Integração de telemática e machine learning para determinação das condições de transporte de cargas

• Ano de defesa: 2023
• Autor(a) principal: Glehn, Fábio Ribeiro von
• Orientador(a): Fonseca, João Paulo da Silva
• Banca de defesa: Fonseca, João Paulo da Silva, Fagundes Neto, Marlipe Garcia, Soares, Anderson da Silva, Santos, Max Mauro Dias
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Engenharia Mecânica
• Departamento: Escola de Engenharia Elétrica, Mecânica e de Computação - EMC (RMG)
• País: Brasil
• Palavras-chave em Português: Telemática; Caminhões pesados; Economia de combustível; Análise de dados; Aprendizado de máquina
• Palavras-chave em Inglês: Telematics; Heavy-duty vehicles; Fuel saving; Big data analytics; Machine learning
• Área do conhecimento CNPq: ENGENHARIAS::ENGENHARIA ELETRICA::MATERIAIS ELETRICOS
• Link de acesso: http://repositorio.bc.ufg.br/tede/handle/tede/13742

This study explores the application of telematics in fleet management to control fuel consumption and pollutant emissions. Beginning with analyzing the economic and environmental impact of diesel usage, the paper identifies the challenge of reducing these effects and highlights telematics and machine learning as possible solutions. Literature reviews validate this approach, demonstrating that combining telematics and machine learning optimizes fuel consumption, identifies driving patterns, and enhances maintenance. The applied research uses vehicle data available in the trucks' communication networks and telematics devices connected to the SAEJ1939 network. This data formed three separate databases: the first was collected from a fleet of 10 trucks belonging to a fuel transporter and the other two from Geotab® equipment installed in two trucks belonging to different transporters: one of sand and the other of food products. The study developed a model with eight different machine learning algorithms to determine whether a truck is loaded or empty. After validating the data, the final analysis after 10 runs of the code revealed, with the first database, an accuracy of over 85% for routes over 1 km or more than 120 seconds of movement. With the Geotab® data, a dashboard was built that made it possible to monitor the daily behavior of the two trucks with the creation of maps, graphs, identification of places of interest and selection of frequent stretches and performance indicators that include the payload transported.