Modelo predictivo para la detección de fallas incipientes mediante aprendizaje automático y PID´s en el sistema de inyección de un Kia Sportage 2.0L

Abstract

This article describes a predictive study on internal combustion engines under real driving conditions using data from the OBD II port, where three scenarios are examined: normal operation, clogged filter, and spark plugs with a 1.3 mm gap, following RDE parameters. The K-Nearest Neighbors algorithm was used with relevant data determined by Random Forest, and the emissions of CO2, CO, NOx, and HC formed the basis of the model, achieving 98% effectiveness in predicting the presented faults.