MBSE Summit: How AI can increase the acceptance of MBSE

This year's MBSE Summmit in Traunkirchen was extremely exciting and characterized by top-class presentations and good discussions. However, the focus of this article is the result of the breakout session on AI and MBSE.

MBSE Summit

The MBSE Summit differs from other events in many ways. The most significant difference is the venue, Traunkirchen in Austria. On the one hand, the location is fantastic and gives you a vacation feeling. But that doesn't make the journey there easy and, above all, time-consuming.

Another difference is the format: the event starts in the afternoon of the first day without any presentations. Instead, the evening is dedicated to networking, first with a joint boat trip, followed by dinner and the subsequent "Fireside Chats"

On the second day, presentations followed in a single track. This time, a large number of high-caliber speakers from the INCOSE scene were represented, such as Sven-Olaf Schulze, Sandy Friedenthal, Chantal Sinnwell and Robert Karban. I particularly enjoyed the exciting presentation by Annika Meijer-Henriksson for the development of military fighter jets.

The final session was an open space, which this year was offered on the four topics of SysML v2, Digital Twin, Cybersecurity and Artificial Intelligence, all in the context of MBSE. I moderated the session on AI. Each moderator was allowed to present the topic in 5 minutes with a few slides.

Sandy Friedenthal: OOSEM with SysML v2

Sandy Friedenthal is co-chair of the OMG Systems Modeling Community (SMC), very active in INCOSE and one of the well-known drivers of SysML v2. It was a pleasure to meet Sandy in person for the first time.

In his presentation, he showed how SysML v2 improves the OOSEM development method. The higher precision, expressiveness and interoperability as well as new functions such as variant modeling make OOSEM much easier to use. In cooperation with the INCOSE OOSEM Working Group, OOSEM is currently being adapted to SysML v2.

Breakout session on AI & MBSE

I showed the following slide deck in preparation for the breakout session. My key points were:

• We have many challenges in product development, driven by competition, complexity, compliance and technological change
• We have been trying to solve these problems with various tools for some time now. AI is not the solution, but it can support solutions.
• It is important to understand that the publication of ChatGPT has led to a paradigm shift: Understanding instead of structure, learning instead of rules. The processes no longer lead, but often the tools do. And what a few years ago Science Fiction has now become the norm.
• AI can only help to a limited extent with the existing problems, but most of all with the tools.

Question

After this brief introduction, I posed the following three questions:

• Can AI unleash the potential of MBSE?
• If so, how?
• Are there already products that try to do this or even already do it?

There were around 30 participants in my breakout session. Together we expanded the list of questions. In particular, we included risk as another important topic:

The most important use cases

We collected a larger number of use cases. We later deduplicated this list and gave each participant the opportunity to identify the most relevant use cases by a show of hands. The exciting thing was that no one responded to many of the use cases. The following "top three" remained:

1. AI for communication and understanding (D)

15 people voted for this use case, under which we had summarized 5 points:

• Reduce the complexity of tools and data access
• Support communication
• Generate views on demand
• Joint cooperation on the model
• Involve AI experts for various disciplines

In particular, we had marked Chantal's last point (Siemens) as a "double plus": Here the idea is that specialized AIs with expert knowledge on risk, mechanics, etc. participate in the collaboration, whereby not only one but also several AIs are involved in the exchange and discuss with each other (and the humans).

None of the participants knew of a tool that could already do this for models. We excluded approaches in which models are first converted into text before they are given to the AI. Such attempts do exist, but they are not particularly promising, at least at the moment.

2. improve model quality (B)

12 people voted for Improve model quality, which we had summarized as "Find gaps and contradictions".

We almost summarized this with "Improve requirements (A)". AI has long been used to improve requirements, which a quarter of the participants knew or already used. However, even though requirements often have traceability to the model, the challenges of this use case for models are much more demanding.

3. migrate documents to models (H)

10 participants voted for this use case. The emphasis on "migrate" is important here: The aim should be that the documents can be disposed of afterwards, i.e. there should be no redundancies or duplications.

The interesting thing about this use case is that Dalus already promises to do just that for customers in a matter of days or weeks.

Conclusion

The MBSE Summit this year was worthwhile and a lot of fun for me. If I can make it, I will attend again next year.

The breakout session on AI & MBSE also provided me personally with new insights. The idea of disciplinary AIs was completely new to me and I think it is very promising. Even though it was rated rather poorly with "only" 6 votes, I did not yet have the use case "AI as MBSE teacher" in my collection either.

I am already looking forward to next year's MBSE Summit and I am sure that the other participants will feel the same.

The raw data