Join the Insider! Subscribe today to receive our weekly insights
David (00:00)
Welcome, you’re listening to the IT/OT Insider Podcast. I’m your host David. Subscribe to get the latest insights shaping the world of Industry 4 .0 and smart manufacturing. Today I am pleased to welcome Klaas Dobbelaere. Klaas is the IIOT Connectivity Director at Electrolux. Electrolux is a multinational home appliance manufacturer. think dishwashers, dryers, kitchen appliances and more.
That means that today we are going to talk about discrete manufacturing, paper to digital, industrial Internet of Things, and much, more. Klaas, thank you for joining us.
KLAAS Dobbelaere (00:36)
Thanks for having me
David (00:38)
Hey, why don’t you kick off by introducing yourself?
KLAAS Dobbelaere (00:42)
So I’m Klaas Dobbelaere as you said, at Electrolux I’m responsible for what we call IIoT and equipment connectivity. And I’ve been with Electrolux for three years. Before that, I worked at a global food manufacturing company. So I also know a bit of the non -discrete industry. And before that, I was with a quite well -known Belgium system integrator to learn the ropes and to understand the industry from the bottom up.
David (01:08)
So that means that what is your background? Are you an electrical engineer or an IT person or?
KLAAS Dobbelaere (01:14)
I’m an electrical engineer with an option in automation. So I come really from the bottom, the wiring and then the PLC and I worked my way into the IT sphere as well a bit.
David (01:26)
That’s interesting because I would say before we jump into that, I would also love to maybe introduce Electrolux a little bit more. Like I already mentioned that you’re in the market of home appliances. But could you describe what are the things you’re producing? How can people know your products? What is it actually to do discrete manufacturing?
KLAAS Dobbelaere (01:50)
So Electrolux, most people, at least in Western Europe, so UK, Belgium, France, Germany, don’t know that much about the brand because the main brand we have in those regions is AEG. And then in Scandinavia, we brand our products under the Electrolux brand. However, we also have a big footprint in North America where everything is branded with Frigidaire. And then different users depending on different listeners, depending on where they are, will know our different brands. And actually, we have two main product line.
And I think with that, can already imagine a bit which products fall under that. care, next to a much smaller business unit, which we call small domestic appliances, well -being, more or less. And within that, we produce, like you said, major appliances, fridges, oven, microwaves, cooktops, hoods, dishwashers, et cetera. All the things that, let’s say, a modern household needs to run an effective operation.
And we sell that across, I think, 120 countries and we produce across 30 plants worldwide in all major, countries.
David (02:59)
Wow, and so you’re producing that at multiple locations. So then we actually already touch, what does it take to produce such an appliance?
KLAAS Dobbelaere (03:10)
That depends really on the appliance. You cannot compare how you produce a fridge compared to an oven or a dishwasher. And that’s also an interesting challenge, how to create common solutions across these sometimes quite varied product categories, which I think is one of the major challenges we face today, how to create global applications that are solving problems in these quite different spheres. That’s why we also have quite a lot of specific solutions for these.
type of production lines. Of course, some things are in common like metal stamping, metal forming. We also have some common digital processes, how we flash systems, how we update the software or implement the software on these dishwashers or electric machines. Because that’s really also where we see the biggest change in terms of the product itself. We add a lot of more digital and software capabilities into these equipments.
David (04:08)
So, yeah, I think when we touch software, we also need to touch data and your role, IIoT, industrial internet of things. So what is the role of data and IOT in a production facility of electronics?
KLAAS Dobbelaere (04:25)
So I think our role is really to facilitate a more data -driven approach to everything what we do with the express purpose to reach some people call this manufacturing excellence. Some people like call it operational excellence to enable the people within our operations, whether they sit in supply chain, inbound or outbound logistics, manufacturing itself, maintenance, everything that we know around the manufacturing side, how to become more data -driven, how to
make their operation more efficient and more cost effective. That’s our main goal. And that we try to achieve by, let’s say, a self -service approach, because we really believe that we need to enable the experts on the shop floor to solve the problem they are facing. However, we also want to do that in a scalable way. And that’s always the bit of duality of a global team within a dispersed or global organization. How can you solve?
practical problems on the shop floor while maintaining global scalability and while making sure that we don’t invent six times the same solution across the different sites.
David (05:32)
Can you give an example of such a solution? And maybe as a second question, is there a difference between plants, for example, from a maturity point of view?
KLAAS Dobbelaere (05:45)
I think there are plenty of examples where in the past, plants have gone their own way because historically as well, Electrolux has quite, let’s say, a good local footprint. And then understandably, site management team wants to solve their problems ad hoc and quite pragmatically and in the short term. That’s why they typically reached out in the past to their factory IT teams and said, look, I need this solution. And I wanted yesterday, I think a lot of people will recognize that. And then the IT teams locally came up with…
sometimes very pragmatic and clever solutions to solve the problem. I think we have 10 or 15 different OA systems, how to measure OA, for example, on the production line. Energy monitoring is also a typical use case where the classical utilities department needs something to capture the metering data across the plant. And we have, I think, at least three, four, or five, if I remember correctly, energy monitoring systems. However, then again, if you want to create an efficient operation, if you want to really scale more advanced use cases in a, let’s say,
cost effective way with a lower total cost of ownership, need to find some scalability across that. But again, to continue to enable local people to solve their day to day challenges, which you cannot from a global perspective because you’re simply too far away. And that’s duality is the big challenge we try to solve day in, day out.
David (07:02)
If I think IOT, I think like wireless sensors, cloud connected devices, is that also what you envision or what you think IOT is? Or is it more like a more generic term about all kinds of sensors? How do you define IOT at Electrolux?
KLAAS Dobbelaere (07:24)
IoT and then specifically our team is consisted of, let’s say, two main pillars. On the one side, we have what we call the OT department that sets the global standards and best practices around PLCs, HMIs, fieldbus networks, and how they then integrate with, let’s say, the typical IT networks that we find in the plants. So they are more, let’s say, the data providers. And they also foresee a way how we can structure in a standardized way the data within the equipment itself, so within the PLC.
really at the edge. Then next to that, we have what you could call the true IoT team, even if that’s, as you said, the definition is sometimes quite blurry. That works on developing a system how to ingest that data from the equipment, push it to the cloud, and there to build data, so analytics and application use cases on top of that within the standardized architecture.
And then to also interface with other systems we have across the organization, the global data platform where all the corporate analytics runs, the maintenance management systems that run sometimes SaaS, sometimes on premises, and are a bit like, say, a generic data aggregator and consumer and producer for all the applications or users further in the IT stack.
David (08:39)
That’s interesting because that means that you’re approaching a data platform first approach. do I get it right? If the business has a data question, then you first want to have that data in a data platform or you want to enhance the functionality of the data platform and then only then deliver the application to the business?
KLAAS Dobbelaere (09:03)
It depends a bit. And this is where the greenfields brownfield approach comes in. If we get the demand from the business on a brownfield side, then chances are real that we have not yet ingested that data into our system. Then indeed we need to figure out how we can get the data from the device. The PLC, the HMI, whatever we find. And I think everybody manufacturing knows there’s a great diversity of things that produce data with little to no standardization across them.
And that’s what we call, and we have two people in the team that are focused on that. And we call that brownfield enablement. They are then responsible to, as much as possible, within the device itself, or in a middle way, layer, transform that into a standardized data structure, which can then be, again, ingested in the platform on top of which we can then build the use case. But that data, let’s say normalization, is for us the first step and happens always before the use case. Because in that way, we can also make sure that other use cases can make use of the same data.
And we can easily then transform or transport one use case from one side to another side that has the same data already ingested in the same data format. For a Greenfield approach, we take a bit of a different area. There, we share with our supplier who will produce the equipment, a set of specifications around PLC, HMI, and then what we call ESMI, so the ElectroLux Standard Machine Interface, which is this data formatting that we want to do within the machine.
We share that with the supplier. They need to encode based on what we call a machine profile, what we expect in terms of data to be available within the machine. This is quality parameters, process parameters, order information, machine status information. There’s energy information in there. And that we, for any new greenfield, ingest into our cloud platform already from the start. By default, then, a number of use cases are enabled, whether that’s then energy monitoring or OA.
Or we can also in the future build new use cases already on that standardized data sets that we get.
David (11:07)
So you mentioned, I would say, own standard, your own internal Electrolux standards. Is that based on certain standards you find in the markets? Is there a reason why you developed your own standard? Or is it more like a refinement of existing standards?
KLAAS Dobbelaere (11:28)
It’s not a refinement of existing standards. We communicate over OPC UA towards, let’s say what we have in our case, KEP server. And then from there, it flows into our Azure cloud platform, which we are using. The standards, and here I want to nuance, it’s more a framework because like I said, how we produce a fridge or a dishwasher or an oven is significantly different. So we have a data formatting framework, which we can use to
structure the data that we want. Because, for example, on a fridge you want to see how the insulation is formed that generates a certain set of data. On an oven, you are interested when you do the testing on what is the power draw and that is within specification. So the type of data is significantly different. So it’s not really a standard. It’s a framework that prescribes a certain way to structure the data. What data then we want, we define in that machine profile that is, let’s say, equipment specific, but we also have types of equipment.
Every press, metal press we have across the globe follows more or less the same, let’s say, machine profile because we have standardized that. And the same for the foaming equipment, the same for other types of similar equipments we have in our operations.
David (12:43)
I’m coming back to the data topic later in our talk, I just want to tackle another topic right now. I want to talk about cyber -physical systems, just for a brief period of time. When we talk about manufacturing, when we make beer or we refine oil or whatever,
we’re creating a product, that product on its own is not a smart product. It’s used to either to consume or to make something else. What you are doing is you are creating a smart product. So the oven or the dishwasher or whatever, they have embedded firmware. I have an app which hopefully is going to be supported for many years. And that means that it, so when,
KLAAS Dobbelaere (13:36)
We’ll do our best.
David (13:40)
When you are creating products, who is creating the software on that product versus what is the role of manufacturing here? How do you define, I would say, a cyber -physical system within Electrolux and who plays what role?
KLAAS Dobbelaere (13:58)
So whenever we design a new product, there’s at regular intervals a close collaboration between manufacturing engineering and R &D to assess, let’s say, the opportunity, the manufacturability of that equipment. And there already we start to investigate, OK, if there are new specific demands in terms of the smartness, the connectivity of the appliance, how we can fit it in with either the existing or the newly built production line that will
manufacture that equipment. Typically, the manufacturing process by the time we add the smartness, the software to the product is already in a quite advanced stage. We typically also source, let’s say the computer chip, the display from suppliers. So we get that as a raw material or a semi -finished, you can call it, within our production line. But then depending on the product, we
towards the end of the assembly line, we flash the software that was developed by R &D and our technical department onto that, let’s say PCB on that CPU. And then we install that as part of the system, the entire appliance, whether that’s in a cooktop or an oven or a dishwasher or a washing machine. And once that is done, it goes to, let’s say, the testing stage. So the first, let’s say, operational interaction
is pulling that firmware that was developed and updated from R &D into the device, flash it, install that, let’s say, computer board into the appliance. And then after that, we have, let’s say, a testing step where we test if both the software and the hardware work together in a good way. And there also, again, we share data with R &D and we see how we can optimize the process, how we can optimize
both the software and the flashing, the implementation of that software on the physical device.
David (15:54)
That’s really interesting. So from a manufacturing point of view or from an OT point of view, the thing you are concerned about is how can we get the software we receive from, I would say the software development team, which we are not really linked to. You have interactions, but that’s not part of your organization. Yeah, it’s not operations. So what you actually want to do is how can we as efficiently as possible flash that software onto the devices while we are producing an appliance?
KLAAS Dobbelaere (16:10)
Not operational.
And it’s actually one of our, let’s say, more successful use cases. We call it flashing analytics, where we look at the reliability of that flashing operation in our production line. And we captured that today across, I think, 15 manufacturing sites. Every flashing operation, especially on our taste, food preparation product line, we monitor and we have detailed statistics on to, and then whether you call that a DevOps way of working, have continuous feedback and a continuous optimization, both on the physical process that’s happening, as well as on the flashing.
digital process the software that’s being flashed to the device.
David (16:58)
That’s really cool. It’s another way of looking at DevOps. DevOps is a concept which comes, I would say, primarily from IT. People who read the blog already for a while, they know that we try to, I would say, adopt concepts from DevOps to the way we work between IT and OT. But this is another dimension. So this is really interesting.
For example, for this flashing process, you’re capturing data. You’re capturing data for energy. You’re capturing data for whatever may be which raw materials are used, et cetera, et cetera. One of the things I know is on your Electrolux roadmap is to really become a data -driven manufacturer. Did you start a
program for that, what are the goals, what does it actually mean to become a data -driven manufacturer for electronics?
KLAAS Dobbelaere (18:00)
We don’t have a big name program like some other companies have to really go the industry 4 .0 way. We are thinking about that, but that’s more, think, a branding exercise, how to create the momentum, the change management, which is also an important aspect. Don’t get me wrong. And such a branding can help. What we are trying to fundamentally pursue is that
David (18:15)
Yeah
KLAAS Dobbelaere (18:27)
everybody in the organization, depending on their level, has the right data transformed into information available for them for their role within the organization. Maintenance technicians have a different data requirement than a production manager. An operator at the line has a much more, let’s say, operational challenge in terms of the information he needs than his team leader or than that production manager. The same for quality. And that’s really how we need to empathize with our users, how we can also not
presumed too much about how our data will be used in the future. That’s why, again, like I said earlier, self -service capabilities are so critical to us. We cannot foresee how data will be used today by the different users across the factories, nor, even less, how it will be used in the future. Look at the whole Gen .AI transformation. If we were three years ago, think nobody in manufacturing could predict that. So we already now are starting to gather data that probably we’ll need in five years for whatever we will imagine.
will exist at that point in time.
David (19:26)
Okay, so now that you mentioned the GNI buzzwords, especially I would say GNI on time series data or more broadly speaking, GNI manufacturing is something which gets quite some traction, but honestly, I haven’t seen too much tangible results. How do you look at this?
KLAAS Dobbelaere (19:48)
It depends a bit for me on the use case. I think there are potential use cases for Gen .AI, maybe not on, let’s say, the time series data. Because what for me is lacking there, one, is the reproducibility. I think I’m not a big expert on Gen .AI training, et cetera. But you already see it in regular AI use cases. And I understand the data requirements of Gen .AI use cases are even bigger. You need a huge set of data.
that is, let’s say, transferable across many, many equipment to really train a model that can provide you input. And typically, if you look at our, at least in our case, in the volume of data that we have captured for specific equipment or specific type of equipment, we have already some, but not yet sufficiently or sufficiently contextualized or with sufficient, let’s say, textual data around for it to interpret. For sure, it will recognize patterns, but
Causation is not correlation. this is where I think, on the other hand, I think there are a number of interesting use cases, for example, where our equipment comes with a ton of documentation. And I believe the role of a maintenance technician with the advancements in the complexity of technology when a new machine enters our operation, where all this documentation could help train on a Gen. AI model, train on all this documentation, could help, for example, a maintenance technician or a manufacturing engineer.
David (20:49)
Yeah.
KLAAS Dobbelaere (21:15)
to solve problems on the line either during commissioning or when it’s operational. And that’s what I mean when we need to look at the use case. need to, again, look how we can enable the people who have the real problem. It’s not about implementing a technology. It’s about using a technology to solve a tangible issue. Because maybe people don’t want to chat with the chatbot to understand what the problem is. Sometimes it’s easier to go to a dashboard and look and and dice some data to see what it’s about. And that’s why for me,
David (21:23)
Yep.
KLAAS Dobbelaere (21:45)
the whole AI journey, the data journey is also significantly about upskilling data, upskilling, excuse me, people, how to work with data, what it means to become data -driven for himself. How can I, instead of work in Excel or write things down on paper, pull in data that has been captured from our manufacturing equipment, use that to analyze it, and draw some meaningful conclusions on that. And I think there, I read somewhere an interesting quote that we should never fear technology
disruption because you always need people to make use of the technology and that learning curve needs to happen to significant extent. There are of course some nuances to that, but I think people who fear that manufacturing jobs will very quickly disappear due to Gen. AI. I’m not that fearful because still we need to move things physically around. And if you look to the current state of advanced robotics, I don’t think despite maybe some comments.
of people like Elon Musk that will have humanoid robots in the next three years, that it will go that quickly and be that versatile as they all think. Might be proven wrong, but that’s my humble opinion at least.
David (22:53)
Actually, are you using robots in your manufacturing process?
KLAAS Dobbelaere (22:57)
Yes, yes. more and more. Again, I think a lot driven from labor costs. I think we still have potential expansion into more, let’s say, the cobot space, the collaborative robots, where I think we need to make a step forward. But especially in the higher wage countries like US or the Western European countries, we see a significant increase, especially on the new projects in the use of robots. I think our more advanced sites
When we look at it, I think nearly 40 % of all the operations we do to produce an appliance is automated, either by a robot or by some other automation system.
David (23:39)
interesting. What are, would say, tasks a robot would do? And that’s 40%.
KLAAS Dobbelaere (23:47)
It’s screwing in screws, it’s moving appliances from one location to another. It’s also in terms of hazardous situation, doing certain manipulations. Again, it’s sometimes tricky to generalize across our many, many production types that we have, but it’s more still, let’s say, an economic decision than anything else.
David (24:11)
It’s also interesting that you mentioned the, I now believe that one of the real challenges you really have is this enormous diversity in your manufacturing location, but not only between manufacturing locations, but even inside one manufacturing location. So that’s interesting. What are, I would say from your point of view, what are your major challenges? What are the things you would like to work on in the coming years, for example?
KLAAS Dobbelaere (24:23)
Mm -hmm.
think inspiring people on how to use digital technology is then I think a personal challenge. think that we need as an organization to propagate again because fundamentally people on the shop floor solve the problem. So we need to enable those people with the right skills and the right technology to solve their day in and out problems. And I think that’s the summary of it. Of course, that branches into a lot of different
sub things, which technologies do that the best? Which trainings do we need to foresee both in terms of hard skills, whether that’s on Power BI or R for some people, or more the soft skills, how can we work with less educated people in lower cost countries on how to become more digitally enabled? And I think there you always need to make sure your technology stack evolves in line with the capabilities you have within the organization, both on the IT side.
people developing the capabilities as well as on the end user side within the operations on people who use it. think that’s the main challenge next to of course that duality like I said earlier, how can we solve local problems in a globally scalable way?
David (25:57)
And if you’ve seen, for example, I can imagine that within these plants, will have people who are digital natives, they, for them using a smartphone, using a tablet, using whatever tool they go like, yeah, whatever. If I don’t know the answer, I’ll Google it. But on the other end, you will probably also have a part of your workforce who goes like, wait a minute, is that also something you see happening?
KLAAS Dobbelaere (26:25)
For sure, and it’s actually a conversation we had even earlier today. I believe with the younger generations coming in, they’re so used to the in and out work with digital technology that the, let’s say, digital divide between their personal life and their work life will become so big that if you are not sufficiently digital advanced as an employer, people will say, what weird things are going on here? I want to do this and I’m not enabled.
So it’s becoming more and more of a retention risk as well and will only become in the future. On the other hand, to your point, you always have the aging population, the people, you’re very knowledgeable senior operator who knows by the sound of the machine that something is working or not, that you also need to get along in that story. And it’s that balance to what I mentioned earlier around upskilling, how you can enable both and retaining their…
their talents and the experience that they have and embedding that somehow in the solutions that we want to create. That is the interesting challenge,
David (27:24)
Super cool. This has been a very, very interesting talk, Thank you so much for joining us. We touched discrete manufacturing, we touched data platforms, and we even ended up with upskilling people. So very interesting things. To our listeners, thank you very much for tuning in to another episode of the ITO. The Insider podcast. If you aren’t subscribed yet, make sure to do so. until we meet again. Klaas thank you very much.
KLAAS Dobbelaere (27:53)
with pleasure.