Sistema de diagnóstico preditivo para veículos a combustão usando machine learning e deep learning.

Abstract

This work presents the development of a predictive diagnostic system for combustion vehicles, using Machine Learning and Deep Learning techniques. The objective is to create a model capable of predicting failures and assessing engine health. For this, data from automotive sensors, such as temperature, pressure, and rotation, were used, and data preprocessing and feature engineering techniques were applied for data optimization. ML models, such as Random Forests and Support Vector Machines, were compared with Convolutional Neural Networks and Recurrent Neural Networks, using time series to detect failure patterns. The validation of the models was carried out in a Hardware-inthe-Loop simulation environment, using Model Based Design, allowing the system to be integrated with real hardware for robust testing. This study reinforces the feasibility of applying artificial intelligence in automotive maintenance, highlighting the potential of technology to predict and prevent failures before they occur.