Join the Insider! Subscribe today to receive our weekly insights
Willem (00:00)
Hi David.
David (00:01)
Hi, Willem
Willem (00:03)
We’re back and this time we have two guests, not one, both from AVEVA. have Roberto Serrano Hernandez and Clemens Schönlein.
David (00:15)
Welcome gentlemen.
Clemens Schönlein (00:17)
Hello, nice to be here.
Roberto Serrano Hernández (00:18)
Thank you.
David (00:19)
Hey, and Willem, I have to say we are amongst friends here today because I’ve been working with Clemens and Roberto for a while now, so I’m super happy that they join our call. They are both technology evangelists at AVEVA. Roberto knows everything there is to know about the CONNECT Data Platform, and Clemens is deep into data analytics and AI. So guys, let’s dive in. Maybe Roberto, first to you.
who are you and what is CONNECT?
Roberto Serrano Hernández (00:50)
Yes. Thank you, David and William, first of all, for having us here. As you said, Roberto Serrano Hernandez I’ve been with the company working in industrial data for around seven years now. I started in 2018 within this journey. Funny thing, I started in tech support, so I started seeing everything that doesn’t work and seeing customers unhappy coding because they were having urgencies and problems. And then I transitioned into a different role, more of a consultative one.
based in pre-sales, right? And identifying which are the use cases that customers want to achieve, based on that legitimacy and realism that I got from tech support, knowing what works, what doesn’t work, and where are sort of the pitfalls. And because of my curious nature, this goes all the way into the implementation, and because of that technical background that I have, I’ll…
often get involved also during implementation. So I get to see the final results the customers are doing. And that’s where I cross paths also with Clemens, because I think Clemens, you work more on the user side of the things.
Clemens Schönlein (01:54)
Yeah, exactly. Nice. Awesome to be here. Clement Schönlein, my name. I’m now 10 years in industry. I started as an engineer system simulation guy and then I had a nice turn towards the reliability space, doing basically risk-based maintenance planning. And then this topic of predictive maintenance popped up in the domain, which made me a data scientist.
playing with Python. So I consider myself as like the hybrid guy, you know, in the past, I came with my Python scripts, asked the customers to give me the data they have in their OSI soft PIE system back then, most of them, but also work together with IT colleagues that can go a lot deeper into the data science aspects back then, focusing also on the business challenge side, like what to actually solve.
David (02:24)
You
Clemens Schönlein (02:52)
And yeah, within AVEVA, I’m now two and half years in the role of a solution consultant for AI and analytics and evangelizing around the topic.
David (03:01)
and we’re just under three minutes in the conversation and we already dropped Python. So it’s going to be that type of conversation, I guess. that’s awesome. That’s awesome. So to start, let us, I would say, introduce the CONNECT platform, maybe also link that a bit too, because what Clemens said, OSIsoft, it’s also OSIsoft PI, which is still a big name, I guess.
Clemens Schönlein (03:07)
Yeah.
David (03:28)
in the industry. So could you, would say, briefly share that history, Roberto, and then also how that leads into CONNECT.
Roberto Serrano Hernández (03:36)
Yeah, for sure. So something to mention too that I forgot to say is that today I’m a member of the CONNECT Incubation team in the Lead Technology Evangelist for the Industrial Data Platform. So that’s where things converge. And there are two ways of seeing this. There’s the first way for people who know nothing about AVEVA, about the company, or about CONNECT and this sort of solution. First thing is what is CONNECT? So CONNECT is a neutral and open Industrial Data Platform or Industrial Intelligence Platform, as we say, where we put together
all of the data coming from the industrial scope, coming from solutions from AVEVA, but not only. And the goal is to unlock the value of that data and let companies enhance their ability to collaborate with their broader ecosystem. The thing that where the OSI software system matches very well the story of CONNECT is that at the end of the day, we are serving industrial companies.
So there is physical assets behind that. And there’s things that really have a footprint on the shop floor. So the goal is how to unlock that data by bringing the critical on-prem systems that need to keep those things running, those factories producing and creating energy for us to be speaking here right now. And how to drive that all the way to the cloud, where today we have technology evolution, user evolution, education evolution.
You cannot just expect everybody to be happy working on the shop floor. And now the tools and the sort of things that we want to do with data today don’t really have a good fit for what you can do with on-prem systems and with the Pi system. Let me tell you, the Pi system is super powerful. And it’s oftentimes enough for, I would say, 80 to 90 % of our customers, right? But if you want to go that additional 10 % and close it up and do everything that’s available to you, and more importantly,
be prepared for the future where new generations, where new tools will require those skills, those needs and those features, you need to be able to move to the cloud. And that’s where CONNECT Closest they look very, very well and that allows for that swift transition from the Pi system, but not only, and to have that industrial data set, rich, contextualized, ready to share with those additional use cases and new users.
David (05:59)
If we link that a bit to our capability map, so where we go from, I would say, connect, contextualize data quality, the data store, analytics, and then also visualization and sharing, if we specifically focus on connect now, like where does connect play a role? Where does it maybe still rely on on-prem components, et cetera?
Roberto Serrano Hernández (06:25)
So I would say that CONNECT has a footprint a little bit everywhere. And we’ve been experts in industrial data for a number of years. So I would say that the core capability is going to be the data management there, data storage, and data delivery. That’s going to be the main capability. Now, for you to be able to serve rich data and create those contexts, you need to have that connectivity. And because we have all of these legacy experience in industrial protocols, industrial information, not only 10 series data.
but also the data that comes from engineering tools that we use to have also within AVEVA, we have a huge heritage there. We’re able to bring the data in. So I would say we have a very strong component on data acquisition. We have a very strong component on data storage and data delivery. And then everything that goes in between, we are going to have…
Oftentimes, service tools for those guys that want to have that single platform for that simple use case that they want to achieve very easily. And we are going to rely on a huge network of partners. And I think you know a little bit about that, David and William, right? To bring those very specific use cases that some of our customers in industry really need.
David (07:35)
and to, I would say, link that into Clemens’ story because, well, I have a data science background myself as well, so I know a bit about the pain. how does a data engineer or a data science profile interact with CONNECT?
Clemens Schönlein (08:01)
Yeah, so from that perspective, connect can be seen as the interface from the OT domain, so sensor data into the IT domain. And that’s also what we do with connect and also the connector to data bricks to the data bricks platform, for example, to target exactly this kind of users with this skills and needs to access that kind of data.
And maybe to add onto that, because from AVEVA’s point of view, there’s not only analytics, let’s say, or we do not only serve the IT domain with data for the analytics, but there’s also rich capabilities as part of CONNECT that are more targeted to the SME subject matter expert on the process side. Because in the end, and that’s also my experience before AVEVA doing these kind of projects is,
it doesn’t stop with the analytics, right? It needs to, all these insights need to go back into the process and improvement that change has to happen. And I think that’s also the advantage of like where connect comes from or the backbone of it, that we also have things in place to get the information back to the shop floor.
Willem (09:19)
Now that we’re talking about making those examples coming back to the shop floor, you guys also brought an example to make this a bit more concrete.
Roberto Serrano Hernández (09:31)
Yes, I can go with it. The example that we’re sharing, by the way, this is a public story. This is just one among many, but this one is published and we have the nice light to decorate the story. This is Amcor. Amcor is a packaging manufacturer. They produce packaging, rigid and flexible for when you buy your lace chips and that sort of thing, or when you buy your blisters of whatever you’re going to eat tonight. That’s they produce.
David (10:00)
Okay.
Roberto Serrano Hernández (10:02)
The things that they have been working with AVEVA for some time, and they wanted, when they started the project, they wanted to use CONNECT and this new set of technologies to try and bring an additional dimension and additional capability to better interact with the data. At the end, it all comes down, oftentimes, to make better informed decisions, right? As Clemens was saying, it’s not just about having the analytics engine and doing, no, it’s not just about the recommendation. If you don’t act upon that recommendation, then it’s worth not much, right?
At the end of day, they wanted to have all of those sources of data that were being used on the shop floor together into one single pane of land so that they could drive this better decision making. And what happens is that what they did is they’re bringing that MES data with historian data to see them at the same time within CONNECT Visualization Services, which is one of the tools for visualization that we have within CONNECT. And together with that, they were able to use additional
advanced analytics tools that we have within CONNECT. So it’s not just as Clemens was saying, feeding data to database where we have a huge set of users that might be willing to get a on the data. We also have these self-service tools that I think are essential for industrial operators to be able to leverage these new technologies with all of the data that they have from years. And what we’re able to do is that
they knew that they were having stoppages, downtime in the machines, but sometimes, you know, as I do, and David, whether you have a background also working in industrial situations, that sometimes it’s a little bit obfuscated. Why didn’t a machine stop? What was happening here with Amcor is that they were having a chained process in which there was one parameter drifting, namely the viewpoint of one of the machines in the sequence of what they were doing. And at the end of the day,
the viscosity of the plastic that they were putting into the molder was not good enough for the molder to work. So that was a problem from the preparation of that plastic that comes from steps before. And what you see is that there’s another machine that’s jumping, but you don’t know why. So they were able to use this advanced analytics engine to find those correlations and then do the mapping between the right set point that you need to put for that specific parameter that you want to monitor. In this case, the…
the viscosity of the plastic that comes from machines before the one that you’re focusing on. And then they were able to close the loop and understand that, hey, this is where I need to act if I want to prevent this downtime. And that was saving three, I think, the figures were there in the story. So they were saving around $3,000 US dollars per hour of downtime. And they have like 160 assets of this one. So you can imagine the amount of savings that can be potentially
unfolded with this, right? And it was done in lightning speed because the data was there, the connection to connect was almost immediate. And then all of these cell service tools allow for those outcomes to be just real and achievable for everybody.
Clemens Schönlein (13:09)
And then for me, if I may add to this, because in my past I did similar projects and what is really impressive here with this technology in that case is the speed. Like how quickly we can close the gap if you want to from the data to the analytics, getting to these insights. Like in my past, I was working two, three months just to get the data and then going through Excel sheets,
David (13:35)
Yeah.
Clemens Schönlein (13:38)
talking with the guy from the automation system, which tech is what, and then finding out there was a gap in the data set and all these things. And that’s what you can basically do with CONNECT now in just a couple of seconds, really, have all this at your fingertips.
David (13:53)
It’s recognizable. You know what I also like and especially for this, again, for this case study you mentioned a bag for bag you use for chips. know, the cool thing from my perspective, working with manufacturing is every time I’m having dinner with my kids in the evening, I sometimes say like, I don’t know, I have a bottle of milk or yogurt or you know, whatever, or like the chips in this case or the ketchup or whatever. And I sometimes say to my kids like, see, so this bottle, daddy didn’t make the bottle.
But at least we are at say the software, know, the software we are selling or supporting or configuring or installing has been used for that.
Clemens Schönlein (14:25)
Yeah.
Willem (14:34)
Is that when they start rolling
their eyes, David? Like, that is sick.
Clemens Schönlein (14:37)
Hahaha
David (14:38)
No, no,
no, they still admire me. Maybe…
Willem (14:40)
They’re still young enough. With me, they reach the age they don’t care. It’s like, I’m not
Roberto Serrano Hernández (14:44)
Thank
Willem (14:48)
a pop star or something like that. Clements, think going into this use case where you mentioned that in the past, as a data scientist, you wasted a lot of time finding data. I think it’s clear those data platforms aim to solve that problem. Of course, you will have new bottlenecks. What I’m seeing out there is that
There’s like two different approaches to data science and analytic solutions. And one of them is the one that you mentioned, quite focused, quite detailed on one very specific problem that might be multiplicative, like you said, with a lot of relatively identical machines. The other one is let’s try to make something more generic that
everybody can use and we can show that we implemented analytics in all our plans. So what are your experiences with both approaches and how would you handle those?
Clemens Schönlein (15:51)
Yeah, I love that you bring up this point. So the philosophy that I believe in is that you need to think in scalability at some point. Like these projects typically take quite some resources to generate the analytics, understand the asset and so on, even if you have a great platform and you have contextualized data.
you still need to put brain power into that. So you need to reach some kind of scalability and you need to find ways how to further reduce that effort. I believe that, well, I’ve seen also that standard solutions work. if you think of anomaly detection, basically, where you say, okay, let’s take the first step and…
and use the data to use these kind of algorithms to identify anomalies and then use our domain expertise to basically interpret those and maybe also use systems for that. Or for example, smart sensors or virtual sensors where you start using standard applications to predict the value like a quality parameter, like the viscosity that Roberto mentioned. So I think there is ways to standardize.
And then depending on the size of the problem, I think it also can make sense to go into more the custom-made approach where you really just build an analytics for one case only. Yeah. So unfortunately, I mean, that’s why the question is so interesting. Both worlds exist. Both worlds have a justification why they exist. If you think about, OK, we want to change something and we want to be
David (17:32)
Yeah.
Clemens Schönlein (17:39)
it has to be cost effective, then the scalability aspect and the solution is the way to.
Willem (17:46)
Imagine that one of the listeners is managing a group of people and has the potential to maybe allocate resources in specific ways. What would make more sense to you have like a central Tiger team going from side to side following some excellence program, but then of course not coming in with all the knowledge? Or would you try to work more with the dispersed?
approach how do you see that working best
Clemens Schönlein (18:18)
Yeah.
Yeah, so I also discussions with customers, specifically a big manufacturer in Germany came to me, we have this data science group. We’re doing great work the couple of years, and we don’t have enough resources to cover all the analytics needs in the company, I need some tooling that helps the domain experts to be enabled and do some of the work itself. So basically, like the
The answer really depends on the maturity of the customer, right? So if you never had such a group, you can of course not do that. But my recommendation in that sense is like to have an excellence group that has like deep analytics skills, but also has the skills to enable less analytics like deep, knowledgeable people to still use tools out there to do analytics. Because that’s again, where the decision happens. So yeah.
excellence group, but then also enable and educate people closer to the process.
David (19:22)
Yeah, a bit of both, a bit of both. Yeah. You know, I always love playing a bit of devil’s advocate. So Roberto Ecclémacy, obviously, with his analytic stuff, obviously is living in the cloud. But you know, there is the OT world as well. So I can imagine that when we start talking to manufacturers about, for example, working in the cloud, bringing data from the shop floor up,
that that’s not always an uneasy discussion. So do you have some stories? Do you have some insights to share? what are the typical, I would say pros and cons you hear from manufacturers when we start talking about some kind of a cloud first approach, or maybe not even cloud first, maybe it’s an end approach.
Roberto Serrano Hernández (20:16)
Yeah, I can go on this one. I used to be a security consultant when I started working a few years ago. And I’m a strong believer that the main security flow that you can have is going to be the human flow. And I read some of your posts, David, where you speak about Stuxnet. Stuxnet was introduced by a human, right? So.
I think going to the cloud takes, of course, the risk of not having the feeling of owning your data. At the same time, you’re relying on people who are certified, who claim and probably can do things better than you. And that’s why you have your bank account and you don’t have your money under the mattress. And that’s why you have your email address from Gmail or from Outlook. It’s just the same principle. Then typically, there’s this fear of the cloud.
David (21:02)
Yeah, yeah.
Roberto Serrano Hernández (21:10)
and that’s a dogma in industry. But at the same time, I think there’s a trend that’s changing because many of these companies, especially the big ones, they already count heavy users. you see many companies are using CRMs, many companies are using IT tracking systems, they are using already the cloud. It’s just that they are not using the cloud for everything because it’s still a little bit stigmatized, right? The way that you can free your secret recipe or how you produce Coke or how you produce the…
your tires, how you produce your unique setting points. At the same time, my experience is that we’ve had customers in food and web, we’ve had customers in pharma, and those are very, very, very related.
sectors. And still, you can manage to find the right use case or the right population of users that want to use that tool that only lives in the cloud to do something better with the data. Or they may have a real need on the shop floor. And not all of the data may be subject to those constraints. So if we take a look, for instance, at Food and Bef, they are going to, we have a good example. That’s Nestle, too. Nestle, are producing food. are producing Nesquite. They are producing
Ferrero Rocher, think, and many of these things, right? I’m sorry if that’s not what they produce, anyway, they produce these things and they are very, very regulated. So they need to track everything that they do well or not with the product. And if there is something that they don’t do well, of course they don’t want that data to be disclosed, right? However…
In order to improve the yield, in order to improve the OEE and all of these fancy things that we measure in industry, there are new tools now that might be more appropriate to work and find different ways of seeing those same calculations that they were doing in the past. You don’t need to unveil all of your process data, especially what concerns your critical control parameters for those processes, right? But you can export some of the machine data to understand how that’s working.
at what time, correlate that to some of the batches, maybe, of production that you have, and then understand where you can have some efficiency improvements and that sort of thing. And we’ve been very successful in those things. The thing is that you need to have the real use case that brings value, and you need to prove that the data you’re putting into the cloud, first, doesn’t have any risk for the company. And also, I think we have these…
I would say commitment and this responsibility to educate and spread the word that this is not so bad in some situations,
David (23:39)
No, and it’s also, I think you said in the IT world, nobody would today still host their own mail server or something like that, right? Just the idea would be like, no.
Roberto Serrano Hernández (23:56)
Of course, of course.
Willem (23:59)
Yeah, Clemens, since we don’t get very often data scientists here, I have an additional question. So you were mentioning those center of excellences would be a pattern that you can start to introduce to take care both of those really highly specialized cases and also working on more broad education. So I assume you’ve seen a couple. My question would be if you’re starting out and you have
a couple of guys, you get the label, center of excellence. What tips would you give them in how they should grow as a center of excellence? Because Roberto, I’m sure the platform can give all the data that they need. I mean, it’s a lot of work and we’ve talked about it before, but in the end, if it’s not used, I mean, it’s just a waste of money. So Claymas, how do we get this return so quickly as we want?
Clemens Schönlein (24:51)
No.
So maybe the situation I’ve been in as an, let’s say, external consultant coming to a plant and saying, hey, I’m a data scientist. Let’s work with these new methods. It’s probably quite comparable to the internal center of excellence going to, I don’t know, site A and say, hi, we are the excellence guys. Let’s work together. And the most painful things that I learned doing this is that
I have my toolbox, I believe in the toolbox, and that’s what I want to do, like use the toolbox. And then I go to a process engineer and ask, hey, I this toolbox, what do we do with this toolbox here? And then they come up with ideas because they have a high level understanding where things go wrong. And then we’re working three months on such a thing. And then we bring back the solution to the operator that has to press the button. And they say, no.
That’s not really the problem. The data is not reliable. This doesn’t work for me. So in that sense, the most important thing is to involve, like if you’re in this center of excellence group, if you start an initiative, you go to the business, involve the person from the very beginning one, even if it’s painful because there’s resistance, right? But involve this person from the very beginning on that has to take the decision. You’re building the analytics.
Willem (26:23)
I like this answer. think, no, I like that you didn’t come up with some sort of a technical thing. It’s like go and talk to the operator and involve them. Because one of the things that you might get the impression when you see all those discussions about clouds, AI everywhere, is that you get some sort of like the manufacturing world and then you have the clean high-tech.
Clemens Schönlein (26:25)
You have another question?
Willem (26:52)
AI world in the cloud and that they’re separate. But I think your example here shows that no matter how clean that world looks like, you still need to go and talk to the guy who’s going to press the button and who knows what’s behind the data.
Clemens Schönlein (27:06)
And that maybe brings us also back to the earlier comment around like this big analytics application you build with a lot of expertise or maybe standardized solutions because one very important aspect is the change management and having the support about the change what you’re doing here. And if you have applications that are closer, like analytics applications that are closer to the end user, you’re from the very beginning on have the support.
of changes, even though maybe your analytics in the beginning is just calculating a value, triggering one simple alarm, but that’s the first step. And then you can think like, what’s the next? And then you will take the person with you and you will have more success than bringing this big, awesome tech and say, this is the new thing you should play with.
David (27:56)
closely related to this is the topic of data management. Whether it’s sexy or not, always debate about this on the podcast. But one of the things we always say is crucial. Crucial for many reasons. Crucial because you can somehow capture process knowledge in asset models in certain contexts. You also don’t want, I would say you want to speed up deployments. You can’t…
I would say, be sure that a data scientist is really aware about all the underlying structures in the plans. So what is the AVEVA take or the CONNECT take on data management? Is this something you do in the cloud? Is this something you do on-prem? Is it a hybrid solution who should be working around context and models, et cetera?
Roberto Serrano Hernández (28:49)
I have a strong opinion on that. And to make it a little bit more tangible from the very beginning, I think we need to tie that to, at the end of the day, what’s it that we’re doing, right? And you need the data to be trusted. And the data needs to drive some action. If you have the data there but no action is driven, then there’s nothing to do. So we need to embark the people on the show for it. And that’s one of the things that we do really, really well with CONNECT because it is so self-service.
It’s very easy for the people on the shop floor who understand the machines and actually need to press the button that they didn’t want to press in Clements story, right? They can trust the data because they are doing all of the flow. Now, in terms of data management, what do we do with CONNECT? We have features that allow for creating context and enriching the data. We have the analytics tools embedded within CONNECT and all of that that allow to create a better data model. But I think in this situation specifically,
that we’re speaking in which data from the shop floor goes to the cloud and then we expect the data to coexist and be able to be cross-functional with some of the pieces of data in different departments. I think everything we will do as in any other industrial data platforms that might be there in the market, that’s not going to be enough. There needs to be something much more, I would say homogeneous. At the end of the day, it’s about standardization.
and you can standardize as much as you want your processes, your data from the shop floor, your SCADA system, your PLC system, if at the end of the day that doesn’t match what you expect to see in your data bridge report, what you cross, what that shift was doing when this product code and the supply chain order that I executed and that I paid for. If I cannot do that correlation, then the…
Data scientist on Databricks is not going to be happy and the user on the show floor is not going to be happy either and comfortable with pushing a button. So I think data management is a topic that, know, David, you’ve been speaking about this for a long time in many episodes in the podcast. There’s two takes to this. If you want to be exhaustive enough, it can be very cumbersome, but that’s not a choice for very big companies who are in that state of maturity.
I think there shouldn’t be anything to worry about at the very beginning when starting to build the value from that platform because you need to start somewhere. This is something that you need to keep in mind. Michelin is one of our biggest customers that have been working with data platforms for over the years. And their take is that start small, think big, and do it fast.
So I think that’s the mantra that we should all keep in mind. And I have seen very successful teams delivering value quickly and standardizing as they go. Because the truth is that you start today with as limited scope of function and outcomes that you will expect out of your data platform. But as it drives value, there are going to be more and more people putting their intake into this novel that’s going to be growing and growing and growing. And you cannot start. Sometimes you can, but.
David (31:29)
Yep.
Roberto Serrano Hernández (31:57)
Most companies don’t have the means and the ability to wait and invest for four or five years until my backbone for data management is completely built before I deliver the first use case. That just doesn’t work in my experience.
Willem (32:10)
I also think that if you don’t test your model out in the field and you just sit on your own and make philosophies about the perfect data model, it’s probably going to be dead on arrival and you will have invested a lot of efforts and made a conclusion that, okay, we have to go back to the drawing boards anyway.
Roberto Serrano Hernández (32:31)
Yes.
David (32:33)
And do it fast. That’s what I remember. Guys, we are at the end of this episode. So, thank you so much, both Roberto and Clemens, for joining me and Willem and for sharing your insights. And obviously to our listeners for tuning in. And if you enjoyed the episode, make sure to subscribe at itotinsider.com. Clemens, Roberto.
Clemens Schönlein (32:39)
you
David (33:02)
Thank you so much and until we meet again.
Clemens Schönlein (33:06)
Thanks for having us.
Roberto Serrano Hernández (33:07)
Thank you guys, was a pleasure.
David (33:09)
Bye bye.