Severin Diepold is managing director of Syskron GmbH. Syskron is the IT house of the mechanical engineering company Krones. Syskron has several locations wolrdwide. The company has two locations in Regensburg – in the business park and Maximilianstrasse.
Hello, first of all thank you very much for the interview and the opportunity to present my company in this context. My name is Severin Diepold and I am the managing director of Syskron GmbH. If one or the other does not know us yet, we are the IT house of the mechanical engineering company Krones.
As a managing director, my areas of responsibility are naturally widely spread. In addition to the operational business, it is my greatest concern to establish my company in the market and to orientate it towards the future.
At Syskron GmbH, we aim to support customers and Krones with the help of data-driven IT solutions. In this field, contact with big data and machine learning is inevitable. To give an example: We access the control data of our machines and evaluate sensor data in order to identify component failures at an early stage, often known as predictive maintenance. Thus, the customer is not surprised by unexpected faults, but can schedule them into the production plan. In this detective work we take the liberty to access the complete kit of algorithmic possibilities. In some cases, clever statistical analyzes meet all requirements, in other cases we use complex procedures from the field of AI.
We maintain close cooperation with the University of Regensburg. We are currently giving over 30 working students an insight into our various departments. In addition, two of our employees in the physics faculty are doing a PhD on questions in the field of AI.
For us as a company, it is of course always beneficial to get new wind and new ideas in our company. We always welcome other points of view. Not to be sneezed at, of course, is the opportunity to get to know highly qualified employees during their studies and to win them over for Syskron.
A new revolution is currently taking place in our industry, so far mechanical engineering has mainly sold steel and the highly developed mechanics and electronics to customers. Now the trend is that only services are sold. Specifically, this means: Guarantees sold that a machine achieves at least a certain performance. In order to meet these high expectations, data must be evaluated on a large scale and in an intelligent manner. The basics in AI are essential for this. It is important to convey concrete practical knowledge in addition to algorithmic knowledge. How do you go e.g. with noisy data or what options do I have if labels are missing.
At the moment we are dealing with various topics. The above example with predictive maintenance is only one of several. In another topic, we observe the dynamics within the line to find out the “scapegoat” of the line, which costs the most production time due to faulty behavior. In a very new exciting topic, we are developing algorithms in cooperation with Krones that intervene directly in the machine. This considerably simplifies the configuration of the machines, shortens the commissioning time and eliminates possible sources of error. So we are currently doing some exciting research projects.
Some of our research topics are currently related to AI. To give an example in the area of predictive maintenance: We are currently researching to have all sensors of the machine automatically monitored with an AI. This poses a wide variety of challenges, since the signals have completely different characteristics and, of course, the customer should not be overwhelmed with alarms.
I think the biggest challenge is to correctly classify media reports in the context of manufacturing industry. There is a lot of talk in the media about AI successes based on immense training data or computing power. For example, the successful AI in the computer game Dota 2 played against itself over 40,000 years until it could beat a human. Of course, our industry can only dream of such a lot of qualitative data and computing power. From this aspect I really appreciate the discussion with my colleagues, who can judge this algorithmically.
I find it most fascinating that an AI finds structures in data, which can then be found in reality. I find the meetings of our data scientists with the experts particularly interesting, in which the two worlds merge and new solutions are found.
We maintain exchanges with other companies at various levels. At the strategic level, for example, we are in exchange to get siggestions for our development process. At the algorithmic level, the exchange with other special machine manufacturers is enriching, since the challenges in the industry are relatively similar. We cooperate with Regensburg companies as well as companies worldwide.
I think the gratest opportunity and challenge for AI in higher education is the extensive collaboration of different faculties. An AI project can only be successful in a cooperation; it will require both the programming science, the mathematical knowledge and intuition for the algorithms, the explicit specialist knowledge from the respective research area as well as the philosophical foresight for possible consequences. This means that an introduction to AI should be included in every individual course. This is necessary in order to be able to assess the topic more realistically and to recognize potential in your own field. In research, it is then necessary that thinking in faculties is broken up and that researchers cooperate in cross-faculty projects. Only in this way can the University of Regensburg, as a pioneer, revolutionize many fields of research and secure the top position in the world market in industry.
“KnowledgeCreatesData” makes exactly the right contribution to carry AI into the various faculties and lay the foundations there. We would be happy to work more closely with “KnowledgeCreatesData” in the near future!
If you allow me, I would like to briefly advertise our newly founded Bavarian Industry association for Applied Artificial Intelligence. In meetups, we promote exchange between research and a wide variety of industries. We already had wonderful meetings in which we discussed, for example, embeddings or biases in big data sets. We’re always happy to see new faces!
Mr. Diepold, thank you for taking the time to answer these questions. We wish you a nice day!