Software Makes Hardware Matter More
Software is a great enabler across industries and disciplines, but for manufacturing and production especially it needs to be grounded in reality. The interconnect with hardware is where this happens, where the software gets information to understand the world it is helping to optimize. We have Rainer Brehm on the podcast to talk through this idea of increased hardware in a software-defined world.
Nick Finberg
00:09
Welcome to the Future Ready Podcast from Siemens. We have Rainer Brehm back on the podcast today. In some of our previous episodes, we’ve talked about building a foundation of software-defined to enable adaptability, we’ve covered some of the best practices for companies starting on this journey, and we dove into the topic of how technology is aiding operators and engineers. But you made a really good point about the extent to which we can do software defined, and I wanted to talk with you about that in depth. You said a business could deploy an extremely generalized machine or a gripper, but it won’t be very effective at most things, just capable. And that this is the dynamic that is playing out even now in the industry, that balance of goals. Would you mind talking a little bit about how hardware will evolve in the software-defined future?
Rainer Brehm
00:57
That’s a good question. And I honestly have also this question in the Siemens internally because we have factories and the factories say, yeah. If everything goes software-defined, do I still have something to work on? And I can calm them down because I guess we will have even more hardware in a software-defined world. And I can I can tell you why, because I think if automation gets more and more adapted the intelligence is gonna be more software defined, but automation interacts with the real world. And in the real world you have sensors and this is you can do soft sensors by the way, yeah, but this would be another episode. But still, you’re gonna have sensors and You definitely have actors because you cannot move something by a software. You need to have a power, and you need to have probably a motor or inverter. So, there will be motors and inverters, and even in a software-defined world, this will not be replaced by software. So, there will be probably if software if automation gets more and more adapted and can Solve more and more tasks which haven’t been solved in the past, you probably beat more hardware than less. Number one. Number two the execution platforms where you execute the intelligence when we talk about software define automation, that might change. That might change from being connected to a hardware to be decoupled from a hardware as you have explained. It might still run embedded, it might run on a virtual machine, it might run in a Kubernetes cluster, it might run on the cloud. So, I think there there’s a change. So, in hardware execution hardware can be more generalized, but can also very much very much empower because when we when we work together, for example, with NVIDIA, which was just announced on the CES, I think a lot of TPUs, or GPUs are necessary to execute AI workload. So, there might be even more compute power be necessary close to the machine because you want to have a certain reaction time and you want to act in a safety manner. so therefore, there might be more compute necessary. but not in dedicated system, but more in general systems.
Nick Finberg
03:13
So, you mentioned NVIDIA and it’s a super interesting partnership. Can you talk a little bit more about that, that powerful partnership as well as some of the other ones that were we’re working on here at Siemens.
Rainer Brehm
03:23
Yeah, well, you know everybody talks about IT OT convergence and that means we are bringing IT Hardware IT mechanisms into the manufacturing world. And that is basically where we’re working on. It doesn’t mean that it always what we use for IT does 100% fit into the OT world. Yeah, maybe there’s some adaptation is necessary, and that’s something we’re working on with our IT partners. it’s with Nvidia for example how you bring how you bring hardware accelerators into more embedded systems yeah where the power consumption is very relevant, where the heat is heat topic is a very relevant one. so that’s one topic. But we also work with them now how we can for example, move into a more autonomous factory. That means how we can bring AI intelligence into the automation systems. And we talk about vision action models or how we can control a robot much more smarter than we when you do today. So how we can go into a goal-based automation, that’s also something we’re working on. And the other one I think which would also be our own episode is a lot nothing around this kind of industrial metaverse, you know, how you can how you can do in a planning phase it much more realistic. Much more even might be even photorealistic. and how you might use synthetic data in order to train networks which are then maybe executed in a factory. In these areas we are working closer together with Nvidia to move forward into a more adaptive and autonomous factor.
Nick Finberg
05:04
Maybe a follow-up to all of that. What should customers consider in deploying this kind of software and architecture? Not all of them, of course, are going to be on the latest and greatest or want to invest all that money right away. What are what are some of the options?
Rainer Brehm
05:23
I mean first of all, connect. Get the data. We see a lot of customers we have which went into standardization have a benefit already. We see a lot of automotive companies, for example They established since I don’t know 20 years a standard and said this is our standard, how I want to produce. And that standard helped them now because if you have a standard, Data is standardized, it’s contextualized, you know, it’s not you don’t have apples and pears, and everything is different. So, standardization I think is a very, very important topic. standardization contextualized because if you want to use the data you need to contextualize it you need to know what this data really means yeah and So ontology is a very important topic. So, building up this one is something you need to consider already when you build something new, because this is gonna be the basis all the time. Even what new technology, what new LLMs comes up, that is something you need to do. And then we come back to where to start, you know, start with a pain point, work with partner, work in open ecosystems. in in order to go in a continuous journey here.
Nick Finberg
06:33
So, your counterpart for these discussions on software defined, Mark Hindsbo, has been helping me examine the operation side of all this, but I’m curious how you see this other half of the discussion, how it integrates with automation.
Rainer Brehm
06:47
Well it’s it Yin Yang it’s basically having a black box system doing a task is very much automation but it need to be orchestrated it need to you know you want to produce something normally you produce something with several steps And you need to have a work order, how you do that, and you have a certain material which you need and need to process. A bill of material, a bill of process, how you translate a bill of material, bill of process, a work order, a batch into the factory, into the automation system. And here operational software comes into play to know to break this down, break down the work orders into the machines, and then let the automation do the job. And so that is the that is the massive interlink between the execution and the job to be done.
Nick Finberg
07:35
What do you think about is the interplay of Software Defined and the Digital Twin? How are those two connecting to each other? Is one supporting the other? Are they working together?
Rainer Brehm
07:52
Siemens, we are really, really promoting a comprehensive digital twin, which is you know having different facets and we are we are promoting a digital twin which goes over the life cycle. And I think that going into the life cycle it’s not implemented in in in a lot of a lot of cases. Currently you use a digital twin, you are planning a product, you’re planning a factory, you’re simulating the hell out of it. And then you are going into maybe a virtual commissioning and then the factory runs and then it operates. Fine. And then if you’re really good, then you take the digital twin if you want to make changes in the factory and kind of simulate again and then do that, apply the change to the factory Also good. I think the next step is you’re using a digital twin and run it in parallel to the real production. So, we go into an executable digital twin. And that executable digital twin could run in a cloud, but could also run very, very close to the OT workload. And I believe when we want to have that kind of, I talked about in software development you have DevOps. So, you do this development. And then a very consistent testing before you deploy something, before you put something into operations. In the OT world, that’s not so easy. Because in the in the software world, if something goes wrong, you do an undo. You know, if you do something on a website and you do a change and then it doesn’t work, you kind of recover it to the old one and then it works. In the real world, if a machine crashed, there’s not an undo button. You cannot undo a crash. And even worse, if you injure a person. Because something went wrong, it’s even worse. So, you can also not even somebody you know get really hurt it, you cannot undo. But I want to get the flexibility of changing and do this kind of devops and in my opinion, that will be only possible if you have the digital twins, which are really representing the line, the machine, the operations. It’s gonna be feeded with the real data as a real operation gets the same data. You have access to the same data than the real program operating a machine and then you simulate with the real data not with synthetic data or whatever old data and then if you see it really everything works fine and then you switch over. Yeah, and I think that and that digital twin will be a prerequisite in order to go into a DevOps of automation into the shop floor. And that is an executable digital twin which runs close to the automation. That’s one element where probably what I talk here, this is more future of maybe three to five years in my opinion, that is really gonna be applied. But there are other nice examples where we have that array today. And this is we have a nice topic which called soft sensor. Where you have a digital twin which runs really in operations. So runs, for example, on our industrial edge system. We have an app called Live Twin. So, the Live Twin runs and it runs the simulation model. And it’s get feeded by real sensor data out of the process. And then it calc it has a simulation model, so it calculates maybe then some outputs. Which they can be then be used to optimize the process. And for example, it can replace sensors because you don’t need to have a sensor, for example, of the energy consumption of a machine, if you know this is the rotation of the motor, this is the viscosity of whatever I gonna do. and I know the speed of the motor and then basically with a simulation model you can already s kind of calculate for example the energy consumption and you don’t need an energy meter because you can calculate it with a soft sensor And those applications even come earlier and especially when sensors are expensive or you cannot put a sensor because there’s a space limitation or it will influence the process if you put a sensor there. So those use cases are already using digital twins today in order to you know simulate or replace sensors. And this is something today, the other one maybe is a little bit more in the future.
Nick Finberg
12:12
Are there any particular industries or even regions where software-defined everything is gonna be adapted earlier than in others. There’s some industries that probably are more advanced than others.
Rainer Brehm
12:24
But it it’s I couldn’t say it’s more in process in in in process industry you have something called advanced process control. which is also AI-based. We have a great AI-based solution which is really optimizing the process. And normally in in process industry, it’s even better because you have centralized systems like a DCSTM is very much centralized It’s very much good described this kind of you know building blocks and there you can apply it quite well. So yes, there are process industry which are using it. Sometimes improvised injury it’s a little bit tricky because it’s regulated and in pharma industry you don’t want to change something if you have a validated line. So, you need to be careful to change something because there might be regulation, there is validated things. like you don’t change so quickly. You could so but the other industry, like always automotive, is front running in some areas. and we have very innovative automotive customers which are if you’re applying a lot of those topics we already discussed about. so yes, automotive We have also some very interesting customer around consumer package goods which are applying this. They have high speed machines, they have massive scale and every small optimization helps them really to get productivity. So, there is an is an area. I see currently in other areas. I mean currently semiconductor is a very interesting one because they need to scale massively. But also, aerospace and defense are areas where currently a lot of investments move in, and then they say again, if I build something new, I want to build the latest and greatest.
Nick Finberg
13:58
How does all this software-defined automation we’ve been talking about fit into the larger context of Siemens Accelerator and maybe merging products together?
Rainer Brehm
14:07
As Siemens, our goal is to combine the real and digital world. And a lot of the things I discussed about is we come from a simulation, and we go into the real production. Yeah. And this is we talked about the engineering side; we talked about op software. And so, all we want to do is Leveraging skills from the digital world and applying it to the real world and doing that very seamlessly. I think that’s very important. Number one, and that’s the reason why we fully pay into the story of that Infinity loop combining the real and digital. I think number one. Number two, we’re very excited about CMAX Accelerator, yeah, which is our it’s a digital platform, a marketplace. Where we also need to bring this into the market, but not only to very big end customers, but also to small mid-size enterprises. So, we talked about a lot of those technologies, and I think we need to find a way how we can enable small mid-sized enterprises also to buy that kind of solution out of a box. They might be for certain use cases in certain industry, but we want to make them accessible not only to large enterprises, sophisticated companies, so also to smaller ones. They also should benefit for on software-defined automation, everything which comes around it. And their accelerator is the way how you’re gonna bring it into the market and to our customers.
Nick Finberg
15:38
Why Siemens? Why is Siemens the right company to start the software-defined journey with?
Rainer Brehm
15:43
Well, you can say now two topics. You can say, why not Siemens, you know, you are the incumbent, you are the global market leader in automation, so why you should be the one you know changing it. Yeah. I I see it differently. I think it’s our obligation. we have the customers trust us that that we support them in on their journey. And it’s about us to bring the latest and greatest technology to solve customer problems, to solve real use cases. And I think we have no I think we have the domain knowledge of the customers. We understand OT mechanisms, we stand the processes of our customers, we have the domain knowledge of what they are doing and what they want to achieve. But we also have the competence around IT, we have the s the competence around industrial software, and we have the capabilities to talk, to bring these new technologies. into the area, into the realm of industry. And that is a reason why you know Nvidia is teaming up with us. The same accounts for AWS, for Microsoft, for Alibaba, and if you’re more looking into China. And I think that is that’s important because they also need want to have partners which are playing on an eye level, and you know adapting then this technology into the area of industry. Where everybody believes that industry probably is the next big thing where AI will take place.
Nick Finberg
17:15
You touched on the concern there at the beginning, but I’m curious maybe to learn a little bit more about the people on the shop floor. how are their roles going to change in the future?
Rainer Brehm
17:27
Well, you could build today already dark factories, but they are very fixed, yeah. You don’t have any flexibility. What I see, the role of the operator, it goes more into a validation than in a creation. Because still need to validate property things. Number one. Number two. Assistance gonna be there and ease up works and might be enable also less skilled workers to fulfill our task? I think that’s another topic which is important. Then you might act more from a remote topic, so you don’t need to necessarily be on site. We talked about they can remote all the containers from a central position. We have customers we want to which want to go into kind of a command center Where you also can operate maybe factories very much remotely from a central point somewhere. so that might change so you don’t need to be Always on site as close as possible to the machine. I think this is gonna be a change, so you have more flexibility in that one. And then at the end you’re always gonna have maintenance people, yeah. But maybe this unscheduled maintenance gonna be reduced because you will have more predictive maintenance, you will have more AI-based predictive maintenance, and therefore probably the call in in the midnight or is might be less because you know the system might identify itself that something goes wrong and you can do a scheduled maintenance and don’t need to, you know, have a firefighting mode. So, I think that’s that probably something which is changing. And at the end it’s necessary that we go into a learning because Also AI needs to be somehow supervised. AI needs to be trained. So, we need to get also that learning mode into the people that you know they evolve further into a, in my opinion, exciting future.
Nick Finberg
19:20
Okay. What talents, skills, roles maybe will be critical in in a software-defined production model? Um, and how will this this reshape work for you? force development.
Rainer Brehm
19:32
Well I think you know operators more become maybe orchestrator and need to understand the big picture because they need to understand maybe how things fit together and that’s an element so not only operating it but understanding the entire line the entire production flows because probably a lot of the Productivity comes not on a single point but comes in the interplay of different things. You need to find what is the root cause, how you optimize overall then again, you know, humans probably are good in problem solving. They might be assisted, yeah, but I think problem solving is important. And you do that in a collaboration, and you do that in a kind of a continuous learning mode. I think this continuous learning is a very, very important role. role. And then yeah, I think new roles will come up. You talked about digital twins, how you build digital twins in a very easy way. How you enable your data to that it can be consumed by AI, so AI engineers. you can manipulate data. I don’t say that everybody should do will do that, but maybe you need data stewards to make sure you know you’re not manipulating because Things can still be manipulated and you want to have a trustworthy operations, and therefore you might need some governance and some security around that one as well. Cybersecurity is a very important topic. I said always you need to connect, connect, connect. But with that, cybersecurity comes into play. And we need to have people also running factories, understanding the threat and how to mitigate it. under the aspect of cybersecurity. So, I think new roles come up, which are more kind of what you see today, maybe also in the IT landscape.
Nick Finberg
21:12
Awesome. Well thank you so much, Rainer. It’s always awesome talking with you.
Rainer Brehm
21:16
Thank you Nick. Thanks for having that remote connection.
Nick Finberg
21:20
And thank you so much to the audience. If you’ve listened to this far into any of our other episodes, you have a pretty good idea of what I’m about to say. But we have so many amazing guests lined up for the series, some moderated by me talking about software defined, but also conversations on AI and the comprehensive digital twin, as well as some more industry-specific conversations. Be sure to subscribe so you don’t miss them. Have a good one and we’ll see you next time.
