Estudo do uso de inteligência artificial como ferramenta de apoio à elaboração de memoriais de instalações na engenharia civil

Abstract

Generative Artificial Intelligence has increasingly expanded its presence in Civil Engineering, particularly as a tool to support the preparation of technical documents. In this context, the present study aimed to analyze the application of Large Language Models (LLMs) in the preparation of descriptive reports for building systems, evaluating their potential, limitations, and implications for the professional practice of civil engineers. To this end, a case study was conducted using the design of an Operational Base of the São Paulo State Highway Military Police. Descriptive reports for the building electrical and plumbing systems were prepared using four different approaches: the traditional manual method and three AI-based platforms, namely ChatGPT, Claude, and Gemini. The resulting documents were compared in terms of information organization, textual clarity, level of technical detail, compliance with applicable technical standards, specification of materials and equipment, and the time required for their preparation. The results demonstrated that Generative Artificial Intelligence tools significantly contributed to improving the efficiency of document production by increasing the speed of text generation and promoting greater standardization of information. However, the study also found that the generated documents require careful technical review, particularly regarding regulatory compliance, information accuracy, and adaptation to the specific characteristics of each project. It is concluded that Generative Artificial Intelligence is a promising tool to support the preparation of descriptive reports, provided that it is used under the supervision of the responsible engineer, without replacing technical expertise, critical analysis, or professional responsibility.