Desenvolvimento de um sistema para economia de combustível com rede neural artificial

Abstract

This article aims to develop a data acquisition board to identify the uphill and downhill stretches of a vehicle through an LSTM (Long Short Term Memory) artificial neural network, and to estimate the reduction in fuel consumption if the electronic injection was adjusted to take the information from the neural network and optimize the fuel cut on deceleration (DFCO - Deceleration Fuel Cut Off). The data acquisition board is composed of three microcontrollers that receive data from the vehicle's sensors. This data is used to train the LSTM artificial neural network, which is able to identify the uphill and downhill stretches in the route taken. From the neural network information, it is possible to optimize the vehicle's DFCO, adjusting the electronic injection map to improve fuel consumption. This is done through an algorithm that uses information from the neural network to estimate the reduction in fuel consumption that can be achieved with changes in the injection map. The results showed that the data acquisition board was able to identify the uphill and downhill stretches with an accuracy of 96.7%, and that the optimization of the DFCO based on the neural network information can result in a significant reduction in consumption. of fuel. This study contributes to the development of technologies that can improve the energy efficiency of vehicles, reducing environmental impact and operating costs.