Anlık yakıt tüketiminin makine öğrenmesi ile tahmin edilerek iyileştirilmesi

With the development of the internal combustion engine, our lives have changed significantly. Although their use has facilitated the development of humanity they have become one of the major contributors to environmental pollution and resource utilization. Although electrical vehicles seem to be the future of transportation, internal combustion engines will continue to be a part of our lives for the foreseeable future. There are a vast multitude of techniques to enhance a vehicles performance and economy, from engine modifications to aerodynamic changes. However, it is difficult to verify the changes without knowing the fuel consumption of the vehicle under various situations. Modern cars are very technologically advanced and rely on sensors and actuators which communicate with control units, therefore it becomes possible to obtain data by using the vehicle sensor data from the controller area network (CAN) bus. Due to its bus structure, it is possible to reach real-time detailed data from sensors inside the vehicle such as O2 sensor voltage, fuel pressure, catalyst temperature etc. This study aims to predict the instantaneous fuel consumption by collecting a large-scale vehicle sensors' data and create a model with machine learning algorithms with the goal of better understand how the multiple variables influence the instantaneous fuel consumption. With this predictive model, it is possible to make fuel experiments and receive improvement results easily such as inlet water injection, side mirror aerodynamic optimization etc. This approach and the experiments can also support original equipment manufacturers in developing and marketing this technology in the future. This work may lead the way to a cleaner environment due to more economical and less polluting vehicles.