20.500.12556/RUL-186803 Možnosti in omejitve uporabe velikih jezikovnih modelov na primeru dimenzioniranja jeklenega preklopnega spoja diplomsko delo Capabilities and Limitations of Large Language Models in the Design of a Steel Lap Connection Diplomska naloga obravnava možnosti in omejitve uporabe velikih jezikovnih modelov v konstrukcijskem projektiranju na primeru dimenzioniranja jeklenega preklopnega vijačenega spoja. Razvoj umetne inteligence v zadnjih letih odpira nove možnosti za podporo inženirskim procesom in avtomatizacijo teh, vendar pa odpira vprašanja glede svoje zanesljivosti in uporabnosti pri nalogah, ki zahtevajo visoko stopnjo strokovnega znanja. V teoretičnem delu naloge so predstavljene osnove umetne inteligence, delovanja velikih jezikovnih modelov in njihova uporaba v gradbeništvu. Poseben poudarek je na prednostih, omejitvah in tveganjih uporabe teh tehnologij v inženirski praksi. Prav tako so predstavljena načela projektiranja po Evrokodu 3, ki je evropski temelj za projektiranje jeklenih konstrukcij. V praktičnem delu pa je izvedeno dimenzioniranje jeklenega preklopnega vijačenega spoja v treh različnih obtežbenih primerih. Najprej so bili izračuni opravljeni ročno v skladu z določili Evrokoda 3, nato pa še z uporabo komercialno dostopnih velikih jezikovnih modelov ChatGPT 5.5 in Claude Opus 4.8. Rezultati so nato med seboj primerjani z vidika pravilnosti postopkov, natančnosti in preglednosti rezultatov. Namen raziskave je bil ugotoviti, v kolikšni meri modeli, ki so trenutno dostopni na trgu, lahko služijo kot podporno orodje inženirjem pri projektiranju konstrukcij. Na podlagi analize so nato podane ugotovitve o njihovi uporabnosti, zanesljivosti in omejitvah pri reševanju zahtevnejših inženirskih problemov. This bachelor's thesis investigates the capabilities and limitations of large language models (LLMs) in structural engineering design through the example of designing a steel bolted lap connection. The rapid development of artificial intelligence in recent years has created new opportunities for supporting and automating engineering processes. At the same time, it has raised important questions regarding the reliability and applicability of these technologies in tasks that require a high level of professional knowledge. The theoretical part of the thesis presents the fundamentals of artificial intelligence, the operating principles of large language models, and their applications in the construction industry. Particular attention is given to the advantages, limitations, and potential risks associated with the use of these technologies in engineering practice. In addition, the basic principles of design according to Eurocode 3, the European standard for the design of steel structures, are introduced. The practical part consists of the design of a steel bolted lap connection subjected to three different load cases. The calculations were first carried out manually in accordance with Eurocode 3 and were then performed using the commercially available large language models ChatGPT 5.5 and Claude Opus 4.8. The obtained results were then compared in terms of the correctness of the applied procedures, the accuracy of the calculations, and the clarity of the presented results. The aim of the research was to determine to what extent currently available large language models can serve as support tools for structural engineers in the design process. Based on the conducted analysis, conclusions are drawn regarding their usefulness, reliability, and limitations in solving more demanding engineering problems. diplomske naloge gradbeništvo umetna inteligenca veliki jezikovni modeli Evrokod jeklene konstrukcij preklopni spoj graduation thesis civil engineering artificial intelligence large language models Eurocode steel structures steel bolted lap connection true false false [P. Huč] Slovenski jezik Angleški jezik Diplomsko delo/naloga 2026-09-05 08:45:05 2026-09-05 08:45:14 2026-09-11 07:35:03 0000-00-00 00:00:00 2026 0 Ljubljana 2026 Ljubljana 1 spletni vir (1 datoteka PDF (VII, 45 str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.8:624.014.2(043.2) 181777 290737923 22881.pdf 22881.pdf 1 4EF72CAB0304CA796DEF3679F2207FFB 8cf87b0e52073202bdaab71a439b2e87198de0ef1bdbe7324698e772dcffd2fc 59eb6ef5-a8f5-11f1-8bc5-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=246173 Fakulteta za gradbeništvo in geodezijo 0 0 0