Identificação de condições funcionais em pavimentos urbanos auxiliada por machine learning

Abstract

The intensive use of road in Brazil imposes on managers the responsibility of maintaining and rehabilitating the implanted structure. However, several municipalities do not have the resources to collect data and assess the status of their roads. Due to the difficulties in several locations for the maintenance of roads, this paper proposes a model for the identification of the functional conditions of urban pavements to assist decision taking by managers. The model involves a Machine Learning algorithm that, together with smartphone sensors, generate georeferenced images and indicate the condition of the pavement. Data from linear, angular acceleration and vehicle location were collected in the city of Fortaleza and processed together with the statistical parameters obtained as a function of acceleration values. It was realized that the model has the potential to become a useful tool for managers as a viable and complementary alternative to more expensive and time-consuming methods.