Redes neurais aplicadas à determinação do nível de serviço da BR-230.

Abstract

Artificial neural networks are a type of Machine Learning (computational model), whose structure resembles the network of neurons in the human brain. They are used in various areas of knowledge, including the transport sector, to obtain information more quickly and accurately. In this sense, the main objective of the work is to develop neural networks, using a matrix simulation software with package for artificial neural networks, that can determine the variation of the service level of the BR-230, in the stretch of km 20 to km 137,38, in the two lanes (JP-CG and CG-JP directions). For that, the stretch monitoring data, provided by DNIT, were evaluated to calculate the values corresponding to density and service level, using the DNIT methodology, considering typical traffic days, and to identify the critical times where there is a reduction in it on the highway understudy, therefore, with the data processed, different architectures for the ANNs were studied and, based on training and tests, it was possible to build a neural network capable of delivering results with an accuracy of around 95% and assessing the behavior the flow of the highway.