Consistently Inconsistent: Ai-Analyzed Patterns in Inconsistency-Situations of Cyber-Physical Systems Development

Inconsistencies are a critical yet underexplored challenge in the engineering of Cyber-Physical Systems (CPS), where interdisciplinary collaboration, diverse toolchains, and evolving development processes often lead to misalignments across models and domains. Understanding such inconsistencies is difficult, as the engineering situations remain unclear and are mostly documented in small and semi-structured datasets. This paper includes a novel, reproducible methodology for extracting and analyzing semi-structured, visually represented data using Generative Artificial Intelligence and introduces patterns of inconsistency found by applying this method. Seven recurring inconsistency patterns were identified, spanning tool limitations, process misalignments, and interdisciplinary collaboration issues. The patterns are validated through an assessment including voices from 21 systems engineering experts.