Feasibility of using AI language models for piping design in shipbuilding

Abstract

This thesis explores the feasibility of using large language models (LLMs), such as ChatGPT and GitHub Copilot, to assist in piping design within shipbuilding. The goal was to assess whether artificial intelligence can support engineers in interpreting standards, performing calculations, and applying design principles to freshwater distribution systems. The study integrated theoretical analysis and practical testing, evaluating both models using tasks based on ISO 15748-2 Ships and marine technology – Potable water supply on ships and marine structures, Part 2: Method of calculation. and typical ship piping layouts. The Results show that while LLMs can provide useful information, perform basic calculations, and explain design concepts, they struggle in realistic engineering scenarios and fail to produce fully correct outcomes. However, their potential value in documentation, data retrieval, and conceptual planning is clear, and continued development could make AI-based tools increasingly beneficial for marine engineering in the future.