Companies face the challenge that their machine data is often not digital or not in the right format. There is often a poor database and/or a lot of IT/OT heterogeneity or data silos. Symptoms of these challenges are slow progress in digitalization projects and rework that is only possible with great effort.
Preventing these symptoms from arising in the first place is the topic of the 111th episode of the IoT Use Case Podcast with Philipp Merklinger, Technical Consultant Digital Transformation at ITK Engineering and, of course, Ing. Madeleine Mickeleit.
Episode 111 at a glance (and click):
- [06:28] Challenges, potentials and status quo – This is what the use case looks like in practice
- [12:26] Solutions, offerings and services – A look at the technologies used
- [26:24] Transferability, scaling and next steps – Here’s how you can use this use case
Podcast episode summary
Philipp Merklinger and Madeleine Mickeleit talk about how companies can manage to filter out the really relevant information in the flood of data. Philipp Merklinger and Madeleine Mickeleit discuss how companies can manage to filter out the truly relevant information in the flood of data.
Philipp Merklinger sheds light on the importance of Overall Equipment Effectiveness (OEE) as a key indicator of production efficiency and addresses the daily challenges and potential solutions his customers face in their operations.
They also take a look at the technologies used, data collection and analysis, and discuss the challenges that arise when their machine data is not digital or in the right format.
This episode concludes with a look at the future, including topics such as 5G connectivity, IoT platform development, and AI requirements for training datasets.
Hello Philipp, nice to have you here today. I’m so happy to have you as a guest on the IoT Use Case Podcast today. How are you right now, and where are you at the moment?
Thank you Madeleine, I’m fine. Thanks also for inviting me to your podcast. You can reach me right now at our location in Rülzheim, which is about half an hour west of Karlsruhe, on the Palatinate side of the Rhine. I am happy to be here today and to be your guest.
Very nice. Do you come from the area, or are you originally from somewhere else?”
I was born in Karlsruhe and also studied there, industrial engineering at KIT. But now I have moved to the other side of the Rhine.
Shoutout to Karlsruhe. You are a wholly owned Bosch subsidiary, I believe, with over 1300 employees according to LinkedIn. You offer cross-industry solutions, including very specific ones, in the field of system and software development. Philipp, you are a technical consultant for digital transformation. What department do you work in exactly, and what customers do you have contact with?
Generally speaking, ITK is organized as a matrix. This means that on the one hand we have business areas that include, for example, rail, healthcare, automotive, or even industry. On the other hand, there are a number of specialized teams. These have specialized in different areas. This ranges from systems engineering and classic software engineering to current trend topics such as AI or cloud technologies. As an engineering service provider, this offers us the opportunity to flexibly staff the projects we have with our customers with different competencies. My department is called Smart Factory Technologies and deals with all areas related to I4.0. These include, for example, the development of interfaces to overcome system discontinuities, the topic of factory simulation, or even the topic of data collection or retrofitting in production.
Cool. Now you’ve already said you work with very different business areas. What kind of customers are they?
We have a wide range of customers and use cases. This ranges from OEM manufacturers to a small machine manufacturer from the SME sector. It’s hard to specify at this point.
In the podcast, I always discuss various practical use cases to understand the business case behind them, why companies engage in these endeavors, and to talk about monetization as well. Can you tell us about the project we’re looking at today and the use cases that go along with it?
Today we want to look at the use case of data collection or retrofit in production. Specifically, the issue is that customers have machinery available in production that provides only limited data, or perhaps none at all. This, of course, makes it difficult to take action. That’s when it’s important to identify which data really help customers make the right decisions to optimize their production planning. Or if we consider this on a long-term horizon, such as which machines I might need to acquire or maintain differently, today’s focus is on assisting in these decisions and determining the data required for them.
Yes, very cool. You have various success stories online, are we allowed to talk about the “Bosch Rexroth plant” case or is that something we can’t discuss in detail today?
We can certainly include that as well, no problem. I can mention that too.
The nice thing about you is that you have so many customer cases. In the Bosch plant, what are the cases that are implemented there? Is this also so classic data collection/retrofit or do other issues come into play?
At the heart of the project is the topic of retrofitting. You have a machine that is not connected. There are also newer machines that already provide key figures to assess their productivity. For example, OEE, Overall Equipment Effectiveness. In concrete terms, this involves retrofitting the existing machines so that the key figures are comparable with the machine park. This is a major challenge at Rexroth, which of course has many plants and many productions itself and also advises customers. That’s where it’s all about developing the whole thing on that level.
[6:28] Challenges, potentials and status quo – This is what the use case looks like in practice
Among other things, it is about cases around the OEE, the Overall Equipment Effectiveness or also about data collection and retrofit. What is the business case of your customers? Why are they doing this today?
The OEE is used to assess the productivity of the machines in production. Particularly in the area of availability, the aim is to optimize downtimes, which can also be planned. Based on that, one can think in the direction of predictive maintenance. How do I create my shift schedule? Are there optimizations I can push in this direction? Finally, the issue of quality is also important. What scrap rates are being produced? Are limits exceeded? From this, measures such as employee training or changes to the production process can be derived. The OEE can indicate this and is the starting point for further analyses. That’s why it’s important for customers to think about OEE.
Yes, many companies have already started down the path, though not all yet. It is possible that some of them already have key figures in mind. What are the challenges to going down this digital path, especially in the context of IoT solutions? What day-to-day challenges do your customers face? What difficulties arise when you want to implement something like this?
One of the biggest challenges, especially when scaling data-driven use cases, is a poor database. As a rule, this does not mean that no data is available; in fact, it is usually available in large quantities. But often there is a lot of IT/OT heterogeneity and data is just buried in so-called data silos. That is, companies are already collecting a lot of data, but the data is not being used. There is no connection to other systems, they cannot communicate with each other. That’s a big challenge, to virtually eliminate such quality problems. This is exactly where we try to support the customer.
Okay, now you’ve already given me the perfect segue. I mean, with the OEE it is almost already defined, but somehow you still have to get to this data first, or also collect the most diverse data types. Can you tell us what kind of data that is? What data do I need to collect for this?
In general, it is possible to implement different connections with the Transparency Toolkit. It depends on the use case. You can then also use standard protocols such as OPC UA or MQTT. On the other hand, we can also extract data directly from control units or from higher-level IT systems such as MES or ERP. Specifically for OEE, it is of course important to get the signals directly from the machine. For this reason, we at Bosch Rexroth recorded and evaluated a total of 40 signals from a PLC. Important parameters are then something like the current material number, the active production program and the current machine status. Are there any error messages? Are there quality indicators such as dimensional inaccuracies?
Now, for example, you mentioned material numbers. Why is it important to include the material number?
For example, your machine may be producing a certain product flawlessly, while another batch keeps producing errors. Therefore, it is critical to isolate these cases from each other later. It’s possible that the issue is related to the product geometry in the milling machine, which can be addressed in this area. That would be my approach to this.
Now you’ve already touched on it, it’s about your Transparency Toolkit. We’ll get into what it is and what it can do in a bit. What are some requirements that you hear over and over again that lead to the use of the solution? What is important there?
An important requirement from the customer, in addition to technical advice, i.e. which data should be collected or stored, is in particular good cooperation between the project teams. Another requirement is that the data we ultimately collect must also be placed in a certain context. Our customers are experts in their production processes, but sometimes they can also be a bit operationally blind in certain areas. At this point, we support the customer in identifying anomalies and also assigning them to a cause together, if necessary, so that customers also have a clear added value.
Yes, that’s for example what you just mentioned, which is to provide context with the material number to allow comparisons and to go deeper into the data.
[12:26] Solutions, offerings and services – A look at the technologies used
What exactly can the Transparency Toolkit do and how was it used at Bosch Rexroth’s plant? How does it work?
In general, the Transparency Toolkit is a prototyping toolkit. That means we can take the existing system and adapt it to the customer. For example, if a customer wants to collect energy data, we add an energy meter and ensure that the components can communicate with each other. The data can then be stored and analyzed via an IoT platform. This illustrates that we strive to select meaningful data that is relevant to the use case at hand and perform the evaluations that are important to the customer based on that data. In the case of Bosch Rexroth, the aim is to automatically calculate the productivity of the machines to enable control of the machine fleet. This is done in particular against the background that Rexroth already carries out OEE calculations in many places. It is important that these calculations are comparable with each other. The Transparency Toolkit ensures that the OEE for this particular machine can be calculated in the same way, enabling comparison and optimization of production.
I just opened an image of your Transparency Toolkit, I’ll link that in the show notes. It can be described as a toolbox that you put there. First of all, these are various devices connected with cables. What is the composition of this product, what do you see?
In general, the Transparency Toolkit consists of various components. We try to be as flexible as possible. We have elements in there that allow us to read out PLC’s. We also have a wireless bridge that allows us to add an inductive sensor in production, for example. Then that one can communicate with our Transparency Toolkit. The collected data is stored together with the data obtained from the PLC, and we can thus make connections between them.
It’s not just a hardware toolbox that you put there, so to speak, which can also make exactly these different connections, but also a piece of software with it. Check out the show notes or have a look at LinkedIn @IoT Use Case. I’ll link it below the podcast post, so you can see exactly how it looks in practice. You said you were collecting PLC data. These are a wide variety of interfaces and protocols. How exactly does data collection work?
Now in the specific use case, it is of course important to realize that such a PLC provides a large amount of data. A Siemens S7 provides a whole range of data fields. Now you need to figure out which are the important data fields. Does this data field really contain the data that is written on it? All of this needs to be made plausible. The exchange with the customer who has all the process knowledge, in this case at Rexroth, is crucial. For example, this customer can provide information on how OEE is calculated on other machines. Then you have some comparability and can calculate the OEE. The customer then has a single source of truth.
Are there certain obstacles with customers who are trying to implement this without this Transparency Kit today? It’s incredibly time-consuming to bring all these data types to the same level in order to enable the connection. Do customers usually handle this in a very elaborate manner, or how does it work? What are obstacles that customers face without you?
Exactly. As you say, a retrofit of a machine is not simply done overnight. Behind this are major projects at customers’ sites, ranging from budget requests to registering devices on machine networks to training employees. A big advantage of the Transparency Toolkit and its flexibility is that we can go into production, collect data over a two-week period, and prototype whether it works. Then you can go from there and further optimize your production. It’s mainly a time saver there.
It is also about the important data fields. What are they anyway? What data is even relevant to establish this context and comparability? Do you have some kind of guide there for our listeners? How do I go about it? What is your approach?
Sure, gladly. So just to keep it in mind again, every project is absolutely customer specific with us. We do not develop standard software. In general, it is always important to formulate a clear goal at the beginning. If we look again at the material number, for the different material types, I want to analyze how they perform on my machine. Then I have to think about what sensors or what data sources are needed to answer that question. In this case, the data was available in the PLC and we were able to access it via an IoT gateway at that point. The next step is to check this data, which is then available, for quality. That is, do I see the correct data here? That’s when I have to talk to the customer again, with the process experts. Findings are then derived together. This is also always very important, because an evaluation alone will never generate added value. You also have to really record the joint findings and also think about whether the whole thing can perhaps also be implemented on another machine, i.e. bring the issue of scalability into it. What we also consider important, especially in the case of changing parameters in production, is the regular review and, if necessary, adjustment of the evaluations carried out during operation. It may happen that new environmental parameters are added that must be taken into account. In such cases, they must be included in the considerations.
You had mentioned context and comparability in the context of scaling these topics. Are you referring to this on a machine-by-machine basis? Or, if it was implemented for one plant, can it be made available for other plants? What does scalability of these cases mean to you?
Scalability comes in different levels. I mean, you don’t start by completely converting your entire facility; instead, you begin with a machine where you are relatively confident that you can achieve it and can demonstrably verify whether this calculation is accurate. Then I go to the next machine and say, okay, maybe I’ll manage to implement the approach on my whole production line. It’s sort of like a cascade. You then transfer what you’ve learned on a small scale to larger contexts.
How does this installation work on site at the machine? You bring your toolkit with you. You have defined beforehand what the data types are, where they come from, what the devices and sensors are that are connected there. How does this installation on site at the machine work?
Generally, this is an installation together with the customer on site. The process owner can show us where to position the Transparency Toolkit. First of all I have to clarify such issues as enabling the Transparency Toolkit for the machine network, i.e. I have to assign a MAC address and an IP address. Then it is basically a joint installation on site, a kind of plug and play. At Bosch Rexroth, there was a very good collaboration on site with a ‘hands-on’ mentality. Often, customers may experience challenges with the Transparency Toolkit that require this certain mindset. This could be problems with the network topology, for example. There I was a few weeks ago in a factory where the network jacks were not properly patched. Although everything was connected properly, it was not visible where the cause of not seeing the data was. Then you try to isolate the error, and eventually you come up with a solution. But the level of ‘hands-on’ mentality is critical to the installation.
I’m sure you have in-house experts who will come along and help with the installation on site. Now you had just said IP or even ports that are opened somewhere also from a security point of view. Is it common for customers to move their data to another server, or do they say, ‘We’ll keep the data here on site’ and use an on-premise solution, or do they consider moving to the cloud?
In principle, it is possible to store the data in the cloud with the Transparency Toolkit. This decision is always up to the customer and whether they can reconcile that with their policies. One advantage, of course, is that if the data is stored in the cloud, we can already access it remotely during the collection period and make adjustments if necessary without having to have someone on site to show us the data in a meeting and confirm whether everything worked. This significantly accelerates the process. On the other hand, and I understand some of the customers here, they want to manage the issue of data security themselves because they ultimately have to take responsibility for it. Then it is also possible in principle to save the data locally with the Transparency Toolkit. Then you can sort of take things with you when you pick them up and then read them in and evaluate them. That’s entirely up to the customer in the end.
Very good. That means I can use your Transparency Toolkit to make the connection to different data pots but also to different systems. I have a certain database to fall back on. I can also identify initial measures. How does the data analysis work?
Basically, the procedure is that when we have collected the data, we have to carry out a plausibility check. In the first instance, it is of course always most helpful if this plausibility check is already made on site by the customer. They ensure that they have the right connections and that the data is coming in to the data pots. Now, however, we may still be checking the data quality downstream when we have the data with us. For example, I collect energy data and I know that a forklift charger has a very specific energy demand, while a machine has a relatively high and continuous power demand. If I see in the data that there is a stable energy demand and the SPS at that point says that the forklift should have been tracked here, we can, by simply looking at the data, see that something may be wrong. Perhaps other data will then need to be included or the data assigned to another type to provide context.
I’ve also learned so far now, I can do connectivity through your Transparency Toolkit to different data pots, but also different systems, whether that’s a PLC or other data. I have a database that I can draw on. You described it, also with different hosting depending on how the customer wants it. And I can also already identify initial measures, then also bring the business case a little closer. Now we have, or you have already talked about context and also comparability of this data. Can you share how this data analysis works now? We don’t have to stay with the example of the material number, but maybe you have another example, just so that people understand, how do you do this data analysis?
Exactly. At this point it should be mentioned that it is of course very important to check the plausibility of the data. This means that, as a rule, the first plausibility check is carried out on site by the customer. They make sure it has the right connections. They make sure that the data gets into the data pots you described. However, it’s also possible that we’ll check the data quality afterward when we have access to the data. Meaning, I verify there, I collect energy data with the Transparency Toolkit, for example, and I know that different devices, like a forklift charger, for example, has a very specific energy requirement. While a machine has a relatively high continuous energy demand, such a charger has generally like a sinusoidal energy demand. And when I now see in the data that there is a stable energy demand, and my PLC suggests that at this point, perhaps the forklift should have been tracked, we can already, by simply looking at the data, say, ‘Yes, something might not be right there.’ We may need to collect additional data in that area, or we can assign the data to a different category to establish the context.
Alright, yes, incredibly interesting. The topic of OEE was just one example. You have very different customers, and each customer probably brings a very different case that can be implemented with it. With the Transparency Kit, it enables these three benefits, connectivity to the systems and to the devices, the creation of a database, and also the identification of initial actions. The use case is probably customer-specific, right?
Exactly, definitely. Exactly, Bosch Rexroth knew relatively well what they wanted, namely to have the productivity of their machinery comparable. There are other customers who may not yet know exactly what they want to do with their machine data. That’s where we go, collect the data, and then try to explore it. Then you can also set different directions on how you want to optimize your production. Is this data that can go into any AI models? Do I have to worry about the fact that this machinery, as it is configured right now, is not optimal. These are all issues that can be addressed through exploratory analysis.
[26:24] Transferability, scaling and next steps – Here’s how you can use this use case
If you’re listening and thinking, ‘Hey, I actually have a different use case,’ please let me know. Feel free to post it in the comments, also on LinkedIn under the podcast post if you’re listening. I would be very interested to know what use cases you are implementing. Philipp, last question for you today. What else can we look forward to in the future? What is still to come? What are you working on right now?
Yes, gladly. In the area of data collection, what is certainly also always exciting is the further development of the individual IoT platforms. There is always added functionality, where we have certainly not yet reached the end of the line. Especially regarding the topic of communication protocols and network, with a focus on 5G connectivity, which will be introduced into the facilities in the future. Also in the future, a big topic will be how I collect data to push topics like artificial intelligence. What data do I need to collect to have a good training data set? We want to provide support at this point, but we are lacking a data set here, so we can use the Transparency Toolkit. These are the cases we want to cover and then you do get a nice solution.
Yes, very nice. This is also a significant driver for many of the use cases that you are already implementing today, further leveraging the potential with AI You can see now what leverage you suddenly have with ChatGPT and the various solutions once you use it. And it’s good to see that you’re continuing to invest in it and that you’ll be adding more functionalities in the future.
Thanks so much for joining us today and sharing some specific use cases. This is not a matter of course, but today we discussed the example of determining OEE at the Bosch Rexroth plant using the ITK Transparency Toolkit. We also covered the topic of retrofit for data collection in general and the use of energy data as a use case, which you brought up again at the end. Thank you very much for this practical episode. I’m very pleased that you were here today. Thank you, Philipp. I now give the last word to you.
Many thanks to you as well. I’m glad to have been here today. I’m excited to see what the future holds.
Yes, me too. I think there are many more cool cases out there. Thank you very much. Bye. Ciao.