of all? How manufacturable is it?
Trade-offs between those processes. And then move forward with a better design, with a better chance of succeeding. Not only that, but you actually have the ability then to perhaps to go back. If you say, hey, it's a different process. What would that design have looked like if it had been that different process? We can actually make those designs that look appropriate and that actually work appropriately for all those manufacturing processes. So it won't eliminate that risk, if you will, of manufacturing being a question. But it will definitely reduce, no question. And today it's true. I mean, first time right, if you will, if that term has been out there in industry is much, much more so than it ever was before. No question.
Jim Anderton: Yeah. Brett, in the big OEM world, and I'm thinking airframers particularly, I understand in a past life, you also worked with a very famous manufacturer of gas turbine aircraft power plants that once made locomotives and light bulbs. They'll remain nameless. But for companies like Boeing, Airbus, for example, the trend appears to have been to outsource more and more of the product and become final assemblers of very large sub-assemblies, entire fuselages, wings, empennage, very, very big chunks of the final product, and to push a lot of the design work down as well. So it's no longer a matter of just we're going to find a contract manufacturer who can make a horizontal stabilizer. They're expecting the suppliers to actually design those components as well.
So you're looking at a collaborative world where you're talking about multiple suppliers doing sort of product criticals for mission critical work that has to integrate with other suppliers, doing things perhaps a continent away. Is everyone using the same platform to do this? There was a time when the big OEM customer would sort of force everyone to then get onto one common platform to make that work.
Brett Chouinard: No, I don't think it has to be one common platform. I don't think that's really necessary that you have one, platform. If you will, not one major CAD platform, for instance. It's not one major simulation platform even. I don't think it has to be one platform. What it does mean though is that all of those tier one and tier two and perhaps lower suppliers, they need this expertise and they need to be able to fit into this ecosystem that will be created for those suppliers. So this system that I'm describing where you actually have requirements to verification, some of that will be, well, that requirement is going to be satisfied by a supplier. Okay? So it doesn't necessarily have to be that they have to be able to go into the physical system and work within that physical system. But it will be managed by that system.
Okay? So that they'll say, okay, here's the supplier. They have a subassembly. They're delivering it. These requirements fit to that subassembly. They will manage those. When they come back into the system and they basically register their design, those requirements will be basically managed. Today, I'm sure some people do it better than others. Of course, some are very, very good at managing and other people are perhaps a little bit chaotic in how they manage that. But they do it very well. There's no question they do it well today. I mean, all of the major automotive and aerospace and the locomotive, as you mentioned, they all do this quite well actually. And like you said, it was born out of necessity from the airframe manufacturers, if you will.
Jim Anderton: Brett, the automotive industry, of course has done something similar. In some cases, entire power plants, Cummins is a major supplier to Stellantis, is one example for the Ram truck program. But that same power plant manufacturer that you worked for innovated things like door modules. The concept of pre-assembling, not just an automotive door, but everything that goes inside it as well for the automotive industry. Boeing has gambled literally billions of dollars with the Dreamliner on an extensively outsourced product in which their finalists have a 737 program pretty similar, you know, entire fuselages now arrive by rail. In the automotive industry, there's always been these sort of centripetal sort of forces that want you to design air conditioning compressors and alternators and starters out of house. Even though of course, this was vertically integrated. But now we've got COVID and we've got broken global supply chains. And tens of thousands of vehicles are sitting incomplete in lots wanting a single microprocessor for completion. Is that going to alter this landscape, do you think? Are we really still going to sort of globalize the design and supply process?
Brett Chouinard: I mean, I think we will. I don't think that's going to go away; it wouldn't be easy to make that go away. But there will be an ebb and flow, I think, regarding this. I mean, some of these critical items, if you will, chips, as we all know, is one of those critical items ... some of them will be more controlled by, if you will, major OEMs and so on. But in addition to that, I think collaboration between them will happen to make sure that they create resources that are more dispersed, maybe globally dispersed. They will work together to a degree, in my opinion, to solve some of these problems. I mean, it won't solve the immediate problem. Probably that will take a little bit different approach.
But I do think that it will ebb and flow. You know, companies will get control more of their process and then they will give some of that control up. It will come and go. I think it has over time and it will continue to do so. But I think that numerically, simulation-wise, we're only getting better at being able to separate systems and subsystems and be able to have those things be in the control of suppliers to be able to still make a really great product, if you will. So it won't be because of the infrastructure for the, if you will, stimulation upfront. But of course, this is actually a supply chain issue, not necessarily an engineering design sort of system. And so, there definitely will be a motivation for them to get control. But I don't think that they'll do it individually. I think they'll do it in groups.
Jim Anderton: Brett, is the fidelity of simulation still an issue? I recall sort of at the birth of this widespread adoption of this technology, your senior engineering managers would frequently say the simulation is nice, but go cut some metal. And I want to see it bolt together.
Brett Chouinard: The fidelity will always be an issue. I mean, there's no question. I mean, there's always seemingly endless amount of demand for more accurate or more. But I would say that probably though in the last, let's say, 10 years, what has been more than fidelity of a single simulation, it's been more simulation, more kinds of simulation. Okay? So it's really spread to all the possible disciplines you can think of in terms of physics, structural and fluid and electromagnetics and system power and power integrity. And, oh my gosh, it's just really, really big, broad list of things that you can simulate. And now it's actually multi-physics. So how do those things work together? But of course, nothing exists if it's not multi-physics. Nothing exists in a single physic. It's not even a thing.
So that's a really big thing that people will be doing more multi-physics so that makes sure that you're getting the interaction of those things and you're not having to send data from one place to the other. But always accuracy and fidelity and everything will be, will be a thing. You know, it's not going to ever go away. Nobody wants it to be not accurate. Okay? But the drawbacks to that are how fast can you do it? How fast can you get a result back? And what's the cost of that accuracy? That's the trade-off that our customers are doing, and we are doing every day.
Jim Anderton: Last question. In a world where we can, with high fidelity, simulate and iterate with blazing speed. You mentioned the possibility of tens of thousands, hundreds of thousands, perhaps millions of ways of simulating something with these advanced tools now. Is it possible now to iterate our way to success? Can we take a design, which we know is wildly suboptimal just to create a starting point and then throw the switch and let the machine essentially lead us to an optimal solution? Am I really talking about sort of a future 10, 20, 30, 50 years from now where we put ourselves out of work and essentially anyone that can sketch a design can become an engineer?
Brett Chouinard: Well, I don't really see it that way. I don't really see that putting ourselves out of work because it seems there's always something more for us to do. I mean, in some ways, we do eliminate some tasks. Okay? So for instance, if you're a software user and that software is very complex and you know every possible way to use that software, is your value how you navigate that software or is your value the result that comes out of that software and how it affects your product? So we're going to have a less of that first one. So less and less of the value of the people who are involved in simulation and in design and product development is going to be in the expertise in the actions that they take and more in, how does it affect the functioning and the performance of the product, the cost and the delivery and all that sort of thing?
So I think that is a really big thing. I don't think anybody's going to be out of a job. I do think artificial intelligence and reduced order models are something that's going to make things super-fast. So a lot of these really big models we have are going to be condensed to be run really fast so they can be combined in this multi-body, or I should say this model based systems approach. And I think it's just going to make us more efficient. I think companies are going to design products faster, they're going to make them better, more times correct the first time, less warranty. All those sorts of things are going to be very real. And of course, they're going to ... we didn't talk about this here so much, but they're going to connect them to digital twins, all that stuff is going to happen.
Jim Anderton: Brett, thanks for talking to me today on the show.
Brett Chouinard: Sure. Thank you. I appreciate it.
Jim Anderton: And thank you for joining us in designing the future. See you next time.