MBSE and AI – Using the FAS Method
Today, I had a chat with ChatGPT. Since I’m giving an INCOSE webinar on FAS, later on, I tested how well ChatPGT can do FAS.
I provided some context in the first prompt but otherwise did not specify anything.
The result is amazing. The functional grouping, which is an architectural decision, is not bad. If you repeat the conversation, sometimes a different grouping comes out, which also has its justification. And if you bring in other framework conditions, it naturally also has an effect on the grouping.
Of course, you can still improve the use cases, etc., which also works if you provide more context or follow up here and there. The exchange with ChatGPT is not a series of function calls but a conversation.
Only the visualization at the end is not successful, and the textual notation of SysML v2 is not quite syntactically correct. But these are easily solvable problems.
It’s amazing how good the results are without fine-tuning or special training of the AI.
You will soon be able to read more about MBSE and the use of AI as an SME in the book AI Assisted MBSE with SysML by Doug Rosenberg, Brian Moberley, Joshua Hauswirth, and Tim Weilkiens.