Detecção de adulteração de gasolina por etanol por meio da análise de imagens infravermelhas por algoritmos de aprendizado de máquina

Abstract

Fossil fuels-based vehicles represent the largest percentage of the world’s active fleet, therefore despite all of the issues gasoline is linked to, it is directly involved in people’s daily lives. There is effort made by the companies to reduce the cost of producing and distributing gasoline to maximize profit, and the use of illegal techniques is not uncommon. One example is the adulteration of that fuel with ethanol, which can lead to damage to the engine’s components when the machines are not properly equipped to handle ethanol. There is a need for an easy, fast and effective way to detect this solvent in gasoline-ethanol blends; the present paper aims to present a method based on infrared imaging (IR) processing, principal component analysis (PCA) and machine learning (ML) algorithms to classify images of operating engines in regards to the content of ethanol that was introduced as fuel blended with gasoline. The study concluded that Artificial Neural Networks (ANR), Support Vector Machines (SVM), K-Nearest Neighbors (KNN) and Decision Tree (DT) algorithms are capable of classifying the images with accuracies of, respectively, 94,6%, 90,8%, 94,7% and 97,2%.