Estimating the effect of cushioners on fuel consumption with deep learning algorithm: An example application

Speed bumps are used to regulate traffic frequently. However, bumps also cause problems such as decreased driving comfort and increased fuel consumption. Nowadays, data can be accessed instantly from the OBD-II port in vehicles. In this thesis, it is aimed to determine the effect of vehicle's slowing down and accelerating over bumps on fuel consumption. For this purpose, an automobile is used on the routes with bumps and real-time speed and fuel consumption data is obtained from the OBD-II port with Arduino. The received data is preprocessed and normalized. Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) deep learning models are developed to predict fuel consumption. The preprocessed data is used to train the models. The sampling time of the data set is 300ms. The system is trained with the window slide method. The window size is selected as 5, 10 and 15. 60% of the data is used for training and 40% for validation. Hyperparameter optimization is conducted in the models. Thus, the number of layers and the units in the layers, the activation functions and the learning rate is specified. The lowest mean square error is obtained as 63‰ in the validation set. The effects of different speed scenarios on fuel consumption are predicted by using the models. In conclusion, the fuel consumption is increased between 16.30% and 31.03% during the impact of the bumps, by using the speed and time generated for the bumps.