Multi-agent retrieval augmented system for domain-specific knowledge in structural engineering

Thumbnail Image

Files

master_Ahmad_Hafiz_2025.pdf (1.14 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Engineering | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-09-29

Department

Major/Subject

Mcode

Degree programme

Master's Programme in Building Technology

Language

en

Pages

65

Series

Abstract

Large Language Models (LLMs) have shown remarkable capabilities in natural language tasks, but they struggle with specialized engineering knowledge like country-specific code provisions. This thesis addresses that gap by integrating the Finnish National Annexes to the Eurocode into an LLM-based assistant. We design a multi-agent Retrieval-Augmented Generation (RAG) system with two agents: one dedicated to retrieving relevant text from Finnish annex documents, and a second that leverages general Eurocode knowledge if the first agent finds no matches. We evaluate the system on structural engineering queries drawn from real design scenarios, comparing different model settings. Our results show that the RAG approach produces accurate, code-compliant answers grounded in the annex text, while the two-agent design guarantees a response for every query. Key findings indicate that smaller LLM variants stay strictly within provided sources by eliminating hallucinations at the cost of leaving some questions partially answered, whereas larger models offer more complete answers by drawing on general knowledge but occasionally introduce unsupported details. The thesis contributions include a novel LLM-based tool for automated code compliance assistance, empirical insights into model behavior with RAG, and a step toward safer AI integration in structural design workflows.

Description

Supervisor

Noureldin, Mohamed

Thesis advisor

Kortelainen, Petri

Keywords

LLM, large language models, RAG, multi agents, agentic RAG, retrieval augmented generation, artificial intelligence, generative AI, agentic AI

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202510208276

Collections

[dipl] Insinööritieteiden korkeakoulu / ENG