Mr. Schlögel, when it comes to the topic of energy and resource consumption, the focus is usually on the end consumer, private consumption. In fact, industry consumes most of the resources, with AI and smart machines you want to change that. But how?
Artificial intelligence is currently on everyone’s lips. With ChatGPT, AI has stepped out of the background into the limelight because here, for the first time, an AI communicates directly with the general public as a chatbot. In the so-called “Industrial Internet of Things”, IIoT, it has been doing this for a long time.
So AI is “old hat” – how can you imagine that in production? I think first of robots.
Robots are symbols of Industry 4.0. For example, the welding robots in the automotive industry, where the degree of automation has been extremely high for some time. Today, the areas of application are much broader. AI is playing an increasingly important role in interlinked production lines where digitization is gaining ground.
It’s not about a single machine, but about the complete production, about a community of machines?
Of course I’m interested in the individual machines. But when it comes to where the biggest pain points are and how I can optimize the output, I have to look at the whole line. The question is also important: What do I want to optimize my line for? Because that depends on the situation. In pharmaceutical production, for example, it’s not about getting the last percentage points out of performance, it’s much more important that there are no errors in the production process. Quality and safety are the stipulations. On the other hand, when I produce kitchen rolls, it tends to increase performance.
This is not trivial. A machine needs sensors to even have an overview of the process.
Sensors have existed in the past and have been little used to relay data to the outside world. They primarily supported the control loop within the machine. Today we look at the complete production process, end-to-end, so from start to finish. This is where the “machine to machine” communication comes in, the following machine should “understand” what the previous machine did and what problems it had. Let’s take kitchen rolls as an example: The speed of production depends, among other things, on the quality of the paper I’m using. The worse the quality, the slower the machine has to run to avoid tearing the paper. And it’s all a matter of sensor technology: Does the machine recognize this and how quickly can it react to it so that I can achieve a stable production process?
Humans cannot process this amount of data. That’s when the fear or the uncanny begins that things are regulated and created without a human being involved.
That doesn’t have to happen. At Körber, we think it makes more sense to combine human experience and machine learning insights.
Man meets machine – how does that happen?
Let’s take kitchen roll again as an example. A tissue production line has up to 400 parameters that the machine operator can set. So far, he has mainly relied on his gut feeling and experience, but now the AI makes suggestions. Our AI software breaks that down from 400 to 40 parameters that have an impact, or just five that are relevant in the given situation. The operator can also reject suggestions, but then has to explain why. This is how the AI learns for future suggestions. And people benefit from the amounts of data and remain able to act. That is our basic understanding. AI is a tool that helps people make better, faster decisions.
But what are the effects? What are you doing this for? What potential for improvement do we have here?
Production processes can be optimized according to different aspects: output, energy efficiency or quality. If I want the highest possible quality, I might not get the highest performance. Or I optimize in the direction of performance, then perhaps the energy consumption is not so important to me. I’ll give you an example: With InspectifAI, we have developed a system that is used for the visual inspection of liquid medicines such as vaccines. Because safety is the top priority here, everything that even looks like a problem is sorted out there, such as a small scratch on the bottle. We also placed an AI system on top of the automated process that learns which deviations are really problematic and which do not have a negative impact on the drug. This enables us to “save” 85 percent of the bottles wrongly sorted out. This saves costs and energy at the same time.
Industrial processes consume a lot of energy and raw materials. But what we consume here in Europe is often produced abroad, and we make good money on it.
We have always taken sustainability seriously. For us as machine builders, however, it is becoming increasingly important what CO2 footprint our machines generate at the customer’s site – this is called Scope 3.
So you are also responsible for the resources your machine generates in, say, Pakistan? And that over the entire life cycle?
Yes, that goes into Scope 3. And that’s right, too, to show that you don’t just take care of what’s happening on your own doorstep, but also bear responsibility for what’s happening three doors down. From the climate perspective, it doesn’t matter where the negative output is generated.
Thanks to AI, the new systems are smarter, more efficient and also deal with the climate more responsibly as a result of regulations. That sounds good, but when will these new green facilities replace the existing production facilities?
We can’t wait for that. There are many systems that have been in operation for 20 to 30 years or even longer. The majority of factories worldwide are “brownfield” plants.
In other words, systems that were built before the “Green Age”.
Yes, exactly. For this reason, we develop AI-based software systems that focus on these old systems in order to then optimize output, quality or energy efficiency accordingly. The impact is much greater there than in new business. The good thing is that you don’t have to extensively rebuild the system. We read out the important data, send it to the cloud, the AI software evaluates it and sends the recommendations directly to the operator’s smartphone.
Are people still important here?
Yes, there is no other way with old systems, otherwise they would have to tear out all the technology and set up fully autonomous systems, which many medium-sized companies cannot afford. This is what makes this support via the cloud and AI so interesting, because it can be added to existing systems. I see this as democratization of AI technology. Not only those who are able to buy the latest machine generations benefit, but also those who are in the existing business. You asked what AI will do for the industry as a whole. That’s difficult to quantify, but if we optimize a system for output, we achieve performance improvements of 20 to 30 percent. It is important that we can extract as much data as possible from the machines and also share it from different companies.
And do the customers join in? Isn’t that a glimpse into the holy of holies?
Yes, this is a challenge that requires a lot of persuasion. Confidence in the security of the data is a prerequisite. However, a lot of the data does not concern any “bedroom secrets” of the companies. As in the example with the inspection and sorting of the bottles. I like to say: Dates are like love. The more love I give, the more love I get back. And the more data I give, the more optimization I can achieve.
