IDP or Semester Project: Build an AI-Assisted Natural Language Interface for Multi- Physics Simulation

Institut

Lehrstuhl für Aerodynamik (TUM-ED)

Typ

Semesterarbeit /

Inhalt

theoretisch /  

Beschreibung

Project Overview and Goal
Would you like to work at the intersection of simulation, artificial intelligence, and software engineering? In this project, you will help make advanced multi-physics simulation easier to use
by developing a natural language interface that can translate user intentions into validated simulation settings.
 

The project focuses on the open-source Python simulator SPHinXsim, which provides an LLMand Python interface for the C++ multi-physics library SPHinXsys. Instead of manually writing
complex configuration files, users should be able to describe a simulation task in natural language and obtain structured, reliable settings that can be checked, edited, visualized, and
eventually used by the simulator.
 

What You Will Do
• Design a structured JSON-based domain-specific language for simulation settings.
• Explore how large language models can generate and modify simulation configurations.
• Use Pydantic to validate the structure of generated settings.
• Develop semantic validation methods that check whether generated settings are physically and logically meaningful.
• Create a workflow that supports both direct editing and AI-assisted interaction with simulation setups.
 

What You Will Learn
This project offers hands-on experience with LLM-based interfaces, JSON schema design, Python software development, validation strategies, and modern simulation workflows. It is well suited for students interested in computational engineering, scientific machine learning, AI-assisted software tools, or open-source simulation.
 

References: To explore SPHinXsys and SPHinXsim, please visit the repository:
github.com/Xiangyu-Hu

Möglicher Beginn

sofort

Kontakt

PD Dr. -Ing. habil. Xiangyu Hu
Raum: MW1636
Tel.: +49.89.289.16152
xiangyu.hutum.de

Ausschreibung