2026-10-202026-10-202026-06-05MARTINS, Edilene de Sousa. Geração de modelo via rede neural artificial para descrever o afundamento de trilha de roda. Orientador: Gérson Jacques Miranda dos Anjos. 2026. 63 f. Trabalho de Curso (Bacharelado em Engenharia Civil) – Faculdade de Engenharia Civil, Instituto de Tecnologia, Universidade Federal do Pará, Belém, 2026. Disponível em: https://bdm.ufpa.br/handle/prefix/10391. Acesso em:.https://bdm.ufpa.br/handle/prefix/10391Road infrastructure is an essential element for Brazil’s socioeconomic development, with flexible pavements representing the predominant solution due to their adaptability and lower initial costs. However, one of the main problems faced by this type of structure is rutting, which compromises road safety, user comfort, and pavement durability. Traditional design methods, such as DNER/1966, have proven insufficient to predict this phenomenon, which motivated the adoption of the mechanistic-empirical method MeDiNa. Nevertheless, the full application of MeDiNa is limited by the high cost and complexity of the required laboratory tests. In this context, this study aimed to develop an Artificial Neural Network (ANN) model capable of predicting rut depth (ATR) based on traffic, base, and subgrade parameters, using a dataset of 1250 simulations generated by MeDiNa. The model was trained using the backpropagation algorithm, validated with statistical metrics (R², RMSE, SSE, MAPE), and compared to multiple linear regression models. The results demonstrated the ANN’s high performance, with R² ≈ 0.998 and minimal dispersion relative to the 1:1 trend line, outperforming traditional linear models. Sensitivity analysis indicated that structural coefficients Ψ4 and Ψ2 were the most influential parameters, followed by Ψ3, while traffic (N) and the Poisson ratio showed secondary influence. It is concluded that ANN constitutes an efficient alternative to MeDiNa, offering methodological and practical contributions to pavement design, particularly in regions where laboratory tests are limited.Acesso AbertoPavimentos flexíveisAfundamento de trilha de rodaMeDiNaRede neural artificialDimensionamento mecanístico-empíricoFlexible pavementsRuttingMeDiNaArtificial neural networkMechanistic-empirical designCNPQ::ENGENHARIAS::ENGENHARIA CIVILTrabalho de Curso - Graduação - MonografiaAttribution-NonCommercial-NoDerivs 3.0 Brazil