Dr. Christoph Müller and Dr. Kevin Cremanns | © CADFEM Germany GmbH

Kevin, artificial intelligence and simulation. Everyone is talking about it - what is the current state of affairs from your point of view? Has AI really already had an impact in the field of simulation in product development?

Absolutely! There are various areas where AI is already having a major impact. Mostly to speed up simulations, but also to gain a deeper understanding from the large amounts of data generated by both simulation and experimentation. Topics such as optimization and predictive maintenance are fields of application where companies are investing in AI and launching projects.

There are countless tools and methods for artificial intelligence and machine learning. Your solution is called STOCHOS. What makes STOCHOS so attractive for use in conjunction with simulation?

There are three main reasons for this: For one, the amount of data available for training AI in the field of simulation and related fields such as real experiments is comparatively small. Google algorithms and large neural networks need much more data. With STOCHOS, we have developed a system that can handle this relatively small amount of data in the AI world.

That sounds logical. What other reasons are there?

There's also the applicability of AI. With AI, you have to know what you are doing. A simulation engineer would need to work hand in hand with an AI expert. STOCHOS was designed for non-AI experts and - to put it bluntly - works at the push of a button. This means that SMEs without AI experts can also use the technology. The third advantage of STOCHOS is that the algorithms run on conventional hardware, i.e. locally.

For all its simplicity, the leap for a simulation engineer from their usual physics-based way of working to using a supposed black box like AI is huge, isn't it?

We often encounter engineers who are suspicious because, as you say, they suddenly have a black box. Of course, this requires persuasion and positive experiences. Reliability and trust grow when the results of the simulation match those of the AI. A validation calculation is particularly advisable for projects that are relevant to reliability. Our software offers the possibility of displaying the influencing factors that the AI has learned, which makes the black box behaviour more transparent.

Another special feature of STOCHOS fits in with this aspect, doesn't it?

Exactly. Because at Stochos, we don't just do machine learning, we do probabilistic machine learning. This means that with our AI, you always get a statement about how safe or unsafe the AI is at any given time. This is made possible by combining the advantages of two methods, neural networks and Gaussian. This is what sets STOCHOS apart and makes our solution unique.

Watch the full interview with Kevin Cremanns on our YouTube channel now!