A

Generating SysML V2 Models from Natural Language Requirements Using Large Language Models

IEEE International Symposium on Systems Engineering, pp. 1–7

Abstract

The rising complexity of automotive systems demands structured methods for system development. Model-Based Systems Engineering (MBSE) methods offer such a structured approach to manage complexity, but its adoption is hindered by the high effort of manually modelling complex relations in modeling languages such as SysML v2. Information needed to author SysML v2 models is embedded in requirements specifications, which are usually stated in natural language. While Large Language Models (LLMs) are capable of processing natural language to extract information and generate SysML v2 code, output quality varies widely without strong guidance. To address this challenge, this paper presents a structured approach for instructing LLMs to more consistently generate SysML v2 models. We evaluate the robustness of this approach through repeated model generation and compare it against other prompting setups with input of varying structure. Results show that using structured prompting substantially improves generating semantically rich and syntactically coherent SysML v2 models. This work highlights the potential of LLMs to support the adoption of model-based systems engineering by automating the transition from textual requirements to formal SysML v2 models.

Authors 4

  1. RWTH Aachen University

    Affiliation as printed

    Institute for Machine Elements and Systems Engineering (MSE), RWTH Aachen University,Aachen,Germany

  2. RWTH Aachen University

    Affiliation as printed

    Institute for Machine Elements and Systems Engineering (MSE), RWTH Aachen University,Aachen,Germany

  3. RWTH Aachen University

    Affiliation as printed

    Institute for Machine Elements and Systems Engineering (MSE), RWTH Aachen University,Aachen,Germany

  4. RWTH Aachen University

    Affiliation as printed

    Institute for Machine Elements and Systems Engineering (MSE), RWTH Aachen University,Aachen,Germany

Cited by 2 stored of 2

2 results

No patents citing this paper on Lens.org (checked 2026-10-06).

References 7

7 results