Digital Twins for Direct Air Capture — How TU Vienna Built a Live Research Platform with cts DataFabric

00:00:10: I'm Clemens and we are here at CTS in the office in Salzburg.

00:00:15: And today, We have our colleague from TU Vienna with us Stefan.

00:00:21: Hi Clemens!

00:00:22: ...and...we're doing an interview about what they did last two-and a half years or maybe more regarding digital twin project.. ..and we will go from Stefans Wieter into technical topic.

00:00:40: Let's introduce ourselves.

00:00:41: Yeah, thanks a lot.

00:00:43: Great to be here.

00:00:45: A couple of details about me.

00:00:46: My name is Stefan Jankovic.

00:00:47: I'm post-doctoral researcher at TUWIN.

00:00:49: I finished my dissertation a few months ago and on the topic of digital twins in industrial plant operation Together with you guys from CTS we set up our digital twin for direct air capture pilot plants.

00:01:06: We'd like talk today technically what we achieved together how your products helped us to get there, and what the next steps are.

00:01:17: Currently I'm working on various research projects mainly with a focus of process simulation, scale-up of industrial plants in technical analysis.

00:01:26: but my main focus is still digital twins so we can connect simulation models from industrial plants using these models And of course in the long term to optimize them.

00:01:42: So,

00:01:42: introduction was really interesting and I know that you not only worked alone on this topic with Digital Twin.

00:01:49: so who is working group or study group around this whole direct air capture project for Digital Twin?

00:01:56: Yes i did my PhD thesis in the research group Industrial Print Engineering and Application of Digital Methods.

00:02:03: it's a bit broad title.

00:02:06: Specifically, we're looking at various thermochemical conversion technologies mostly regarding renewable fuels production.

00:02:13: So a production of renewable synthetic natural gas, various liquid or other gaseous energy carriers.

00:02:21: but we have also expanded into carbon capture technologies such as the one that were talking about today.

00:02:26: and let's say methodologically We are always looking to combine three different approaches on the one hand the experimental research then the measurement technology needed to measure these things.

00:02:39: And also, the modeling aspect so that we actually understand on a deeper level what's happening in this processes and can afterwards replicate it as well into virtual world.

00:02:52: if you combine those three approaches exactly where do get to digital twin which is our new big topic?

00:03:02: So more

00:03:05: or less you're following a three tier model, so to say and all combined.

00:03:09: You know let's get the best out of what try to get the past out off your system Or that process around it?

00:03:14: And optimize it.

00:03:15: I think last time when we visited with you he mentioned you are Moreless now in a state where you finished The overall topic for direct air capture but now you looking into edge cases and tried to find Let's see critical values or critical procedures where the direct air capture might not be as efficient, it could.

00:03:38: I wouldn't say that we're actually quite finished with research.

00:03:43: The technology is there and works but still a lot of topics need to be investigated.

00:03:48: But you are right!

00:03:49: One important question definitely What's the operational window that the technology can operate in?

00:03:55: On one hand what's best operating point?

00:03:58: on another how do we expand this window into edge cases where the plant didn't work as well up to now.

00:04:05: Inside the plant we have a different combination of parameters, We can use somewhere let's say more energy efficient.

00:04:12: some allow us get larger amount CO₂ out of air.

00:04:16: and here is just trade-off which operating point you want actually used?

00:04:20: And in combination with our simulation models... ...we can rigorously defend these assumptions that we take for the operating points.

00:04:31: Maybe we need to go a step back and let you explain what exactly is the direct air capture

00:04:37: for?

00:04:38: What can this be used in industry?

00:04:40: or maybe not only an industry may goes into general society?

00:04:45: Yeah, gladly so.

00:04:47: In general it's a carbon capture or carbon dioxide removal refers to technologies that remove carbon dioxide as already inside atmosphere.

00:04:56: So not technologies.

00:04:57: That's mitigates the emissions but actually removed carbon dioxide is already out there.

00:05:03: There are various processes, usually with solid or liquid sorbents that selectively capture the CO-II out of the ambient air and then this CO-I can either be used for various industrial processes Or on the other hand it could also be permanently stored underground.

00:05:21: The two options in general And our technology specifically.

00:05:26: we look at a solid Sorbent DAC process so solid particles where we blow ambient air through.

00:05:32: CO₂ is selectively adsorbed at the surface of these particles and the ambient air with less or no CO₀ exits a system, then this absorbent that's full of CO₄ is transported to another chamber.

00:05:44: were than temperature has increased We add steam as a purging agent And reduce pressure.

00:05:52: All three effects help get the COₒ off again and we have a new fresh regenerated adsorbent that you can use again at the beginning of your process.

00:06:03: And, uh...the gas released from particles is then mixture over CO₂ steam.

00:06:09: We can cool it off.

00:06:10: The steam condenses becomes water with pure CO₀ stream Again for applications I mentioned before either for utilization in various industrial processes or for primary storage.

00:06:23: Okay When we now take your research project and compare it to already existing direct air capture or CO₂ absorber, CO₀ separator separators can you describe the difference?

00:06:38: Or is this maybe too much in detail which are more less under heavy restrictions.

00:06:46: In general I can say that our great advantage is that we separate these two chambers from each other.

00:06:53: so adsorption and desorption which means that we don't have a large thermal mass, that needs to heat and cool every time.

00:07:00: We switch from adsorption to regeneration or desorption.

00:07:05: So definitely the defining characteristic is simply having lower energy demand than other competing technologies.

00:07:12: Okay

00:07:13: And since energy is money Exactly That's good.

00:07:17: Yeah

00:07:17: Definitely.

00:07:19: Additionally, a large part of the energy needed for the technology can come from waste heat.

00:07:27: So we need some amount of electricity but a larger amount also comes just from low-temperature waste heat which again broadens application because you don't have to put in lots high value electricity into process.

00:07:40: Maybe I'm a bit biased as i already saw your twenty foot container and everything resides inside it.

00:07:47: regarding the current setup of your CO₂ separator, which parts were done within your research group?

00:07:55: So I think everything but... Which was done in particular by you or my colleagues.

00:08:02: To be honest it is a collaboration with an external investor and during duration of project two startups are also created that this is a cooperation for all parties.

00:08:14: My part was less on the technology side itself, but more in the digitalization and modeling sides.

00:08:20: But no basically everything from ground up is done within a research group And also of course builds upon knowledge that we had for other projects over the past ten or fifteen years.

00:08:30: Let's

00:08:31: hear your definition about Digital Twin because this isn't our very first one.

00:08:35: so you already have several Digital Twin projects.

00:08:38: So let's elaborate.

00:08:42: I wrote my PhD thesis on the topic and researched it quite a bit.

00:08:46: And found that in fifty different research papers you have hundred definitions, so its very vaguely.

00:08:53: digital twin actually is what should be capable of doing.

00:08:57: What i for myself then chose rather broad definition.

00:09:02: Digital Twin needs to have three things.

00:09:04: So there need's an actual plant that it connects to.

00:09:09: So simply a simulation model by itself is not digital twin, then this model needs to be connected with live data from the plants.

00:09:18: so its'nt enough.

00:09:19: you manually copy some exodata into simulation models and calculate afterwards.

00:09:24: thats not a digital twin.

00:09:26: I mean it depends strongly on technology whether real time or live data means millisecond resolution.

00:09:34: but for other technologies, an hour or day resolution will also be enough.

00:09:38: But still it needs to close enough the actual events that you can influence future events.

00:09:44: and finally it simply need to be able fulfill a set of requirements we said for Digital Twin.

00:09:52: so there's not one service.

00:09:54: that Digital Twin is provided doesn't have mean every Digital Twin needs help with predictive maintenance plant monitoring, or plant optimization.

00:10:04: These are all possible applications but there is not one thing that specifically defines a digital twin.

00:10:12: You mentioned the resolution within the Digital Twin.

00:10:15: so talking about milliseconds and seconds of maybe hours or longer Is your process slow process?

00:10:21: Or how would you describe it?

00:10:24: Yes I will describe it as rather slow.

00:10:26: So its a batch process.

00:10:28: And uh... One batch takes let's say from an hour to a couple of hours depending on the operating point and The main conditions that influence the process such as let's say ambient Ambient conditions, just temperature humidity also change rather slowly.

00:10:44: It's in the minute-to-hour range.

00:10:47: In our case I believe yeah Let's say quarter hour two an hour or two hours would be the derange we will looking at

00:10:54: too

00:10:55: to monitor the plant.

00:10:56: Let's say just for performance of a plant, if it is about some safety issues then we need much faster if pipe bursts or temperature rises over limit.

00:11:09: but these things are in my opinion not should be part of digital twin and actually parts directly from the control system which has simply higher demands.

00:11:22: Yeah, for safety.

00:11:25: Then the digital twin which let's say has many more services and possibilities to serve but doesn't have requirements to be up I don't know.

00:11:34: ninety nine point nine nine percent of time.

00:11:37: Okay And regarding the direct control system this was also written by your team member?

00:11:44: Oh not directly us But a partner company that actually did the plant manufacturing.

00:11:50: The plant control system was developed and then of course updated during the first tests by a partner company that did manufacturing for us.

00:12:01: Regarding cyclic times, these are also rather slow compared to other processes in other industries.

00:12:08: so is this something that benefits you?

00:12:15: I wouldn't actually say that it benefits us a lot.

00:12:17: It makes you the bit easier, we don't need to have very large computing capacity... ...that's when our models run in millisecond or second time.

00:12:27: but for us doesn't really make any difference whether cycle-time is five minutes an hour because the model usually takes somewhere between five, ten and fifteen seconds to run.

00:12:36: so anything more than this is basically the same.

00:12:39: Okay

00:12:41: Already a lot about technical stuff, but what were the requirements from the other parties?

00:12:47: From external investors maybe and also from the Techler University.

00:12:54: Oh yeah so there we had actually two let's say different wishes or directions that they wanted us to develop in On one hand point of view for plant operators.

00:13:10: For them, the digital twin was actually there to just assist in daily operation of a pilot plant.

00:13:16: To help with their operations itself.

00:13:18: but also they helped afterwards with validation or evaluation of data.

00:13:25: So for then we designed various nice looking dashboards so that you can see real time what's happening catch any problems early on and We design standardized reports after an experimental campaign.

00:13:40: They don't have to export data, manually copy paste it into the Excel sheet and do their calculations there but just click on a button.

00:13:47: they get nice couple pages long PDF report where you can see how well experiment went.

00:13:52: And of course for them prerequisites that we even got was that We had a nice centralized datastorage because... Of course we have one PLC inside this as mentioned in the pilot plan designed container.

00:14:07: We also use a couple of external measurement devices.

00:14:11: They're not bound to this pilot plant itself, but they are used for other experiments.

00:14:16: So we simply have three four five different data sources that all use different protocols That all yes save their measurements in a different way.

00:14:26: so Just to build the basis, even get a digital twin.

00:14:30: We needed to have a nice centralized way with consistent timestamps and tag names just so we had everything in one spot.

00:14:39: And building up from that then as I said, we did dashboards reporting data export for further analysis.

00:14:48: On other hand, on this side of investor is long-term project development.

00:14:55: There we looked at possibilities how these twins could actually benefit the operation in the long term.

00:15:01: So, um... We wanted to see How can it optimize plant performance?

00:15:07: And maybe help people who don't understand technology as deeply as we do with troubleshooting or maybe even allow us to help them remotely with the problems they have.

00:15:17: So this is more, we only partly implemented that in our part of plan but it was a demo or proof-of-concept showcase that will be presented then also for investors.

00:15:31: so all those prerequisites and requirements pushed into direction of emission I would say.

00:15:38: But was this also the decision path for you, or how was this whole process around choosing the right tool?

00:15:45: For a job.

00:15:46: And did you have to more-or less argue with your investors and with Technical University... ...to use this one?

00:15:54: How were these settings around it?

00:15:57: Luckily no!

00:15:58: Luckily we had big freedom from one side of the project investors but also our supervisors at Technical University.

00:16:07: We have, let's say a research goal that we try to reach but it is our own choice which tools we use concretely.

00:16:15: To be honest I hadn't heard of InMation up until the moment when they started using it.

00:16:20: A colleague of mine then heard about our use case for problem and mentioned because he could try using this.

00:16:28: It was perfect fit from first moments Because actually did what you didn't want.

00:16:36: It's not so difficult to click on export and then get some Excel data, but it still just boring annoying.

00:16:44: And there is also again another source for errors because you know... Some of these are trivial things like measurement devices use the local time zone or some use UTC.

00:16:53: You probably never remember which is which Just spend your time on useless thing that can be spent more important questions in our research.

00:17:02: So this where innovation definitely helped us.

00:17:05: And no, it was a pretty straightforward process.

00:17:08: So after the colleague recommended this software to me we contacted you and say in couple of weeks I had prototype running saw that worked just started building on from there never looked back.

00:17:22: luckily if now compare let's say starting phase with information knowing even the word or company around it, and compared to your previously done digital twins.

00:17:35: What are their biggest differences here?

00:17:39: So yeah for one use case we had already developed a Digital Twin that was different project where looked at conversion of wood waste with other waste materials into synthetic natural gas.

00:17:54: And through this we develop a Digital twin based on a Microsoft Azure cloud solution.

00:18:00: To my opinion, it was also a really great successor project.

00:18:03: We got where we wanted and managed to connect again simulation models with the system.

00:18:09: You could calculate variables that couldn't be measured before or at least couldn't been measured in real time Before... ...we manage to add a multi-projective controller so the plant itself did.

00:18:22: the whole system definitely worked.

00:18:24: But we saw a couple of downsides with the same Microsoft Azure Cloud solution, not specifically with the Azure Cloud Solution but in general with let's say self-done and self-programmed solutions.

00:18:36: There was definitely lot of overhead so things that our minds were actually would be just trivial like creating some dashboards or simple graphs on input fields suddenly required actual programming work expertise that we didn't have in-house, that luckily project partners had.

00:18:56: But still it's just another layer you get through and the hoop to jump too.

00:19:02: so we did not have all of these configuration possibilities on our hands.

00:19:07: It took definitely longer In this case let us say a month before we got first data from our plant into cloud then dashboard or simulation models.

00:19:21: These are things that we managed with Inmation in, let's say a couple of days.

00:19:27: Simply huge time difference.

00:19:29: and the other big thing is simply need less expertise.

00:19:34: I can train a colleague at least basics of Inmations day or two but definitely won't get cloud programmer that knows all of MS Azure's tools in this time frame.

00:19:46: So I wouldn't want to only say negative things about the cloud solution, we definitely had some...I would say it is even more open and a bit more freer.

00:19:56: In Mation you do have let us say some framework if you need follow but yeah If you get used to it then I wouldn´t say that it has ever actually hindered us from getting those tools as much as possible would also possibly speak a bit more in the direction of the cloud solution, or it is in general to self-programmed solutions.

00:20:17: It's that they have lower license fees and less vendor lock-in.

00:20:21: but our case we just did the math with tradeoffs and chose the animation software simply because... ...it saved us so much time and headache let's say!

00:20:33: And this works out as a box.

00:20:34: you don't need to do a lot of troubleshooting.

00:20:37: Your curtain says?

00:20:38: Thanks.

00:20:39: I'm really interested in this whole, let's say cloud situation and also Microsoft think since a few years to have this.

00:20:45: it's not data fabric but is called Microsoft Fabric so probably an evolution of what maybe you use back then.

00:20:51: possibly that

00:20:52: didn't had your hands on this?

00:20:53: No we haven't Back.

00:20:54: Then We used the Microsoft IoT Hub Let's Say as a central connection node for everything.

00:21:00: Possibly it's easier nowadays, but no we haven't had experience with that concrete software solution.

00:21:06: So after those months integrating the cloud solutions and having let us say a starting phase where you personally was involved in two or three month if compared this one now to Emation and said In Lua development you probably have more coding involved, let's say coding analogies.

00:21:27: Maybe using JSON which wasn't maybe known before.

00:21:31: and how was this for you?

00:21:34: Yeah if I say it is a bit of roller-coaster of emotions at the beginning.

00:21:38: Yeah, roller coaster

00:21:39: of emotions yeah!

00:21:40: At Lua

00:21:40: its

00:21:42: hard to learn in

00:21:44: the beginning... Definitely not only Lua but in the begining the emotion really great.

00:21:48: everything worked.

00:21:48: You had nice green light data coming But then when you realistically try to get some more complex tools, it's not trivial.

00:21:59: It does take time getting into but after out of the low finally understand what we're actually doing and how individual objects work with this object is possible.

00:22:11: that also helps us a lot for that.

00:22:15: You are welcome!

00:22:18: When I actually had the feeling that they understood how this software works, then it was really great again because yeah.

00:22:23: Yeah i knew what the possibilities were and limitations where And yeah i mean Lua development is not easy.

00:22:29: but to be honest Im also a real programmer.

00:22:31: I understand how code work so can do some simple coding tasks.

00:22:35: But you know there's just many AI tools out here.

00:22:40: We're not building any mission critical tool or safety critical tools.

00:22:45: So Basically, nowadays we use AI tools to program the code snippets for us.

00:22:52: We test them and then we use it.

00:22:55: So definitely I wouldn't say that i ever became a professional but an advanced user And now there is in my opinion still only scratch surface and use probably ten-twenty percent of available tools inside the software, but they work for us and we're really happy with them.

00:23:16: I need to contact you much less now.

00:23:18: so it gave me a great running start!

00:23:21: And am able do most of the troubleshooting or new service development myself?

00:23:28: How many hours are put in a week on average?

00:23:33: learn and get yourself running in this environment.

00:23:35: because when I think back to my days, When i started reformation.I was totally new to OPC pretty well understood the network topologies And how things are connected on a physical basis.

00:23:47: so this helped me lot.

00:23:48: but having access To the physical equipment eases things up.

00:23:52: if you understand them You understand the bigger picture.

00:23:57: How was this for you and how much effort did it put in to get on the level where we are now?

00:24:02: It's difficult, I have to say.

00:24:06: My starting point is pretty similar to yours.

00:24:09: so yeah...I knew some basics of network technology but had absolutely no idea about OPC UA or other MQTT.

00:24:17: mod was nothing that only came with time.

00:24:21: at the beginning i would say I had to put in probably two, three or four months really dive into it and during this time i tried mainly the animation programming work.

00:24:34: so let's say twenty-thirty hours a week.

00:24:36: This is also that we actually built and developed tools which were using now.

00:24:41: afterwards.

00:24:42: doing next phase was more troubleshooting adding of individual components.

00:24:49: nowadays maybe dedicate half a day or week to that, if even sometimes less.

00:24:55: On the one hand things work easier now because I know what i'm doing and on the other hand simply we have most of the tools that you need.

00:25:05: my colleagues which use it everyday they say used basic services available in Data Studio and Web Studio but don't develop any new tools right now.

00:25:17: You previously said it's rather easy to onboard a new employee, your emmation system or whole project.

00:25:24: And I know you're already crafting in the background educational program.

00:25:29: how is this in combination working?

00:25:32: and yeah do you think about bringing more people into their projects?

00:25:35: Or are these small as right now covered so that they can look forward for new projects?

00:25:40: So let us say this project was discovered by staffs we have But we're using the software now also for completely different pilot plans.

00:25:51: So again, as I said this is a renewable fuels sector and here it simply see that its intuitive user interface all looks like normal Windows program.

00:26:05: you have lots of drag-and-drop right click context menu.

00:26:08: everything just more or less works as expected.

00:26:12: so at least with basic tasks such creating diagrams, looking up values.

00:26:21: It's very quick to onboard someone to innovation.

00:26:24: if we will go a bit deeper into as I said developing of tools and reporting items it would take longer but still shorter than me at the beginning because simply now there is one expert in-house that can guide next generation or how you want to call.

00:26:43: But yeah, it's definitely much simpler than having to explain the PLC programming or then actually design visualizations on a PLC.

00:26:54: This is more intuitive in animation.

00:26:57: Also also difficult phase came from that not only we had to develop our own tools but I have feeling for some cases using the software edge of what was designed.

00:27:09: So I'm pretty sure it was not designed to directly communicate with simulation models and use that data as input data.

00:27:17: Or let's say we pretended some sort of sensor data, but even in those cases It worked well robustly And yeah very happy about that.

00:27:28: Just maybe add that Why?

00:27:34: also the hurdle to get into using Emationals may be a bit higher for us Because you don't just use it further than standard as I said, data connectivity logging and things like that.

00:27:45: But yeah try to add additional maybe unusual tools.

00:27:50: So um... As far as i know you have other third party applications so also MATLAB?

00:27:56: I think using MATLABB to calculate the models.

00:27:58: I guess

00:27:59: Exactly!

00:28:00: We had two different layers of modeling.

00:28:04: we have let's say reactor scale model where we actually look what's happening inside the reactor and was also happening on the particle itself.

00:28:14: This is done with say physics-based equations, some adaptation parameters that we got from experiments And this all programmed in MATLAB.

00:28:28: more plant level, we look at the entire plants including both reactors for adsorption and desorption.

00:28:34: Including all of the peripheries such as the vacuum pump steam generator things like that.

00:28:40: And this is implemented in Ipsu Pro.

00:28:42: it's a commercially available simulation software where We use custom our own model libraries the measurement data from the plant, then pass it on first to the MATLAB model and get results there.

00:29:05: Then pass again on measurement data combined with MATLABA results through the simulation model that actually runs for example the plant optimization algorithm.

00:29:17: Like you said this is already things that experienced users do and not using the standard configuration objects of information, just let's say pushing data to the cloud or other easy tasks.

00:29:28: So this is really more sophisticated, deeply involved into your processes

00:29:34: but still it worked well.

00:29:36: And if we had to maybe roll this out at large scale, maybe would then implement a simulation model in the cloud with some API endpoint that you could just send data into and get your simulation results back.

00:29:50: But yeah, our case... We actually like everything local close to us easily manageable because it changes all of time!

00:29:59: It's still a pilot plan For us, it's not interesting to run the plan at a standard operating point for eight thousand hours per year.

00:30:08: We want to get to the edge trying without new things and see how these changes impact our results.

00:30:14: Is

00:30:14: this also something where you choose Ipset Pro or what has been the decision path here?

00:30:19: Definitely yes!

00:30:21: In comparison with other simulation tools in the market... ...Ipset pro maybe has less capabilities built-in but it has a very open model development possibility.

00:30:37: So over the past, I don't know, fifteen or twenty years inside our research group we have developed models for this and can adapt them exactly to our needs.

00:30:46: And additionally YPSU Pro also has the so-called validation mode.

00:30:51: that helps us a lot where we cannot only calculate let's say defined system of equations only calculate one operating point inside the plant.

00:31:03: But we can use a sensor data that we have, put it into the model and then have an over-determined system of equations which means that you can consolidate different measurements.

00:31:16: We could maybe reduce the measurement errors that we had simply by having redundant measurements for the same thing.

00:31:22: And in this simulation model itself calculating a balanced operating points where Yeah, this is actually then calculated.

00:31:32: The best fit so to say?

00:31:33: Yes it fits the best!

00:31:35: If you come back with a very purpose of your project and how inMation got into this.

00:31:42: if try to describe from A to Z what InMation does maybe especially good or not-so-good InMations are capable And possible feature requests where we think that would make my life much easier.

00:31:57: Let's discuss on that.

00:31:59: So basically, the main thing why we even chose to get a software was that we have couple of different PLCs and different measurement devices that are not integrated in the PLC that you wanted it combine.

00:32:14: just for us as a single point-of-truth We wanted to reduce some manual tasks such as exporting data from different measurement devices and combining them into one extra sheet, then having to work with that.

00:32:32: So the main thing really is that InMation can connect all of these disparate data sources and store them in a centralized way so we could use this data for our different tools and applications.

00:32:46: From there on they just started building tried-different different tools, services such as dashboards visualizations for our plant operators directly during the operation itself.

00:33:00: Then after finishing specific experiments we added a reporting tool so that when experiment is finished after seven or eight hours The operators can really quickly generate a, let's say five to seven page PDF report where they could see the most important metrics.

00:33:19: KPI is a couple of graphs.

00:33:22: This has been undertaken with this report generator animation and conditionally we would look at it further.

00:33:30: not only what can help us right now during pilot plant operation but how should be integrated into larger scale finished product that actually in market.

00:33:43: And here we're looking more at things such as real-time optimization.

00:33:49: So how can the operation of a plant be adapted to changing environmental conditions or maybe changing price conditions on the electricity market?

00:34:00: Additionally, if you are able to gain data from lots of plants that out there would also look into applications like predictive maintenance possibly even say try to help our customers with remote access, remote maintenance if they see that their plan is not running as it should.

00:34:29: So combined everything you said now powered or covered by Mation for instance we go into this topic of reporting dashboarding.

00:34:39: Is there something more less directly connected to your stakeholders?

00:34:44: or is this maybe just for the technical sake, how's that?

00:34:51: I would say both.

00:34:53: The dashboards are actually just for us during plant operation but the reports themselves really used when also presenting their results to the stakeholders.

00:35:03: and of course it will be possible even without animation which was more complex in my opinion useless additional effort that can be spent differently.

00:35:16: But yeah, we don't have a specific way that was prescribed how to do it but they're really happy with the way it looks now.

00:35:24: because I mean yes It's not just about experimental data About insides of big game and also has look nice at end-of-the day and then be nice to use.

00:35:34: When I think back our, let's say closer collaboration where we had multiple meetings in a month get you on your team running how connect PLC?

00:35:44: because do have multiple vendor PLCs not only Siemens as i remember the two other ones?

00:35:53: To be honest can't remember right now.

00:35:55: but yeah with multiple PLCs.

00:35:57: also these external measurement devices which I don't know, some use MQTTs and Modbus.

00:36:04: It's just a bit chaotic because again if we have a measurement device from twenty years ago that still works by new one but it will be used old even though it doesn't have OPC UA.

00:36:15: But this is

00:36:15: the typical brownfield that's more or less the same in the industry.

00:36:19: so yeah... We really wanted to integrate all of these completely different data sources.

00:36:27: It just works.

00:36:28: And it's really nice because now afterwards, if we want to have a new service... ...it is either quick configuration in Inmation itself or call the Web API and then we can have another service running for couple of hours, couple days.

00:36:44: I will start over with my question because Moly has interrupted myself when asked

00:36:50: about

00:36:50: vendors.

00:36:53: So our collaboration When we worked closer together to get you running in reporting, to get your running in Web Studio dashboards and also connecting different third-party tools.

00:37:05: And how to interact with maybe MATLAB or other applications.

00:37:10: for me personally this was a new experience because most of our customers don't have the time even though they may be equal to your skill level but their lack be that dedicated to an animation project.

00:37:28: So, for me it was a very refreshing experience because I saw you more or less gaining your experience and skill by skill in the short period of time giving some bits & pieces And after two weeks Stefan came back saying oh yeah look i build this and then i need your aid here Cool.

00:37:54: And this was something that I experienced never before from the other customers, maybe a bit but not in that let's say and this mature setup where we have really complex components working together.

00:38:08: so... From my integrated view it is just more or less outstanding.

00:38:15: Thank you very much.

00:38:18: Even you didn't have a programmer background, so technical backgrounds and more or less love to solve problems.

00:38:29: I wouldn't say that it's really just a plug-and-play game.

00:38:35: but if we put in the work for two or three months with more intense work within Mation you can definitely gain very good understanding of it.

00:38:47: And it was worth a lot to get some insights from you, some code examples.

00:38:52: But for us it's simply very interesting to have an expert, let say directly on site or really close with the plant.

00:39:00: Because in discussion of colleagues then new ideas and use cases pop up And I can at least get a gut feeling whether its possible do that or not?

00:39:10: And i could really whip-up a prototype within couple hours if necessary.

00:39:14: If not just leave you message!

00:39:17: And gain some help.

00:39:19: But yeah, if you put in the time it's not let us say a matter of years until we can use that at really high level.

00:39:31: And I would love to see these views.

00:39:36: Have you worked during the prototype with air capture separator?

00:39:41: have your work with other integrators or other suppliers too Solve problems

00:39:50: to be honest.

00:39:50: No, I mean we.

00:39:52: We had a contact with Sergio.

00:39:55: It sounds good because basically let's just try it and see how it goes.

00:39:58: And basically everything went.

00:40:00: the went very well from beginning on so he didn't have that they need to do somewhere else But I just know it from different pilot plants that we have.

00:40:10: That is Just a bit annoying Because it's not our central expertise.

00:40:16: So usually it's just some quick and dirty fix where we either started data directly on the PLC, then just export it as a CSV.

00:40:23: Then work with its offline but no... We actually didn't really ever look deeper into this topic.

00:40:37: What you could say is that your very first prerequisite was more or less to your success?

00:40:49: Definitely, yes.

00:40:51: We have these various external measurement devices that we use for different plants as well and we just want one simple nice way where all our data is in space.

00:41:01: but you don't need think about the tags or timestamps.

00:41:07: it should work basically.

00:41:09: so be safe from any manual user error.

00:41:13: And also on the underhand, we want something that's easy to be backed up.

00:41:18: So yeah it is really no big deal just regularly back-up our measurement data from inMation and I wouldn't say this one great service made us choose InMation but its a prerequisite for any of other more advanced between services than we then adapted.

00:41:43: Just because of this thought, I don't know if it came from you or one of the stakeholders.

00:41:48: A colleague?

00:41:48: Or whoever but starting with these and positively ending up.

00:41:54: there is a pretty cool journey!

00:42:00: The great thing was that's also what i tried to emphasize in my dissertation... ...is even between projects.. You need start somewhere and usually start small just be able prove that it even makes sense to do.

00:42:16: So you choose some small part of it, gather the required data and develop a service.

00:42:22: And you should have at least some sort result within couple months so you can prove your sponsors this actually works.

00:42:30: Then afterwards you add on or scale different services.

00:42:35: but is exactly something which work really well with Inmation because if today there's already I wouldn't say trivial, but it's really very easy just to expand and make dashboards a bit nicer.

00:42:48: Make the reports a little more extensive add new simulation software for the mix something that we value.

00:42:53: I

00:42:54: would like try have forecast from you regarding your upcoming project mentioned at the beginning maybe other than form of your position where executive researcher If you plan to use delimation in the new semester, what educational purposes do try to aim for?

00:43:25: Directly within the redirect air capture project we want to expand the use cases I mentioned quickly at the beginning.

00:43:34: We wanted to run a plant with different operating points depending on external conditions.

00:43:40: These are on the one hand environmental conditions such as ambient air temperature and humidity.

00:43:46: And, on the other hand prices such as electricity prices in the market.

00:43:51: Depending these conditions we want to then find optimal operating points either for CO₂ amount or energy efficiency... ...or let's say long term than actual cost of CO₀ captured.

00:44:05: This would be first at a local level unit of one plant.

00:44:12: But at least in my dreams, what it could then be is that we directly sell the digital twin as some kind of service such a software-as-a-service just Digital Twin as a service As sort of add on to their physical plant itself.

00:44:28: That we then aid operators either with troubleshooting if something's not working With predictive maintenance applications or exactly with these optimization measures, depending on their location where they put the plant up.

00:44:43: So we could say that a small server unit of PC is being shipped within the twenty-foot container running an Emation Core service if you assume version one.

00:44:57: otherwise it would be a node as we learned.

00:45:00: so... We have an Emasion Core then for every container and Yeah, running everything locally without any online connectivity needed.

00:45:10: That would be my dream.

00:45:11: yeah I wouldn't even say with out any connectivity needed but it should be able to work without connectivity.

00:45:16: But then again if its were connected the support portfolio could of course be even broader Because on one hand we can use data from the plant and improve our own models And on the other hand, then we could use our models to actually improve operation of plants wherever they are in world.

00:45:38: So that would definitely be my long-term dream for this project.

00:45:43: Then within our research group will continue using it for various pilot plant applications Again two different types.

00:45:53: On the one hand, to support our operators really doing experiments in our lab.

00:45:57: Again regarding visualization reporting things like that and on the other hand we're always looking towards large scale up towards industrial plants.

00:46:07: so actually show digital twins make sense.

00:46:13: We always have to think about that and create some case studies, proof of concepts.

00:46:18: How can a digital twin be used to improve a large scale plan that runs for eight thousand hours per year?

00:46:26: And to be honest, I mean... ...I like Invasion very much.

00:46:30: So if it were up to me we would definitely continue using it.

00:46:34: We'd love to use it and hope that you'll continue our cooperation.

00:46:39: For your last question i'm also preparing a course on Digital Twins which should then be offered within within the chemical engineering studies at Tailveen.

00:46:57: In that course I would like to present some general definitions, what is a digital twin?

00:47:02: What can it do and how's it used?

00:47:05: simply because this word is a buzzword everywhere but not everyone understands something differently under that term.

00:47:13: so yeah we'd like to represent to students exactly what i believe in these terms means.

00:47:19: then really do some hands-on exercises.

00:47:21: So we already developed a model that would then serve as a virtual plant.

00:47:28: It runs in real time, gives out operational data over OPC UA and I'd like the students to connect with this data source get their data stored prepare some visualizations maybe even connected into simple simulation models so they can get feeling how workflow there actually is.

00:47:49: What would await them if they chose this career path for their future?

00:47:53: So the students can see theoretical side, have hands-on experience using an industry software.

00:48:02: And additionally you'll benefit to actually seeing a plant even though it's just small pilot plants but... It is physical world so to say and they combine everything together.

00:48:14: Exactly, yeah.

00:48:15: Because it's still very abstract.

00:48:17: I know during my studies there was something that i believe is a bit missing from because you always hear about some large numbers You learn all the theory but rarely get to touch actually or just connect two cables together And this is something that I would like to, at least offer them the opportunity.

00:48:38: Ideally with a software which they already know very well and want to use... ...and it's something you might get into contact if working in an industry.

00:48:47: Sounds promising!

00:48:49: Very interesting yeah?

00:48:51: If i think about my last question here Is there anything we have missed maybe?

00:48:58: Thinking of these questions I know that the Technical University in Vienna also holds license agreements with Espentec regarding other software components.

00:49:10: Let's put it this way, is there something you consider within your study or maybe an add-on to have another Espentech product linked into your emmation setup now?

00:49:23: We will definitely look at it.

00:49:24: up until now we've used mostly Matlab and YPC Pro as before But there are use cases where, I don't know, Aspen plants and Aspen hyacis would definitely be the better choice simply because they have better models for some processes.

00:49:40: Better thermodynamic property data.

00:49:42: so yes it's definitely on our list.

00:49:45: let say if you write a project with correct interesting process chain comes along.

00:49:52: Very interesting because I personally thought also the director of as the crisis and I honestly never had my hands on this but, uh i know from other customers in industry that it's a very well known and very very very properly working software.

00:50:07: Yeah so I thought maybe this could be fit for you?

00:50:12: No definitely we have the license through the academic program And they're using it, let's say partly for master students.

00:50:25: Trying to gain expertise in this direction and then at some point also really connected with life plans data as our simulation models as well.

00:50:33: This already drives us into the direction of what will you do if your new role?

00:50:38: So I think nearly half a year now that you are an executive researcher... What are their next steps personally?

00:50:45: or I'm not wanting to say, we had to use your selfie.

00:50:52: Going a bit into this direction what would be the next projects for you personally?

00:50:58: Well yeah... To be honest now as a postdoc it's required of me that actually build the NGT partnerships gain new project Be they really contract research work or or for example, projects still coming in through public funding.

00:51:18: So I will definitely put my focus on getting new project in.

00:51:25: Personally yeah...I would love to stay in the let's say general digital twin space.

00:51:31: somehow I also really like the State University and enjoy the academic freedom i have here.

00:51:38: And to be honest, time will tell.

00:51:41: So I think then we should look into this camera or that and say as a tech community... We'll have some updates for you soon!

00:51:50: If you want so!

00:51:52: Gladly yeah?

00:51:54: Yeah i really thinking of putting your key take away also taking maybe an inspirational push from the professional side To share some insights which can be shared with a broader community.

00:52:08: and if obviously it's possible saying, yeah, CTS was supporting here quite well.

00:52:13: I would be happy to hear that but in any other case i'm also happy to have you more or less here on board and continue our collaboration in the future.

00:52:22: Thankyou very much!

00:52:22: It is really really happy working with you.

00:52:25: everything works really well And...yeah..I'd love to continue and even expand our work together.

00:52:31: thank you very much.

00:52:32: Thanks thanks too.

00:52:34: We appreciate.