Methods, not technologies, are the key to an industrial future
It may not be everywhere, but transformation is already happening on the shop floor to bring AI technologies to operators. The conversation isn’t even about the technology, in many cases it is not the limiting factor. Listen or read below to hear from Rainer Brehm about the future of industry and the shop floor.
Nick Finberg
00:09
It’s time for another episode of the Future Ready Podcast from Siemens. I’m Nick Finberg, your moderator for Conversations on All Things Software Defined. We’re continuing our discussion with Rainer Brehm, CTO and COO of the automation business at Siemens Digital Industries on the role of software-defined everything in the evolving future of industry. In the last episode, we talked about how AI technologies can be leveraged on top of the foundation of software-defined automation, and it sounded super interesting with customizations for the operators on the shop floor. How fast is that transformation happening, Rainer?
Rainer Brehm
00:42
The technology is there. I mean number one. So, it’s not a topic around technology. I think it’s a topic of how You bring it down and bring it into the shop floor, yeah, because still maybe some compute is too high and you need to have a lot of storage. Yeah, that’s a topic we’re we are working on. For example, when we talk about AI bringing not large language model but smaller language model, more specialized language models down to the shop floor, number one. Number two, in order to do it, to create this magic, you need to train them, you need to access the data of the specific machine, of the specific side, of the specific production. So, you need to have the working manuals, you need to have the lock book, how this machine will operate, you need to feed this data into the system that it can you know leverage that data. And we see that this integration work You need to invest money in doing that. So, but if you can do that smarter and smarter, then this initial investment also is easier, and we’re working also on that. And I think if we can do that, I think there will be a good ad a great adaptation if we can automate that integration, yeah. And but we have a lot of customers which already see a great value in and using it. I had a had a customer which even an executive person was looking at that because also for that company it was very new and we deployed it and in with a with a pilot customer that kind of co-pilot for operations. And that executive c came from an Arabic country. And the interesting thing was, you know, you could communicate them with a system and kind of you know you don’t need to type something, you can you can talk with the system And the interesting thing is that person could even talk with the system in Arabic and got responses in Arabic, even so there was no training on Arabic languages. And if you see now whether it’s you have something you, do it in in in in English, but you maybe something somebody with Portuguese skills or Spanish skills. comes over, you know you could interact even in in different languages and system takes care about it. So that’s also something to make it more accessible because you can interact maybe with your mother tongue. with a machine if you’re not capable of you know having maybe some specific terms in in English or whatever is the machine was designed in.
Nick Finberg
03:05
That’s really interesting. Yeah, because there’s some language that’s just very hard to think about. Especially with technical information. Like yeah, there are some cognates and things that are just transient across all languages, but there’s some things that are just very hard to conceptualize in another language. If you’re an expert with something other than English, like you learned physics or something in in Portuguese for instance. It can be very hard very hard to translate.
Rainer Brehm
03:38
And the other and the other topic is I think we have a lot of global customers. which say hey I designed something, but I need to deploy the machines. I have factories all over the world and I want to use; I need to go into standardization and currently what they’re doing they are translating all the things into all different languages and really program it hard coded in in the future. And then probably again one language is missing because the operator just came from whatever country. And I think that it gives much more flexibility.
Nick Finberg
04:11
So, you talked about that a little bit, the distinction between doing this all at once for maybe a startup or a new business. how does that change between like a greenfield or a brown field situation? What’s the what’s the distinction between those two?
Rainer Brehm
04:30
Well, in a in a greenfield situation, I mean you can do a lot of things already from us the from the first start and we see a lot of customers when we talk about reshoring manufacturing that those customers say now if I now rebuild a factory which maybe has been I was producing only in one location in Asia, yeah, and I want to kind of re reshore it to the US, to Europe, whatever. they want to build a state-of-the-art factory and that’s already going very much into digitalization. On the one side. On the other side, the biggest market is always Brownfield because there are existing factories which you are upgrading, which you are updating. And youdon’t rip out something because you know you won’t be fancy because it needs a business case. So, at the first topic, this is what I said also getting first data connectivity. is the first thing to be done, especially in brownfields. And then if you have this kind of a data blanket, then you can start adding features on there and save and
Nick Finberg
05:36
Okay. So, this is we’ve definitely been talking about like single site or even multi-site, but it’s very much a…
Rainer Brehm
05:48
Yeah, absolutely. I mean we have those customers as well. And one nice referent is they are producing thousands of containers, and they put it somewhere remotely, but they want to control those containers centrally. And they even might offer that as a service. And they have different ways and they need to have a mass deployment. Like today, you do a software development, you change something and deploy it instantly. on all different units and that means you need to have a different way how you manage devices so you need to manage devices not On site because it might be a container in a very remote area, so you need to have a very good way of how you centrally manage devices. And or kind of the container. At the end, the container consists of a controller that needs to be centrally managed and updated. And we need to have that kind of flexible deployment mechanism. And you need to do changes as quick because maybe the clay is changing or you know the surrounding conditions are changing. So, you need to be very flexible acting. And we see that, or we have another customer which is a similar topic It was around COVID, but it also now comes that you know a lot of micro factories gonna be built. We have customers do micro factories to produce pizzas. We have customers doing microfactories to produce individualized medicine. so basically you have container where you flexibly put them somewhere in the world where it’s needed. Or to treat water, for example. Another good example. Maybe in crisis or you have a flood and you need to put something there. And there soft feed automation helps a lot because You’re centrally managing it, you’re easily deploying it, and that’s also another use case which where we don’t have a fixed factory which is built and run for twenty, thirty years.
Nick Finberg
07:37
Yeah, you’re thinking it with variables rather than hard coding everything so you can make it deployed very, very quickly for any number of situations that you’re kind of thinking about.
Rainer Brehm
07:48
Exactly.
Nick Finberg
07:49
Yeah. Yeah. What role does AI play in in this case? We’ve talked about it a little bit. We’ve talked about the customization. We’ve talked about data acquisition and manipulation for optimization, but is there something else that you’re that you’re kinda looking at?
Rainer Brehm
08:05
I mean first of all I told already on AI at the beginning, you know, if you can go from rule-based automation to a goal-based automation, that would be a dream. Yeah, you tell the system what needs to be done, what is the goal, what I wanna be done, and the system figures out itself how it’s gonna be done. And that is AI. And here we talk about Topics like physical AI, where you might have vision action models which are running close to the to the controller or on the controller and doing that task which was not thought before which is you know you automate the unknown. That’s one element. I see we are we are can be much better in optimizing and automating workflows. And I give you an example today. Like when you do an engineering, you do a you engineer your machine. and using our C A D tools and team centered doing machine automation. Then you start simulating and you do a behavior simulation and you and you do you do a then you do a sizing what size of motors you need to put in in order to you know to do the task, what mechanics you have and so on. So, you do this. And then if you have that, then you select maybe your automation equipment and say, okay, for doing this I need this kind of motor, this inverter. you know, I want to show this on my a on my on my panels, I need to have these switches, this cybersecurity, and so on and so on. and then if you have done this Then you say now I need to put it in a cabinet. You’re starting doing electrical engineering, electrical design, you build your cabinet, and then you start having your cabinet, you go into automation again and you do your hardware configuration in the automation environment. and then you’re writing a program and then you are testing and validating a program. You do it in a virtual world with a maybe a virtual plc and then you deploy it and you test it and so on. Andthen you want to further up optimizing it. So, there is a complete chain of maybe even different engineers doing different tasks in order to get something done.
Nick Finberg
10:13
It sounds like it’s a dance between like the hardware-defined and the software-defined worlds just kinda merging. together a little bit.
Rainer Brehm
10:20
Yeah, and then we always have the topic, you know, does it need to be done in in in a in a sequence or you can do concurrent parallel engineering? Then you just have the handover from the mechanical to the electrical to the automation engineers and then maybe something is going wrong there. And I could imagine that in the in the in the in the future those workflows are automated. So, we don’t only talk about automating automation from an execution perspective, but also from an engineering perspective. And we’re working here, for example, with other companies because again, we need to talk about the ecosystem. It’s not only Siemens, a lot of companies out there and also competitors of Siemens, which we need to somehow bring together because if our customers want to be more effective, they need to automate workflows. And here the logic around agents come into play. So, there will be agents which do this work, do this orchestration of all these different tasks. And for that, we also need to bring then AI into our tools and make them accessible for agents. So, we need to enhance our tools that agents can say, you know. I have now this bill of material and now I build a cabinet and maybe then we go to a to a company if One word Mark Lita’s E-Plan and E-Plan says, okay, I take the data automatically from an agent and build maybe with a co-pilot automatically this this cabinet. and then give the data back and do the automation engineering with it. So, we need to have that also in our mind how AI play a role. So that kind of Agent- based workflow optimization with a topic. And then if you are now on the shop floor, agents are gonna be executing things. And this is we talk about we will talk about physical AI. where you have then different machines which have different skills or maybe have a variety of skills which maybe you can even update on the fly. and then there are agents to get something done, to execute on a bit of process, they will automatically, you know, connect different machines, different skills to get something done. So, you can extend that logic of workflow optimization of agents then to the physical world.
Nick Finberg
12:34
Okay, that is the perfect way to circle back to the start of our discussion on copilots. And it makes a lot of sense since we are always talking about interconnecting development and operations. I have so many more questions for you, Reiner, but I think we should wait for our last episode. Thanks for listening and make sure to subscribe so you don’t miss the last part of our discussion with Reiner Brehm. Until then. Check out the description for more ways to learn about software to find everything.
