“Unpack, plug in, log on. There was a small instruction manual and then we were off and running. You don’t have to be an IT guru, have studied for it, or have cloud knowledge.” – Uwe Richter, Ziehl-Abegg
This podcast episode with Balluff and Ziehl-Abegg is about condition monitoring – condition monitoring of machines and systems – and the challenges of heterogeneous shop floor networking in practice.
Episode 56 at a glance (and click):
- [07:24] Challenges, potentials and status quo – This is what the use case looks like in practice
- [11:46] Solutions, offerings and services – A look at the technologies used
- [24:23] Results, Business Models and Best Practices – How Success is Measured
- [29:30] Transferability, Scaling, and Next Steps – Here’s how you can use this use case.
Podcast episode summary
Condition monitoring with a commissioning time of less than 10 minutes – this is what Balluff GmbH promises with its shop floor solution. The user – in this use case Ziehl-Abegg SE – receives a turnkey product. A few cables that need to be connected, the step into the cloud where a user account needs to be created, and off you go. No intervention in the control system is required to implement the solution.
Balluff GmbH is known worldwide as a supplier for all areas of automation, one of the largest manufacturers of sensor technology, and a digital expert for the use and integration of valuable industrial data. Ziehl-Abegg is one of the leading international companies in the field of air, control and drive technology: “Wherever air is moved, we are at home”. The high-tech company employs 2,400 people in production plants in southern Germany. Worldwide, 4,300 employees work for the company. Agriculture, wind turbines or data center cooling are application areas of their technology.
How does Balluff support Ziehl-Abegg in this use case? With complete monitoring over their assets, regardless of global location. Not only do long distances on the map have to be overcome, but also the heterogeneous plant landscapes. The challenge was to create a system that would function completely autonomously from the prevailing structure and IT infrastructure. Whether and how this was mastered can be heard in this podcast.
This episode’s expert guests:
- Robert Tilch (Strategic Incubation Manager (Startup), Balluff)
- Uwe Richter (Head of Quality Management International and Technology Scout, Ziehl-Abegg)
If you’ve been following along for a while, you know there’s already a podcast episode with Balluff on IoT Use Case. In episode 26. At that time, everything revolved around the sensor-based condition monitoring of so-called punching machines and transport carriage tracking with RFID in intralogistics. It’s also worth listening here!
Happy Birthday 100th birthday, Balluff!
Robert, would you like to briefly introduce yourself and what exactly you do at Balluff or in your segment?
Balluff is globally active. We have 3,600 employees worldwide and come from a classic sensor business, i.e. a classic hardware business for the last 100 years. Now we have thought about how we can go along with this change of digitization and join in the fight, how we can make it better? We launched internal startups in early 2020. One of them, you could say, I run. Internally, we call this “Strategic Incubation Programs” – that’s why I call myself “Strategic Incubation Manager”. Since 2020, we have been on the road there with an issue, also in the area of condition monitoring, and have developed a holistic solution there.
Which customers and partners do you work with there? Are these the classic sensor industry customers or do you have new partners there?
They are actually similar customers on the one hand. Spread across the entire spectrum of industries: Automotive, but also packaging, which we are pushing very strongly at Balluff. Our focus with the system, which was developed in my team, is first and foremost to reach out to end users. That is, not so much those who build machines, but also those who operate the machines. In this way, we are trying to develop smart solutions, intelligent solutions for the shop floor – not only on the hardware side, but through the entire value chain. Software, gateway connectivity and so on.
And Ziehl-Abegg is one of them?
Exactly. Ziehl-Abegg is partly hybrid, of course, because: manufactures machines, but above all, USES the machines. With Uwe, we have someone who is primarily interested in utilization.
Uwe, would you also like to introduce yourself briefly and say something about you, about the core business?
My name is Uwe Richter, I am responsible for Quality Management International and also act as a technology scout within the company. That means I also look for new technologies, new trends on the market, and see what we can use from them and where we can implement them and make sense of them. Ziehl-Abegg – we have around 4,300 employees. The core business is industrial fans, then drive technology with elevators. Then we also have the right control technology for all these applications. We are actually everywhere at home where ventilation is required. That’s in agriculture, in wind turbines, in data center cooling. Wherever air is moved, we are at home.
Who are your customers then? The operators of such data centers or also the manufacturers of such elevators?
On the one hand, the manufacturers of the elevators. They buy the drive technology and the control technology from us. In air technology, the supply chain is a bit more complex. We are then the manufacturer of the fan that goes into the control system or ventilation system. The ventilation system then goes into a building, and eventually at some point there is someone who runs the whole thing.
That was what Robert had just rightly mentioned – you are quasi hybrid on the road, i.e. in two roles. Today, however, we will be talking about your own shop floor, or rather about digitization within your company. What is your vision with IoT and digitization? Where do you want to go, internally as well as perhaps externally?
In terms of our factories: Everything we buy new is already IoT-enabled. We are already on the move, and we are networking and using this. Of course, we also have some machines that are not yet networked. For example, we are looking for very simple solutions, plug-and-play solutions that we can install relatively easily. Where I put together a package and send it to one of our production facilities around the world: Someone there can unpack it, adapt it, and then the whole thing runs and we can already collect data.
Challenges, potentials and status quo - This is what the use case looks like in practice [07:24]
I always try to get a virtual picture in my head. You said you work in quality management or as a technology scout – what are actually the challenges you deal with on a daily basis? What does your daily task look like and what are perhaps the challenges for such a heterogeneous brownfield?
You touch on everything, of course, when you get into the whole subject. That starts with, what’s all the connectivity like? That is, from IT security – how do I get the data collected? How do I retrieve the data? How do I get them into the cloud? How do I get them back out from there and then evaluate them? This whole path, that’s already the whole complexity. Then, of course, you have different development characteristics of machines that are already really completely networked and collect masses of data. For example, with us in plastics technology. But then you also have machines that are simply a bit older: They have the data buried deep inside the controls, which you first have to tap into and extract from there, process, and then make available. So there you touch on everything and then you have to react individually.
Plastics production, that is, you have, so to speak, for example, an extruder machine there, or probably several – or what do the machines look like there?
These are larger plastic injection molding machines. This goes from 1500 to 2300 tons, where we then inject our fan parts – mostly impellers.
It’s about connectivity, it’s about the huge issue of how do I get this heterogeneous structure into the cloud – do you have an idea, what would it take to make these machines smarter? What is the approach to achieve connectivity?
Our basic requirement was to look for a system that was as simple as possible in the initial design, especially for all the existing machines, where I don’t have to put an additional controller, an additional PLC over it, which picks up the data for me and then has to generate the data via umpteen gateways. But a plug-and-play solution to get relatively easy access to the information that the machine provides me.
Robert, do you see this challenge with other customers of yours? It’s probably always a similar issue – you have a heterogeneous infrastructure, old controls, maybe also an S5 control, or old machines. Do you know this issue?
That’s right, S5, S7, whatever, Mitsubishi. Any company that operates machines has an uncanny heterogeneity. The challenge then is to still get the relevant data out of the system quickly and easily – most of the time this is not such a terribly large number of data points, it reduces to quite little. I see this with quite a few customers who are also struggling with this wide brownfield.
Uwe, you said you needed some kind of construction kit. Is that the idea then, to say you actually want to tackle this once and then roll it out to other works? That you can also get started relatively quickly with the use cases? Or what exactly did you mean by kit?
That was the plug-and-play solution, where I can put together a package relatively easily, can then send that somewhere. You probably referred to that construction kit. Then I say, “dear colleague in China, unpack this, install this on ten machines.” Then I already have the OEE, for example, and can collect data on the machines – is it running? Is it not running? Do I have any malfunctions ? This is the idea where you need something simple. Because here, too, the rule is 80/20. With maybe 20 percent effort to get 80 percent, you have already won relatively much.
That is, the goal behind it is to monitor your systems, both in China and in the plants here in Germany, or wherever they are globally. Knowing when the system is running, when it might not be running? These are probably topics that move you?
Solutions, offerings and services - A look at the technologies used [11:46]
Robert and Uwe, you’ve probably been working together for a while. You also said you come from the sensor business. With your startup, you are of course also close to the end users as a single-source solution provider. So you chose Balluff and said – technology partner for condition monitoring, that fits your challenge? How did this path come about? How did the collaboration come about?
The cooperation with Balluff is not only about condition monitoring, but starts long before that. We have various solutions in use for Balluff and have been working together for a very long time and very, very well. Balluff is a technology partner of ours. We exchange information on a regular basis. That is, what are we up to, what is Balluff up to? Then the topic of condition monitoring came up and we exchanged ideas. The information came that there was a smart, nice, fast solution from Balluff – so we said, okay, let’s test it!
How exactly do you do that in practice? For example, I imagine you have a plant with different heterogeneous machines. How does condition monitoring work, what is behind it?
The issues that Uwe described we found with quite a few machine users – end users, as we call them. So heterogeneity, which I just mentioned, but also a complicated IT infrastructure. Many different machines. The challenge for us now was to create a system that works completely independently of the prevailing structure, IT infrastructure, machine structure and captures and generates condition monitoring data. The second point was that it must work everywhere and globally, and be very quick and easy to get up and running. So we first looked at the requirements that existed in the market. From this, we put together a package, and now, in a whole lot of sprints and iterative steps, we have arrived at a solution that we can also offer. It runs at customers such as Ziehl-Abegg and now at others as well, and is in use globally.
From practice, if I imagine I have my machine, I have the sensor. I have to connect the data in the first place and perhaps manage it. That is perhaps a bit of the first step. How do I do it today? How does this data recording work?
I may have something to say about how the unboxing of the Balluff sensor, or condition monitoring device, worked. That’s when the package came. I unpacked it, logged the sensor into the cloud. Then I put the sensor on my notebook and could already run the first condition monitoring of my notebook.
This means that the first key figures can be seen there relatively quickly. Then the next step is, now the data has to be collected somewhere and monitored. How does it work? Do you have a database there?
When we set up the system, we decided that we wanted to keep the data in the cloud. That means, one is the hardware. We always promise that this can be put into operation in under ten minutes. It’s really just two or three cables that you have to connect. Then the step towards the cloud, where I only have to register with a user account. Because our system, our gateway, sends data via mobile radio, data is of course sent on-the-fly very quickly after the device is switched on, i.e. within two minutes. I then also see this data immediately after I have registered in the cloud. This is very simple, as I said, in under ten minutes. That was our goal and several customers have confirmed it to us. The example with the laptop is the best for this. Applying that from the laptop to the machine and carrying it over is then just a tiny step.
How exactly does it work with this laptop? Someone stands at the machine and sees this data in real time?
I put the sensor on my laptop. Then I could see how IT runs. Does it vibrate? Is the fan running? Is it warm? You can see all that then. I was virtually supervising my own work.
Perhaps as a detail: Because the data is in the cloud – we work with Microsoft Azure – you can actually access the data from anywhere. The laptop where the data, graphs and information are displayed does not have to be right next to the machine. Instead, in case of doubt, as is now planned at Ziehl-Abegg, the machine is located in Asia, and Uwe can monitor the data from headquarters – and even on a smartphone or on a laptop or tablet. Any device can be used and it is totally location-independent.
That is, I have the sensor, it gives me data directly on vibration, temperature, whatever I need to calculate OEE, and I have the whole thing on a dashboard – which is what this buzzword condition monitoring represents in this case, right?
Exactly. The sensor specifically measures three-axis vibration, contact temperature, humidity and ambient pressure. We have found that this is the most important information needed to monitor the condition of equipment. The nice thing is that this works with us in one device. You don’t need five different sensors.
Do you also go one step further – that you say the next step is to analyze the data as well? One might also work with historical data. Is that the next step, predictive maintenance, or is that perhaps already happening today? Of course, it is also interesting to gain insights from this.
Currently, we are still collecting the data. We have different pots and are already evaluating them. We have our own software that sits on top of that and processes and visualizes the information.
The next step would then be to really say you’re going a bit further, in the direction of predictive maintenance or even artificial intelligence. But this is probably the supreme discipline?
I see our tool as a bit of a data collector that quickly collects and provides the data. However, we have also implemented approaches and initial proofs of concept there to export the data. The easiest way is a CSV interface. This comes as a feature. But also via a REST API or a cloud-to-cloud connection. As I said, we have the data in Microsoft Azure. However, if the end customer wants to have the data in another cloud, we also have options there and are working on expanding the connectivity so that, for example, this data can be aggregated at a central location and linked to other system and machine data. Then, of course, I can develop in the direction of artificial intelligence. But that’s a very broad term. I really see our system as an enabler. Creating and generating the data in the first place; having it digital in the first place.
REST API and so forth, which you just mentioned, that’s just the interface that’s then created to route data through where you need it accordingly. Perhaps also for use cases that will come down the line.
Uwe, what prior knowledge was actually necessary for this? You specifically come from a quality management background. You have to spend some time with it first, don’t you?
Sure, I come from a quality management background – but I love technology. Therefore, it was not a big challenge for me to unpack the device. As we have already discussed: This is really real plug-and-play. Unpack, plug in, log on. There was a little instruction included and then it actually started. You don’t have to be an IT guru, have studied for it and somehow have cloud knowledge or anything else. But, yes, it was relatively simple.
I still remember an early feedback Uwe gave me after the first start-up. From the laptop to the machine – it was an injection molding machine at the time – it was really only a few steps. Also this trial and error then. So to pack the sensor times at place A, then at place B. We have a magnetic holder for the sensor, so I can simply try out the most diverse locations on a metal housing. This is what often happens in the end: The basics are delivered, and then the user tries out and tries to extract the valuable information for himself.
Uwe, perhaps another leap into practice. How do I actually have to imagine this: I have the sensor, unpack it and then probably still have a cable – or how do I get that to the laptop? You also said there’s another modem in between; that means I have the sensor. The sensor must be connected to the modem and that then goes to the cloud via mobile communications?
Exactly. Simply plug together, screw together. These are all really industrial-grade connectors, too. At the end you have a Euro plug, you plug it into the socket. There is a button that you turn on. Then the sensor is activated, the data transmission – and that’s it. That’s all I have to do on the device.
Super. I was just imagining if I started this tomorrow, what would I have to do now? That is, I have this component, unpack that, plug in the sensor. There is also a gateway that transmits to the cloud. What does that actually look like? Do I have a modem with a SIM card on it?
I always say as big as a normal smartphone. Pretty much six inches by dimension.
Perhaps one more example. In terms of usability, it’s about as difficult to use as a clinical thermometer.
Okay, we can all do that, very well. Perfect. If I summarize that a little bit again: We have the first step to connect and manage the devices. Robert, you’re right there with Balluff – connecting, managing. Subsequently, in the second step to collect and perhaps monitor.And then comes, Uwe, your part: You have software that analyzes the data. Then all the intelligence, the condition monitoring, which takes place in the cloud, is consolidated via a Microsoft Azure service.Did I get that right or would you describe it differently?
Basically, the software also comes from us. That means we really deliver a complete package: from the sensor, to the gateway, which sends the data to the cloud, and the display of the data in the cloud. We built a dashboard there, which displays the essential information, which is super simple and intuitive to use. – And these advanced analytics are ultimately then customer-specific requirements and projects that we can then implement with people from my team or other teams or with third-party providers, for example. If there is a need in the direction of predictive maintenance and further analysis.
If I want to start tomorrow, what components do I need?
The sensor, the gateway and the software. But you buy the whole thing in a package from us. You have a package price for the hardware. Then a software subscription so you also get regular updates and upgrades in the cloud. There are two options to choose from. One thing is, what type of sensor do you need? Only vibration and temperature? Or the sensor that can also measure humidity and ambient pressure? And, do you need a power supply or not. That is also still an option. Once you answer the two questions, you get that packed in a box and delivered.
This is probably also very use-case specific as to exactly what data I need. Depending on customer requirements. With you, Uwe, it was the injection molding machine where you said such and such data is relevant to you. Someone else might choose different data points there.
Results, Business Models and Best Practices - How Success is Measured [24:23]
To the business case. The bottom line is that we are not digitizing the shop floor for no reason. Uwe, you described your vision at the beginning. But do you also have a kind of business case, return on investment, for this very specific case, where you say you can already calculate that? At what point do you start saving money with this technology, have you worked that out for yourself?
On the one hand, of course, before calculating the ROI, you have to weigh up, does the technology make sense? Does it take you further if you develop in that direction? I think that’s a very important aspect, and that’s how it was with us. We also thought, how can you calculate something like that?
It’s a very simple calculation: if I say I have 16 working hours every day; I have 250 working days a year; then I’m at 4000 hours a year. And if I calculate the 4,000 hours with a flat hourly rate – really flat, not ours – of 100 euros, then I arrive at 400,000 euros. Now let’s get one percent of that, then that’s 4,000 euros. And one percent – I’ll get that quickly if I go with 80/20. So from an ROI perspective, there’s nothing to be said against it.
That’s a quick return on investment. Every technology has a price – but the bottom line is to earn money with it and, of course, to invest in new technologies. That’s a nice example. You’re the one doing the technology scouting; you’ve probably looked at a wide variety of technologies and then, of course, have to make a good choice.
Why actually Balluff? You said you’ve known each other for a while, probably from the past – worked well together and built it up?
Exactly. What works well should be allowed to continue.
Still on the amortization: In fact, we have system costs of less than 1,000 euros here, which we call for this portable monitoring system. We have heard that this pays for itself after a very short time. Also this question, why did you choose Balluff: Yes, there are plenty of vibration and condition monitoring systems on the market. But once that it is really very easy to implement. And we have also already heard from customers, the industrial appearance of the solution is decisive for many. Because we as Balluff really have an industrial-grade solution and can offer the entire value chain despite everything, from sensor technology to the cloud.
We really had a lot more issues in the pipeline. I can also adapt the sensor to the system without directly performing any intervention. That’s also an important point again; because many other providers really intervene. Take current sensors that they clip over somewhere, over cables, tap into the controls, and so on. If I then intervene too deeply in the machine, I have to take CE into account, and then I quickly become the manufacturer of the machine. That’s a whole different league then. For us, this was clearly an argument in favor of Balluff for this solution, if we really do not have to intervene in the control system.
Also, do you guys have any insights on how many sensors you’re going to put there? You probably have hundreds of machines with hundreds of sensors.
There are some already connected. There are also more and more. Whether we really connect every last machine is something we’ll have to see. There, off-topic, we don’t have a clear roadmap currently on how we want to launch that. The idea that we have here is really the package that we briefly touched on earlier. That we say, you go out, you install this in a manufacturing area, you do several lean projects there using the information, the data. Then you uninstall the whole thing, then go to the next spot and continue there. This was also the original concept.
In other words, you can transfer a package to several sites, so to speak, and thus allow the competencies to merge to some extent across sites as well, as far as these projects are concerned?
Exactly, because with the installation of the sensors alone you still have no improvement. But of course you have to evaluate the data, you have to analyze it, you have to run continuous improvement projects over it. That way it will be better. They need time to take effect. Meanwhile, you can then install the measures, can go somewhere else, can elicit new measures there, install them there. And then go back, do the measurement again – was the whole thing effective, what you did, or do you need to follow up again? Pure quality management.
Transferability, Scaling, and Next Steps - Here's how you can use this use case. [29:30]
Robert, you have different use cases that we present together as well. If I am a listener and may have similar challenges. Might not have an injection molding machine, though, but any other – do you see issues like this with other of your customers? Is it always similar? Is it possible to transfer this project?
Definitely you can transfer that very well. Anyone who operates a plant struggles with maintenance and repair. Most of the time it doesn’t work as planned – but failures always happen at the worst times. If you still measure manually today, if you invest a lot of time to generate condition monitoring data, I would say in very modest self-interest: Then the first step is to have the data digitally; to attach this to the machine very simply. Then this can really be a super transfer point for our system. Many also start with something like a proof of concept; with a system, try that out and are excited about it, and roll it out across individual sites at different facilities. Or even across the different sites. That’s kind of the progression we see with our customers. This is, of course, quite nice to observe, and can be adapted to any type of plant that needs to be maintained in any way.
I think if there are listeners now who have similar challenges who would like to discuss – off-topic if you like – with you, Robert, or even together: I think you are reachable via LinkedIn, I will link that accordingly in the show notes. Otherwise also via Balluff directly. Listeners are welcome to contact there. Incidentally, we have presented two really exciting cases that also go in this direction, with the Schaeffler company – this is about the Schaeffler plants being networked – and the Liebherr company. So feel free to listen in there as well, with two similar topics; it’s just as much about heterogeneous-shopfloors connectivity. A very exciting environment. Robert, do you want to add anything? What else is coming up for you in the future?
We are in really startup-like structures in my team. Through the cloud and our IIoT-enabled components, we have continuous upgrades that we deliver and ship. I can’t tease too much yet, but there will always be new exciting things to come. I’m also super interested in exchange and customer feedback, and in adding new topics to the roadmap at any time.