Recurrent Neural Network Algorithm for Dew Point Detection

The automotive industry is constantly seeking innovative solutions to address the challenges related to vehicle emissions and efficiency. In this context, condensation in the exhaust system represents a significant issue, as it can impact the performance and reliability of emission control systems. This thesis aims to tackle the problem of condensation in the exhaust system through the analysis and development of a machine learning model. The first chapter introduces the context and challenges associated with condensation in the exhaust system, also presenting a solution proposed by Bosch and a machine learning model to address these limitations. The second chapter describes the tool developed for droplet detection, providing an analysis of the EGS-Li sensor and presenting the tool implementation. The third chapter focuses on the recurrent neural network algorithm used to address the problem, detailing the construction of the Ground Truth, data preparation, and the use of K-Fold for dataset training and testing. The fourth chapter presents the obtained results, analyzing the architecture of the Recurrent Neural Network and showcasing various case studies. To conclude, the final chapter presents the conclusions and outlines future steps to improve the performance of the network.