Coleta de dados de inventário e levantamento de defeitos com smartphones e tratamento por meio de aprendizado de máquina para uso em sistemas de gerência de pavimentos

Abstract

In the context of Brazilian cities, the need for constant Maintenance and Rehabilitation (M&R) of the road infrastructure becomes a priority for its proper functioning, development, and meeting the interests of its users in terms of comfort and safety, and Operating Cost of the Vehicle (VOC). In most of these cities, the management of the pavements is carried out by city halls, which often, due to the size of the network, the lack of specialized labor, and the use of conventional assessment methods, tend to not comply - or even fail - appropriate periodicity and M&R practices, especially at the network level. On the other hand, there has been, in the last decade, the development of new technologies to aid in the assessment of pavement conditions. The use of sensors in current smartphones has proven to be a viable option for the collection of data used in Pavement Management Systems (SGP), due to its low cost and high productivity. The exploration of tools available in these devices, such as GNSS (Global Navigation Satellite System) receivers, accelerometers, and gyroscopes, as performed in this research, aims to collect, classify, treat and analyze data using Artificial Neural Networks (RNA) for the identification of irregularities in the pavement, to compose an Urban Pavement Management System (UPMS) platform. Data collection took place at the Pici Campus of the Federal University of Ceará, with a passenger vehicle and a smartphone with the Android operating system. The proposed models were able to identify irregularities in the pavement, such as potholes, with an average accuracy of 84%, as well as different types of pavement, with an average accuracy of 98%. The results obtained, therefore, attest in favor of the technique as being a modern alternative or complement to the traditional methods of assessing pavement conditions regarding the aforementioned characteristics. Therefore, this work intends to contribute to , guiding the manager in the assignment and prioritization of critical or defective segments that require intervention or complimentary assessment, assisting in the most appropriate decision making regarding the strategies of M&R to be carried out.