OPC UA Forge: Unified Namespace for Legacy Factory Data

Today, in this English episode of the IoT Use Case Podcast, a true insider in OPC UA talks about developments and challenges of this very relevant standard. With Casa Mendes Gonçalves, we have a Portuguese company from the food and beverage industry that is ambitiously underway on their digitalization journey, with all the related challenges. They share their learnings and successes very openly.
Our guests today are Jouni Aro, the CTO from the Finnish company Prosys OPC and a member of the advisory council of the OPC Foundation, Luís Estriga, Head of Automation, and Vasim Tana, Automation Engineer, from the company Casa Mendes Gonçalves.
Do you think the collaboration works between a Portuguese and a Finnish company? Let’s find out. Hello and welcome.

Today a special “bom dia” for this episode of the IoT Use Case Podcast. I’m your podcast co-host, Dr. Peter Schopf, your favorite substitute for Madeleine Mickeleit, who is having a very special project for the next months. Not the digital twins we talk about quite often, but real twins—an entirely different set of challenges she will face. But before we meet our guests, we introduce ourselves in more detail. Luís, why should the listeners tune in until the end of today?

Luís

Hey everyone, I’m Luís Estriga, working for Casa Mendes Gonçalves, and I’m based in Portugal. You should tune in for today’s podcast because there are exciting technical details regarding OPC and digitalization of the industry. We’ll probably address AI as well at the end, so you should really tune in for this podcast today.

Great. And then we have Vasim. From your point of view, what do you want to share today?

Vasim

I think the most important thing is to talk a little bit about this relationship that we have with Prosys and how we developed the whole solution internally with their help and support. As Luís mentioned, we have some interesting projects going on, most importantly related to AI, data monetization, and the ease of getting all of this data across all of the machines in the factory. We’re going to share some interesting topics today. Hopefully, the listeners will enjoy it.

Perfect. And then, Jouni, last but not least, what do you want to share?

Jouni

I’ve been the CTO at Prosys OPC for 20 to 30 years, and we’ve been working on OPC UA for most of this time. I would like to emphasize the advanced parts of OPC UA that most people still haven’t been able to use, and focus on what OPC UA can enable in the long term—not just as a communication protocol, but as something much more.

Let’s start with you as well, talking about Prosys as a company and you as a person. You’ve been with the OPC Foundation as an advisor for quite some time, so that is quite interesting: being really close to that important aspect of digital transformation in factories. Maybe you can talk a little bit about yourself, where you’re located currently, and your company.

Jouni

We’re located in Finland, next to Aalto University, which is the technical university background of most of the people that work here. The company just celebrated 30 years, and I also joined around the same time I was finishing my studies.
We started working on a better future for the automation industry in general, bringing the best software development practices to automation software development. Then OPC came along around the same time, and we started taking benefit of that. Somehow, we continued that journey all these years and therefore became very close to the OPC Foundation, which has now worked 20 years with OPC UA standardization. Before that, there was OPC Classic, which is the strong basis for the development.
You could imagine that the OPC Foundation has been hosting working groups for all these 20 years. They meet every week and improve the standard, which is trying to cover a lot. Some people think it’s trying to cover too much. If you think about industrial automation, there are many topics that need to be resolved, and the more you can standardize, the better. That’s been our mission as well: to help the industry take advantage of this fantastic technology.
In the future, if everything is kind of ready for the standard, then you can start thinking about higher ideas, for example semantic interoperability instead of just communication interoperability.

That is also some commitment because you have Prosys OPC, you have “OPC” already in the name of the company. That is quite a dedication. How would you reply if I challenged you a little bit and said: “OPC isn’t that just a protocol?”

Jouni

The reason why people think like this is because, in OT, we tend to think that we need data from devices and machines, and you need some kind of protocol to communicate with these machines. Traditionally, it’s been a lot of legacy, vendor-specific communication technologies. The original mission of OPC was to standardize that so we don’t need to understand all those vendor-specific protocols.
There are successful products that convert all kinds of protocols to OPC and OPC UA, and that’s been very successful. But we haven’t really got much further than that, although the mission of OPC UA especially was to take things higher, not just be another communication protocol, but put together everything related to industrial automation.
We speak of current measurements exchanged through basic, tag-based communication, and then bring in events, alarms, historical access, and on top of this all kinds of information models that are more domain-specific. That’s the advantage of OPC UA: you can model your devices, your procedures, your systems. When these are standardized, we can also make machines interoperable with higher-level applications.
But unfortunately, products and the market are still concentrating on the basic level of communication and enabling communication-level standardization, whereas I would like to talk about higher-level communication and higher-level standardization.

One last question in terms of OPC UA before we go into the application phase with our other guests. When there are so many members—more than a thousand members—there’s lots of interest and probably lots of interest groups. How fast is the development? You said there are working groups every week, but can you come up with decisions fast enough for the market requirements to be addressed? How is it progressing, and in what direction is it progressing currently?

Jouni

The industry has a lot of different requirements and requests, and we see a lot of organizations looking at OPC UA as the solution. We can talk about the Open Process Automation Forum, the module type package defined by NAMUR and the PROFIBUS organization, and many different challenges the industry is facing.
In Germany especially, we have VDMA concentrating on information modeling for different domains. That’s where most of these new requirements come to OPC UA, especially related to the information modeling part.
In fact, the standard is moving forward quite fast and quite far, but then it boils down to the end users really needing these features, and products making these features available so we can take advantage of everything the standard provides. That’s the main challenge. The standard is further ahead and gives high promises, so we also see people disappointed when they still cannot use all these standardized features.
That’s the reality: products are always lagging behind, and the market is lagging behind. That’s why this development takes decades instead of years in practice, before we get everything in place in factories.
So you say that the standard is already enabling much more than many of the products. That’s why we have this new product that we will talk about later as well.

Very good. Let’s come to the market and to specific requirements, because I’m happy to hear that the standard is already providing more and it is not the standard that is the bottleneck, but rather implementing it with products.
Luís, you have a very interesting company in Portugal. Can you talk about yourself and your company? What are you doing? And potentially already: why did you decide to start digitalization? What was the initial motivation?

Luís

I’ve been working for Mendes Gonçalves for the past five years now. Mendes Gonçalves has been in the market for around 40 years, producing sauces like ketchup, mayonnaise, mustard, and all these kinds of things you see every day in the supermarket. We produce hot sauces as well and some other different projects.
When I joined back in 2020, there was little software: one machine where we were collecting data. We were outsourcing it to a provider. It was not really what we were looking for because it was really manual. We had to fix and configure every single detail, and then we had all these other machines communicating in their own dialect. We found it quite hard to integrate new machines into this solution.
More than a year ago, we decided: why don’t we develop our own software, our own solution? So that’s what we did. We started to look into the market to see the solutions that were there for us to use. We tried quite a few, and then we got to Prosys. When we tested it, we had no doubts: this is exactly what we wanted and what we were looking for.
We were not looking for a solution that was right away 100% fail-proof, but what made us choose Prosys was the support and the openness to meet with us, check what we need, and implement it. Each day, we integrate more machines and more data into our solution. We are collecting around 500 data points each minute, so that’s already a lot of data.
The main goal is to cut costs, optimize our factory, and take decisions based on data.

Can you elaborate a little bit more on how you looked into the market? Many listeners are probably in a similar position: they have solutions that are maybe not yet satisfying and they need to start looking for something else in a more structured way. How did you go about it? Did you have criteria that you used to rate different companies? Can you talk a little bit more about this journey until you got to Prosys OPC?

Luís

We were looking for a solution that was not super expensive—there are some super expensive solutions in the market. There are solutions that are plug-and-play, but then we have no control. We wanted to develop our own dashboards; we are using Grafana as well. We were looking for something stable. We were looking for a partnership that felt close to us.

That’s interesting that you say “close”, because you work with a Finnish company and you sit in Portugal. From my stereotype, you would say that Finnish and Portuguese might be somehow different. You confirmed several times that you are very collaborative and work very well together, and I’m happy to hear that. Maybe you can address this topic: how is this interaction with the Finnish guys? And what have been your highlights?

Vasim

Definitely the most important thing for me is the support that we received from Prosys. As Luís said, I have been working at Casa Mendes Gonçalves for one year as an automation engineer, so it is my first experience and interaction with the market and with all of these technologies.
Within this year, as we were searching for the right protocol, we contacted many different providers, and we got the idea right away that this seems right. It’s not perfect, but we can work with this. The support they gave all the time was really on point, so we decided to go on and explore with them.
We got to the point where we became super comfortable and we started even suggesting enhancements, or reporting whenever we found something not working well. If there was a small problem, we would do a small report by email or send the log files of the errors. Right away they would help us and try to understand what’s going on. The support team is even better than the OPC solution.

That’s great. This interaction is really important for such an early phase when you enter these uncharted waters of digitalization. Did you have any highlights yet in your daily production where you say: these insights we did not have before, or could not share with our colleagues on the shop floor? What has been happening there?

Vasim

We have many situations. For example, one of the main products of our factory is vinegar, and we produce somewhere between 80 and 100 tons of vinegar every day.
The vinegar yard is the oldest part of the factory, and that’s how the factory started 30 or 40 years ago. Since it has old equipment, it was quite difficult to modernize the data monitoring and visualization process. We created various dashboards using Grafana with the data that Prosys helped us to obtain. The integration was really simple.
Whoever works in factories and in automation understands how difficult it is to connect old equipment to modern solutions. Considering that, it was quite an easy task. Once we got the data from the machinery—old and new—we were able to set up alerts completely for free within Teams, or within any other applications.
Getting the data is the main challenge. Once we have the data, we can do whatever we want with it.
Another example is monitoring the consumption of the factory: electricity, water, steam, gas. We were never able to know how much we consumed until we got the bills at the end of the month. Once we got the data, we were not only able to know how much we consume, we were able to prevent leakage, accidents, or any dangerous situations, once we noticed excess electricity consumption in one place, or excess gas or water consumption.
One situation was with compressed air. Once we got the data and visualized it on Grafana, we noticed that every Saturday and Sunday we had excess air consumption. That didn’t make sense because these zones are not working on Saturday and Sunday. After a small investigation, we understood that the machine operators did not close the air valve. They left it open on Friday when they went home.
We were able to save 28,000 euros per year just because of the amount of air that was being leaked into the production lines while they were not producing. That’s one clear example of how obtaining data can help reduce costs and make the work environment safer.
The most important thing is not just to obtain the data, but to obtain it easily and efficiently. If you have a solution where you can obtain data but it’s always failing, it’s not efficient and it’s difficult to use, then it complicates the whole process. However, with Prosys it was quite a simple integration.
The manual was one of the easiest guides I’ve ever interacted with: how to install, how to configure the node, how to configure the communications, the IPs. It was straightforward. It wasn’t a 200-plus-page manual. It was direct: do step one, step two, step three, and you’re done.

Boiling it down into simple aspects and topics is very important. What you described with the leakage example: often when you talk about how we need to obtain data and invest in transparency, people ask: what’s the business case? It’s difficult to calculate, because such incidents and insights cannot really be predicted. It’s like little gold nuggets: you need to find many, and then you have your handful of gold. Normally the silver bullet is not there. It’s lots of different insights and optimizations based on stable data.
Jouni, in terms of your solution, can you explain a little bit more what has been employed at Casa Mendes Gonçalves? What has been the product, and maybe the advantages and the challenges you still see, because support is still required? What are the day-to-day challenges in the collaboration?

Jouni

What we’re talking about is Prosys OPC UA Forge, which is still a rather new application that we brought to the market a couple of years ago.
The challenge in the industry has been to get products that take advantage of advanced OPC UA features like information modeling and the security features. After looking for good solutions, we couldn’t really find any good enough and decided to make our own.
The main idea with Forge originally was to enable the industry to start using information models before they can use them within PLCs, within machines, within different systems, to enable integration of different machines to higher-level systems like MES. That was the original idea. But it turned out we were a little bit ahead of time.
The main challenge for customers is to see all the data they have in their production process and make it available in one single place. This relates to unified namespace, which is a great idea: simplify the complexity of factories that have a lot of data, but it’s scattered around in different applications, systems, and machines, used for local purposes for certain parts of control and machine functionality. It’s difficult to put all that information together.
I’ve discussed with many companies over the last five to ten years who have this challenge. They may have bought a lot of factories around the world, and each factory is completely different. Even if they produce the same products, they do it with different applications and machinery, and it’s very difficult to standardize processes.
OPC UA would enable streamlining and standardizing how you communicate from any machine to the namespace. But products don’t fully enable that even today.
Unified namespace is an idea that you put everything together in a nice hierarchical fashion where you can find all your production data. Then you can create solutions on top of that to understand what’s happening in your factory and pinpoint problems easily.
It sounds easy, but it’s really complicated because of the multitude of different systems, applications, and communication protocols. That’s the big challenge we are facing, and we wanted to create a modern solution based on OPC UA 100%.
The core of the application is an OPC UA server. We bring in information from everywhere to that same OPC UA server. It’s called an aggregation server: it aggregates data from different underlying systems. Within Forge, you can manipulate that data, raise the level of how you handle data, and convert it to higher-level semantic information, which you need to take advantage of OPC UA information models.
Most of the customers that have been really happy with the performance and what Forge enables have still had minor issues that are real challenges, and we have worked with them to improve those within the application.

I would have so many questions on how you implement it and the possibilities. However, looking at the time, we probably need to come to an end. From your perspective, Luís, I would be very interested: how do you want to continue developing the automation space? What are the next steps and the vision?

Luís

There’s no way around AI. At the beginning of this year, we already started to look into an AI solution developed 100% internally, with help from Forge OPC on collecting the data. The AI part will be developed by us internally.
We started to test some solutions, and now we are getting there. By the end of the year, we should have a chatbot working so every operator can use it. Every engineer can ask questions to our chatbot regarding our data.
In the future, maybe next year, we plan to cross different types of data like water consumption, what we produce in our production lines, and what electricity we are consuming. We want to cross that data and try to get insights that are not possible to see by the naked eye. Hopefully that’s our plan for next year.

Sounds great. Generative AI within an industrial setting is not easy to implement because of reliability topics, but coaching in terms of troubleshooting and making knowledge accessible is a very good way to go.
I’ll give you the last word. Thank you very much for sharing these insights and interesting topics. Vasim, do you want to start?

Vasim

The audience might want to know more use cases, so I can give a few extra examples.
One industrial-level example is alerting. In our factory, we have different equipment constantly being monitored by the quality team, such as the cooling chambers where we have our raw materials—eggs, tomatoes—since we are in the food industry. Monitoring the temperature all the time is quite important.
With the older system, the system would go down and we had to manually restart it to receive alerts. Sometimes there were bugs where the alert was triggered but the conditions were okay.
With the solution we created, we read the data from the cooling chamber sensors each minute and associate alerts in Grafana, which is open source. Then we get these alerts in real time at a much reduced cost because we’re not paying for each SMS. We have around 16 cooling chambers.
We did the same for vinegar. In the fermentation process, temperature is quite important. It needs to be between 28 to 30 degrees Celsius, up to 33 degrees Celsius. Higher or lower, the bacteria would die, and then you have around 10 or 20 tons of vinegar you cannot use anymore. Monitoring this is quite important.
Another part is monitoring. Before, if something happened during a weekend—if there was a problem, the machine was not working, or there was an error—operators had to leave their houses, go to the factory, and see what was going on on the machine HMI. Now we have dashboards. They just connect to their VPN from home, from their phones, and they can monitor and see what’s going on in real time, minute to minute. We could reduce it to 30 seconds, but for our industry and machinery we thought the lowest interval is one minute. So we’re getting data every minute.
Another example is analyzing large amounts of data. We have data from machines working 16 to 20 hours per day. Before, to analyze it at the end of the month, you had to download an Excel and create your own graphs. We can now do that with one click within Grafana. We saved a lot of time for our engineers, operators, and any team who wants to manipulate or use this data. They don’t need to download or extract it elsewhere. They apply any time filter they want and choose how they want to visualize the data, and it’s right there.
These examples can go on, but time is an issue.

Still great. Thanks a lot for these examples. Luís, your last words.

Luís

As a final takeaway: just start. In industrial digitalization, waiting for the perfect system design, the perfect tools, or the perfect timing only leads to stagnation. You will never start.
Real progress doesn’t come from planning the ideal architecture on paper. It comes from connecting the first devices, validating the first data points, and learning as you go.
Every mature digital factory you see today started with small steps: a few sensors, a basic OPC setup, simple dashboards. From there, the system evolved. The team gained insights, identified problems, improved cycle after cycle, and you discover what really needs to be fixed or optimized once the data starts flowing.
If you never begin, nothing will change. Once you take the first steps, the path forward becomes very clear.

Thanks. Jouni?

Jouni

I still want to mention semantic modeling. It doesn’t sound very fancy, but what is semantics? It’s the words and meanings we give to different things.
When you talk about AI, you cannot be very clever if you don’t know what you’re talking about. When we define semantic models and put AI working on top of that, there’s a much higher chance of getting something meaningful out than just putting AI on top of plain data that nobody knows what it is.
That’s where OPC UA, in my opinion, has a high promise: by standardizing semantic models and semantics within the industry, that can enable a strong basis for successful AI algorithms as well.

That’s a great point. Semantic modeling would merit an episode by itself to go into details. Thanks for your insights, and all the best for your continuous digitalization journey. All the best to the listeners, and see you in the next episode.