Estudio de la señal del sensor de oxígeno y de los ajustes de combustible de corto y largo plazo debido a la influencia de las operaciones de conducción basados en arquitecturas de aprendizaje autónomo.

Abstract

The present research deals with the application of machine learning architectures for the identification of different driving operations and their effects on PIDs due to the influence they have on engine behavior. Through a bibliographic review it is justified that among the most used PIDs for the diagnosis of injection systems are the signal of the Oxygen Sensor and the short and long term fuel adjustments. These variables are altered by the environmental characteristics of high altitude cities such as Cuenca, so the standard data and the dynamic behavior of the signals are not the same as at sea level, which is why we seek to determine the normal behavior of those mentioned. PIDs during different driving operations such as starting, acceleration, deceleration and cruising speed. For this purpose, it is shown that the application of different machine learning architectures such as fuzzy logic, neural networks, vector support machines, K-means and classification trees serve for the detection and classification of different driving maneuvers.