Automação da elaboração de quadros de cargas de projetos elétricos residenciais utilizando visão computacional
Data
Autores
Título da Revista
ISSN da Revista
Título de Volume
Resumo
O presente trabalho apresenta o desenvolvimento e validação de uma aplicação computacional capaz de automatizar a elaboração de quadro de cargas em projetos de instalações elétricas residenciais, a partir da análise de imagens de plantas baixas. O sistema integra técnicas de visão computacional, redes neurais convolucionais e algoritmos de processamento de imagens, combinando-os com regras estabelecidas pela NBR 5410. A solução proposta utiliza uma primeira rede neural para identificação e segmentação dos cômodos, associando cada ambiente ao seu nome por meio de Optical Character Recognition - OCR, e uma segunda rede neural para detecção automática de pontos elétricos, como luminárias e tomadas. Com essas informações, um algoritmo orientado a objetos realiza o cálculo das potências, a divisão dos circuitos e a geração do quadro de cargas final. Uma interface gráfica desenvolvida em Python permite ao usuário carregar a planta baixa, definir parâmetros técnicos e exportar o quadro em formato de planilha. Na etapa de validação da aplicação, foram realizados testes de usabilidade com plantas baixas reais e simuladas, comparando os resultados obtidos com o método manual tradicional. Os testes demonstraram redução significativa no tempo de elaboração, manutenção da precisão necessária e alta confiabilidade nos resultados, conforme avaliado pelos participantes. Conclui-se que a ferramenta desenvolvida é capaz de mitigar erros humanos, padronizar processos e aumentar a eficiência na elaboração de projetos elétricos residenciais, representando uma contribuição relevante ao setor de engenharia de instalações elétricas.
This work presents the development and validation of a computational application designed to automate the generation of load schedules for residential building electrical installation projects based on the analysis of architectural floor plan images. The system integrates computer vision techniques, convolutional neural networks, and image processing algorithms, combined with the technical requirements established by NBR 5410. The proposed solution employs a first neural network to identify and to segment building rooms, associating each environment with its name using Optical Character Recognition - OCR, and a second neural network to detect electrical elements, such as lighting fixtures and power outlets. Based on the extracted information, an object-oriented algorithm performs load calculations, circuit distribution, and the generation of the final load schedule. A graphical interface developed in Python enables the user to import floor plans, configure technical parameters, and export the resulting schedule to an spreadsheet. To validate the application, usability tests were conducted using real and simulated floor plans, comparing the automated results with those obtained through traditional manual methods. The tests demonstrated a significant reduction in execution time, while maintaining the required accuracy and high reliability, as reported by the participants. The results show that the proposed tool effectively reduces human error, standardizes procedures, and increases efficiency in residential electrical project development. Furthermore, the approach presents potential for expansion to other engineering fields.
This work presents the development and validation of a computational application designed to automate the generation of load schedules for residential building electrical installation projects based on the analysis of architectural floor plan images. The system integrates computer vision techniques, convolutional neural networks, and image processing algorithms, combined with the technical requirements established by NBR 5410. The proposed solution employs a first neural network to identify and to segment building rooms, associating each environment with its name using Optical Character Recognition - OCR, and a second neural network to detect electrical elements, such as lighting fixtures and power outlets. Based on the extracted information, an object-oriented algorithm performs load calculations, circuit distribution, and the generation of the final load schedule. A graphical interface developed in Python enables the user to import floor plans, configure technical parameters, and export the resulting schedule to an spreadsheet. To validate the application, usability tests were conducted using real and simulated floor plans, comparing the automated results with those obtained through traditional manual methods. The tests demonstrated a significant reduction in execution time, while maintaining the required accuracy and high reliability, as reported by the participants. The results show that the proposed tool effectively reduces human error, standardizes procedures, and increases efficiency in residential electrical project development. Furthermore, the approach presents potential for expansion to other engineering fields.
Descrição
Citação
SOUZA, Guilherme Leão. Automação da elaboração de quadros de cargas de projetos elétricos residenciais utilizando visão computacional. 2026. Trabalho de Conclusão de Curso (Bacharelado em Engenharia Elétrica) – Instituto Federal de Santa Catarina, Itajaí, 2026.
