Data forms the foundation of AI and smart manufacturing. But before companies can truly learn from it, they must first consolidate the information coming from various machines, sensors, and systems. This is precisely where one of the biggest obstacles for many industrial companies still lies. In this interview, Prof. Dr.-Ing. Mario Studer, professor of simulation and design at OST (University of Applied Sciences of Eastern Switzerland), explains why the institution is testing new approaches to the smart factory in collaboration with AWS, why domain expertise remains indispensable despite AI, and why not all production processes should be digitized to the fullest extent.
About the individual and the University of Applied Sciences of Eastern Switzerland (OST)
Role at the Institute of Materials Technology and Plastics Processing
Prof. Dr. Mario Studer: At OST, I am head of the Simulation and Design division within the Institute of Materials Technology and Plastics Processing. My work is divided roughly equally between teaching and research. In my teaching, I focus in particular on simulation methods and am interested in quality prediction. Within the institute, we also work on numerous industrial and research projects.
Applied research in collaboration with industry
As a university of applied sciences, OST has a strong practical focus. What does this mean in concrete terms for your work? We emphasize applied research in collaboration with industry. Many projects are carried out in partnership with Innosuisse and, in particular, with SMEs. Our goal is to strengthen companies through applied research. We therefore maintain very close ties with the industrial sector.
From the smart factory to data integration
The starting point and the real challenge
In collaboration with AWS, you implemented an initial project in the field of the smart factory. What was the starting point? We wanted to explore how data could be aggregated in a more structured way using cloud solutions and how to derive insights from it as efficiently as possible. The project was initially launched by my colleague, Professor Roman Hänggi, in collaboration with Christoph Schniderig from AWS. Part of the initial project has already been implemented, while another part is still underway. For us, the first step is to familiarize ourselves with the various tools and services and gain experience in using them. In a second phase, we hope to translate these approaches into an “innovation factory.” In your opinion, what has been the real obstacle on the path to the smart factory so far? We had already developed various approaches, but these were sometimes very isolated from one another. Added to this were disparate data sources and local silos. The real challenge was to meaningfully consolidate data from different machines and systems. That is precisely where a large part of the manual work still lies today. Before we can truly learn from production data—and, for example, identify correlations or quality issues—the data must first be harmonized and linked together. As part of the joint project with AWS, we addressed this issue for the first time in a highly targeted and structured manner. It is essential for us to further automate the process of transforming raw data into actionable insights and thereby arrive more quickly at actual knowledge. What exactly do you mean by “data mapping”? By that I mean the merging of production data from different machines, plants, and systems. This data often comes in a variety of formats and structures and must first be harmonized and linked together. In practice, this poses a considerable challenge. In many of our projects, we’ve spent about 80% of our time just establishing high-quality “data matching.” Only then can we begin to truly leverage the data and draw insights from it.
Benefits of the cloud, reservations, and limitations of portability
Is this precisely the greatest added value of cloud platforms and related services? I would say so. For companies, the decisive factor is the effort they must expend before they can even begin to draw insights from their data. If we can reduce that effort, digitization becomes significantly more economically attractive. That is precisely why we also want to show that cloud solutions can be an option for small and medium-sized businesses. But it’s precisely when it comes to industrial data that we often see some reluctance toward the cloud. Do you see the same thing? We work on many projects involving confidential data. Companies are therefore very concerned about where their data is stored or processed. This is also why companies remain hesitant about cloud solutions, even though they use other cloud-based services. That’s also why it’s important for us to gain hands-on experience and be able to present different options. Can a solution developed once be easily adapted for other manufacturing companies? No, it’s not that simple. Each product and many of the underlying processes are highly specific. Depending on the production line, different sensors and quality criteria are used. The basic framework can be adapted. But you always need experts who understand the specific process and know which signals and quality attributes are relevant. Installing a “black box” and assuming that everything will work from there is not a realistic approach. Technology can simplify processes and draw insights from data. But it does not replace expertise. What role can OST play in this regard for SMEs? We can provide support as a strategic partner. In the plastics industry in particular, we have industry- and process-specific expertise that we can combine with technological solutions. At the same time, we aim to compare different approaches and analyze which systems offer advantages and disadvantages for which requirements. However, we cannot and do not want to develop all technological capabilities on our own. That is why partnerships—such as the one we have established with AWS, for example—are essential.
The “Innovation Factory” and the Transformation of the Higher Education Institution
The “innovation factory”: a testing ground for students, service providers, and industry
The next step is to create an “innovation factory.” What is the idea behind this project? The “innovation factory” must fulfill several functions at once. Our students must work in a factory of the future that is as close to reality as possible and become familiar with technologies directly on the machines and within production processes. At the same time, technology providers will be able to test their solutions under realistic conditions. These may include cloud technologies such as those offered by AWS, as well as machine vision systems or new measuring devices. This gives industrial companies the opportunity to discover and evaluate these technologies in practice. Finally, we can apply the results of research projects in this environment. For example, we can start by developing solutions in our laboratories and then transfer the results to the factory environment. Can such an environment also help bridge the gap between research and industrial application? In some cases, certainly. However, there are various reasons why research projects do not lead to industrial implementation. This is particularly the case for long-term projects, where the economic context may change. At the same time, technology is evolving so rapidly today that a project may, at some point, take too long to deliver the benefits initially anticipated. The “innovation factory” can help test developments under conditions close to real-world practice and bring them to light more quickly. But of course, it does not solve all the problems associated with technology transfer.
Higher Education in Transition: Fundamental Principles, New Tools, New Educational Programs
How should higher education institutions adapt in the face of such rapid changes in technology and industry demands? This is a major challenge. I believe, however, that foundational knowledge remains essential. Physics doesn’t change simply because new AI tools are available. Students must understand the fundamental principles in order to use these new systems effectively and evaluate their results critically. On the other hand, it is the way we use these tools that must evolve. We need to ask ourselves more about what students will need to be able to do on their own in the future, about the tasks that AI can handle, and about how we will continue to assess their knowledge. If, for example, a report can be written using AI, the question inevitably arises as to what such work still reveals about a student’s actual understanding. In my view, the mission of higher education is not diminishing, but evolving. We must continue to teach the fundamentals while equipping students with the skills to use new technologies competently, as well as to question and contextualize their findings. Does this also change the range of programs offered? At OST, new programs are emerging, including, for example, the “AI-Transformation” curriculum. In computer science as well, new specializations are emerging, and in mechanical engineering, topics such as data analysis are playing an increasingly important role. In principle, one might even ask whether students first need some kind of basic training in the use of these new tools before going on to acquire the fundamentals of their discipline with the help of these tools. We’re not quite there yet. Will this bring universities and industry even closer together in the future? Not necessarily. In some cases, the opposite might even happen. Thanks to AI, many companies are developing their own capabilities that they may not have had before. As a result, they have less need for support from universities in certain areas. Our strength still lies in infrastructure, process expertise, and domain-specific knowledge. Added to this is the breadth of a university education. Those who work directly in companies often develop their skills in a highly focused manner within a specific field of application. A university curriculum, on the other hand, offers a broader perspective on various topics, methods, and contexts over the course of several years. Industry will, in fact, continue to need this long-term perspective.
AI in Manufacturing: Opportunities, Limitations, and Outlook for 2035
Where Is AI Overestimated—and Where Does Digitalization Create New Risks?
In which areas are the possibilities of AI in manufacturing currently overestimated? For example, when it comes to the reliability of systems. In our project as well, we’ve found that while the systems deliver results, they can certainly contain errors. Without domain knowledge, it’s possible to miss these errors. That is precisely what is critical. At the same time, we must not underestimate the capabilities of these systems. When they have sufficient data, they can analyze far more information than a human and identify correlations that we might not see. The next question, then, is: do we really need all this information? So, does a greater amount of data not automatically imply greater utility? Exactly. This can be clearly illustrated, for example, with simulation. Computing power continues to increase. Yet a simulation doesn’t necessarily take less time. On the contrary, we’re simulating with ever-greater detail and precision. The richness of the information increases, but this must also translate into an economic benefit. The same applies to measurements. The more precisely we measure, the more discrepancies we discover. This raises the question of whether these discrepancies actually constitute a problem. New information also leads to new decisions, and each of these decisions entails effort and costs. Can ever-increasing digitization even create new risks for production? Yes, because networking creates new dependencies. For example, a machine might be functioning perfectly from a technical standpoint and continue to produce good parts, but it may stop transmitting data due to a software error. If the entire production cell depends on it, production may still come to a halt. For complex systems, I see significant added value in digitization and automation. On the other hand, for simple systems, it’s worth asking whether the extra effort is really worth it. We shouldn’t overdo digitization.
Cost-benefit analysis and the Swiss industry in 2035
How, then, can companies decide where the effort is worth it? For new systems, I would take the relevant data and structures into account from the very beginning. That’s where changes can be made at a relatively low cost. For existing systems, the situation is different. We need to ask ourselves very seriously whether it’s worth carrying out a complete modernization just to have every conceivable piece of data. If a product and a process are functioning without issues, there’s no need to automatically invest large sums of money just to collect additional data. On the other hand, there are now production processes so complex that we can barely grasp the big picture without digitization. That’s why there’s no one-size-fits-all solution. In conclusion, looking ahead to Swiss industry in 2035: which skills will remain crucial? Traditional technical skills will remain important. In sectors such as the plastics industry, in particular, we need people capable of setting up machines and manufacturing tools. Without this expertise, no production will be able to function in the future. In my opinion, it is primarily the process landscape that will evolve. Thanks to new AI tools, specialists can automate a much larger portion of tasks themselves, implement their ideas more quickly, and demonstrate what a process might look like. Previously, this often required the involvement of specialized IT professionals. Professional implementation may still require some support. But the path from idea to a working solution will be significantly shorter. Is the “innovation factory,” then, also an attempt to replicate today a part of tomorrow’s production? These are precisely the kinds of questions we want to explore there. We manage production processes ourselves, which allows us to observe new technologies in a realistic environment. At the same time, of course, we don’t know today what production will actually look like in five or ten years. That is precisely why it’s important to have a dynamic environment in which students, researchers, and industry stakeholders can test new approaches.
