Engineering with Artificial Intelligence: This Tool May Change Everything

Siemens’ Mike Nicolai on how AI will positively change generative, simulation and systems engineering, among others

This video was sponsored by Siemens.

“Artificial intelligence” is perhaps the most overused and misunderstood term in computer technology today. Incorporated into key engineering tools however, it’s showing promise as an indispensable part of modern, holistic design processes. Once thought to be a sort of “virtual engineer”, in practice AI is evolving into a way to integrate generative design and simulation into the traditional creative design process, with a twist. New systems promise to work with human engineers well before the rendering stage, helping with the conceptual portion of design as well as optimization. Modern systems can seem like magic, and can supercharge human engineering teams with better designs, less risk, and faster time-to-market. Siemens’ Mike Nicolai leads a team that develops advanced design engineering systems that incorporate AI.

Learn more about Siemens Simcenter Studio.

The transcript below has been edited for clarity.

Jim Anderton: In popular culture, the future is all about technology, and by technology we frequently mean software. The most talked about, misunderstood yet widely anticipated technology is artificial intelligence. AI means different things to different people, but it’s evolved out of the laboratory into a truly useful tool for engineering processes, tools that will change the way engineers design and develop new products.

Joining me to discuss what AI can do here and now is Mike Nicolai, Senior Product Line Manager for Siemens Industry Software. Mike currently leads research, development and product dev teams for a new product in the Siemens’ Simcenter portfolio. He carries an engineering degree in electrical and electronics from the Technical University of Braunschweig and a doctorate in mechanical engineering from Rheinisch-Westfälische Technische Hochschule Aachen (RWTH Aachen University).

Mike, welcome to the show.

Mike Nicolai: Hi James, thanks for having me.

Jim Anderton: Artificial intelligence is being used in popular culture everywhere. You can’t turn on social media or television without hearing this word used. From an engineering standpoint, what do we mean when we talk about artificial intelligence?

Mike Nicolai: That’s an extremely funny question because it changes over the time, right? I mean it was invented in the 50s from the last century, and since then it was basically a moving goalpost, because at the certain point it was just that you that you can say where your car should go, but no one would call your GPS system now “artificial intelligence”.

Basically, the idea of artificial intelligence is artificial intelligence, meaning the idea of that a computer can behave intelligently, artificially, and then it’s all on the definition of what is intelligence. And that’s also what you find in in textbooks.

Jim Anderton:  We use the word “intelligence” very loosely. I think in in the history of computational technology for engineering uses, we think of a calculating machine – essentially a slide rule to a mechanical adding machine. But it was always about an engineer, a designer, a human being who puts input into a device, turns the crank metaphorically, and then the solution pops out the other end. But determining which data to put in and interpreting the output was still a human function.

With AI are we going to cross that barrier? Will truly artificial intelligence decide what to calculate and how to analyze results?

Mike Nicolai: No, I don’t think so. You can also call it “augmented intelligence”. We really see it that it’s helping someone to do a job – meaning, you can tell the system what you want, and the system is then trying to find out the right design.

So with our new tools, Simcenter Studio, we did that on the system architecture level. You can say “OK, I would like to have such a system, let’s say in hybrid electric vehicle. I have these kinds of settings and would like to fulfill these requirements”. And this is given to the AI and the AI is then generating a lot of alternatives. But then it’s again on the human to decide what’s good and what’s not good. It’s not up to the AI to decide.

Jim Anderton:  These days we speak a lot about model-based engineering, systems-based engineering, architecture and platforms. Is the long-range goal of experts like yourself in this industry, are you trying to create a virtual engineer – an artificial engineer? Will the will the AI or the system itself be like a second human engineer standing next to you?

Mike Nicolai: I don’t see it yet, so I think that’s not the end goal. The end goal is really giving the right value to our customers.

So, the tool is generating a lot of architectures – so that in itself is nice. But with a lot of these architectures, you cannot do much. So, let’s say you have 1000 architecture; and if you have 1000 architectures then you need another AI then – for example, one driven by machine learning, and the other one driven by machine reasoning and machine learning – to really find out what’s good.

So, we also have recommendation systems in because you may have a couple of stakeholders – one saying “oh, I like this one because the costs are good” and the other saying “this one is good because the design is good”. And all this information can be tied together and then it’s recommending that maybe this is the design in the right range. But it’s always in support, it’s not necessarily the Terminator style “I decide what’s good for you”.

Jim Anderton:  Mike, traditionally CAD/CAM helped engineers develop very sophisticated, very complex systems. But the information flow – the data stream – was always top down. A designer rendered something that was then created. It might have been prototyped and tested with a limited amount of information feeding back from perhaps a testing procedure, and then design changes would be made, and then the product would be locked for production and then it would go.

And then, often in the Automotive World where I come from, improvements or changes were often simply held until the next scheduled iteration, perhaps the next model year or the next platform change.

But we’re looking at a world now where the ability for information to flow in both directions from end users from production lines back up to the design office is basically unlimited. Data is essentially free. Does this play into your automation, your AI strategy from an engineering design standpoint?

Mike Nicolai: Yes, yes, so what we developed in the last 20-30 years are really sophisticated simulators so we can have this testing phase really in the design phase. We can build something virtually and we can test it – we can look at behaviors and so forth. And we can decide,” oh, that’s good, or that’s not good”. So, we could reduce the amount of prototypes to build, then to validate the actual design.

What we are doing now is the next step with this generative engineering is that we are going one step further: we let the computer not only help us with designing things – or with the design in itself, the cut design – but also with the different mechanisms which you can create.

So, the AI is generating a lot of different mechanisms and then the simulators can evaluate them. And hence you have a really a bigger range of possibilities which you can select from; and this ultimately helps you to create in the same amount of time and the same amount of effort, costs and so forth to create better products. And that’s what we are standing for.

Jim Anderton:  Now you mentioned generative design. Of course, that’s a hot topic right now in design engineering. Originally, one of the key aspects of an experienced engineer was their ability to start in the middle to not have to iterate many, many times to get to a workable design, just based on past experience and knowledge. But generative design promises a new world where you could start from a completely impractical rendering or design and simply allow a billion iterations through simulation and then produce a better product than a human could produce.

Does this mean the starting point now is anywhere? Can you get a monkey to design a bracket and then let the machine take over?

Mike Nicolai: No, no. A monkey I think can’t do it. You have to say what you want, right? You have to know what you want, and you can give some constraints. And with these constraints, the system can generate.

The problem is, if you start from a certain design then you have something like a design fixation, because you are fixated on a known design – that’s a good thing – it’s a healthy thing if you if you’re alone that you know something, you have some knowledge from the past and this knowledge brings it further to the next design.

But if you can look at more options and try to consciously design and remove this “design fixation”, that really opens you to new worlds. And that’s why we call a generative engineering and not calling it only generative design inside Siemens. Because generative design is for us, usually in the CAD world – you really decide the boundary conditions are, where the loads are, etc. and where you can build something and then you create the part.

But generative engineering takes this idea of generating designs further. It goes in the early concept stages where you think about the different system architectures. But it also goes to the later phases where you go to generating control. So it’s a little bit broader than only the generative design, which is on a component level.

Jim Anderton:  One of the hallmarks of generative, of course, is that ability to simply iterate in ways that are faster, cheaper than would be humanly possible by prototyping. In the original way, defining the requirements, defining specifications was the most important part of the process. You couldn’t begin properly unless it was clearly understood what the device must do. In that world, making a change to those requirements was often impossible, or it was highly restricted. And this would often happen, in my experience, where the end user would say “we’ve made a change, we must add 50% to the strain capability” and the answer was “no impossible”, or we must go back and start again from the beginning.

Does generative mean that we can be more flexible, perhaps with the downstream part of the engineering process. So, the end user can make some changes or make some alterations or change their mind without causing chaos upstream in the process.

Mike Nicolai: Yes, you nailed it because the idea is you can say what you want, the specifications, the requirements – it’s not yet the textual requirements. It’s really specifications that you put in. You put “I want to have this, and this area should not be used”, or “these systems should not be connected” – are not yet requirements. But it’s a higher-level description of what you want to have.

And then the system is generating a lot of alternatives. And then you find out, maybe through simulations or just looking at the generated assets “ah, that’s not good”. And then you can really quickly iterate back to what you want and you’re faster.

Jim Anderton:  So, Mike, it sounds like we’re moving from a world in which we have one solution, the answer or the optimized answer to a spectrum of possible solutions to this. So now are we looking at a possibility where there is no one design anymore – now there are 15 or 20 possible designs, all of which will work?

Mike Nicolai: But we already have this, right? I mean if you look at the car – look outside on the street – you have so many different designs. There is not “a” car. There’s an idea of a car, and then there are a lot of instances already running on the streets.

And what the generative engineering approach is now allowing is that in one company you can look at 100 or 1000 or however many designs you want to have and then decide with a larger set of stakeholders “that’s the car we want to bring or the product line which we want to bring”. So, it’s already there, but it makes it a bit more feasible for the engineers to handle.

Jim Anderton:  In typical design or typical engineering teams, it’s quite common to have one member who is the crazy thinker. He’s the creative one who comes up with an idea or a concept which is completely different and sometimes something that seems unacceptable – very frequently it’s simply not cost effective to take that idea and develop it, because the probability of success is perhaps low. The payoff if it works is tremendous – but the risk is too great to proceed. Will this technology change that? Will this be possible now to try the crazy idea?

Mike Nicolai: Yes, that’s basically what we see already. Our customers have tried that. We have this tool. It generates a lot of different designs.

And one of our customers, he generated 3000 different system architectures. And then we have the tool called Discover where you where you look at all of these. You put your preferences in, your likes and dislikes, as you find in Netflix, for example, or as like and dislike buttons found in social media apps; and then at the end the system pops up a ranking of designs and his design was in and he was presenting it to his management and management said that’s good. That’s indeed then the right choice to go.

And we’ll see when it comes to the market. If it comes to the market, I hope so.

Jim Anderton:  Mike, it’s a very common thing in engineering for an individual to train in one discipline but end up working in a completely different discipline. Very frequently I’ve seen individuals who train in electrical engineering who end up designing mechanical systems sometimes the other way around. Will this AI enable technology help individuals make that transition from one discipline to one that perhaps they’re not as trained?

Mike Nicolai: Well, if you train in one thing, let it be mechanical engineering or electrical engineering. I mean you learn the basics and so I think this will stay hard because you really need to go a deep and so forth. But what we see multiphysics simulations is that you anyhow have to build up this system. And you will simulate it. I mean, it’s already relatively easy to build up a thermal system as a mechanical engineer – put some vehicle dynamics in and some batteries and simulate it without really understanding 100% the solution of the thermal system, you don’t write the solver anymore, I mean you can do that.

But basic understanding is always good.

Jim Anderton:  Mike in in the early days of automation or the engineering design development process, a criticism was that an electrical engineer would say “I wish to design a power supply. I don’t want to become a coder. I don’t want to spend time learning how to use the tools; I want to use the tools to get to my end product”. Are these low code/no code solutions, are we talking about a world where this will liberate the engineer to do their core purpose, or will they have to go down that road of thinking about how I work with the tool rather than just using it?

Mike Nicolai: No, I mean my role in product management is really thinking about who are the users of our tools and then make it as easy as possible to use the tools.

I mean, we’re also using their AI. We need to support them. I mean, if you create your model and I mean it’s a system model and model-based systems engineering mode, so, you have a component, you have your systems, you have assets connected to certain components, you have connection. And what you can see in studio, for example, you can press on the component on the sketch, on the on the textual representation, and then we have a helping system which is directly popping out so you can do this, or you can do that.

It’s an inspector which is really supporting you, so you don’t have to look up the documentation. Or it’s also recommending that maybe you want to have look at this example and so forth. These are the little things which we have to bring from product management to really support our users, because it’s not about coding. It’s creating a design. That’s where I want to go with my tools.

Jim Anderton:  Mike, in the beginning of CAD/CAM systems they were machine-centric, and it required very powerful, very expensive workstations. And now we’re moving toward a cloud-based model where a relatively simple laptop or even a handheld device can be the tool which the designer uses. Tell me about cloud connectivity, is it necessary for using products like yours, and if so, what advantage does it bring?

Mike Nicolai: Simcenter Studio is a cloud native application. What does it mean? You basically access it via your browser. And how the service is running depends a little bit on the customer. We don’t have a software as a service solution, at the moment. So, the customers are usually installing it on premise. But you can also install it on your private or even on your open AWS or Asia or whatever cloud provider you’re using. You still need the power to run your calculations or simulations and so forth. But you can view and edit the information then on your laptop if you like – if you have really small fingers – or on your iPad or tablet.

Jim Anderton:  Security is definitely an issue with any cloud connected device and I speak with many individuals in engineering firms who work with some things that are contains sensitive information for preps military contracts, some that are restricted by regulation, ITER in America, for example, and some that simply have very valuable proprietary knowledge that they’re reluctant to release to the world. Are your applications or things you’re working on? They can operate in an intranet or in a closed system as well as an open system?

Mike Nicolai: Absolutely closed system, so we have some. And some of these systems installed in the US on a completely island basis? No connection to Internet needed. It’s really an on-premise solution which you can use no problem. I mean that’s that was a design initially. You don’t need to download here a library, or they are reliable, that’s that all in the system. This this, these are some of the software requirements which we have.

Jim Anderton:  In the early days of computer aided technologies there was a mismatch in capability and power between very large and wealthy firms and smaller firms. Of course, the large firms were early adopters, and they ran with the technology and there was concern amongst smaller engineering firms that they would have difficulty keeping up and also even have difficulty retaining staff who would tend to migrate to larger companies that had the leading-edge cutting-edge code with technologies like yours, will this have a democratization function? Do you think that will this be accessible to even the one or two-person design office as well as for Volkswagen for example?

Mike Nicolai: I would not see why not. I think the world is a little bit more complicated than “you have small companies, and you have big companies”. You also have venture capital, which is pushed into small startups, so they have actually more capital to use on a business case and can go then even faster than the small companies can buy them. I think is not OK to think “I have a small company with three people and I’m not moving”. No, you can be moving fast, and we help you to move faster with our tools.

So, you don’t need to invent everything yourself, you can have one seat or two seats. That’s no problem.

Jim Anderton:  Mike, a final question. With AI enabled-advanced technologies like this will the future require engineers? Will it be possible for anyone to design anything simply with the power of the machine?

Mike Nicolai: I think we will move more to the ingenuity part rather than the tool usage part. I mean, the generative idea is you really say what you want – and we want to make that as easy as possible – and the system will generate a lot of things.

So, what you have to do now, if you use a tool, you have to understand “how do I move these blocks? How do I connect these? Why can I not connect these blocks? Or how do I, in the CAD world, how do I extrude this one? Or how do I connect this one?” And this is tedious work, I mean pushing the mouse around. I think this will move away, that’s the same thing which we had in the in the CAD world.

In the CAD world, you had these 50 engineers working on the different views and everyone was working on that. It was also great because we had 50 engineers really understanding it. But then CAD systems like Anix came and they revolutionized the way to use a CAD in this sense.

And it will happen also with system simulation, for example. You will move less the mouse to do something, but the system will do more for you that can be creating views that can be easier  creating simulation and so forth. I think engineers will be used more and more and there will be more people becoming engineers. You talked about monkeys pushing the mouse, I think that’s less and less. And I think that’s great. I think I personally like it. I hope also your viewers will like it.

Jim Anderton:  A bright future. Mike Nicolai of Siemens Industry Software, thanks for joining me on the show.

Mike Nicolai: Thanks Jim. It was a pleasure talking to you.

Jim Anderton (to audience):  And thank you for joining us on Designing the Future. See you next time.

Written by

James Anderton

Jim Anderton is the Director of Content for Mr. Anderton was formerly editor of Canadian Metalworking Magazine and has contributed to a wide range of print and on-line publications, including Design Engineering, Canadian Plastics, Service Station and Garage Management, Autovision, and the National Post. He also brings prior industry experience in quality and part design for a Tier One automotive supplier.