Digital Twin-Based Fuel Consumption Model of Locomotive Diesel Engine

Abstract

In this study, we developed a digital twin (DT) model of a diesel engine in TÜLOMSAŞ. We estimated the fuel consumption of the engine using the designed DT model. For this purpose, we first created the physical model of fuel consumption. We measured the parameters of the physical model that can be measured directly or other parameters related to these parameters through sensors attached to the engine. We demonstrated that all the parameters of the physical model are essentially interrelated by examining the correlations between the observed data and fuel consumption. Using the measured data for fuel consumption, air consumption, rpm, and combustion temperature, we created two Artificial Neural Networks (ANN) with a single hidden layer and a double hidden layer. By analyzing the results of the models, we created, we showed that the ANN with a single hidden layer gave more accurate results in predicting fuel consumption. This model has an error rate of 2.3% and estimates fuel consumption with an average error of 7.34 L. The created DT is a model that can help in many aspects of planning, such as trip scheduling and preventive maintenance. Using this model, the ideal driving speed between stations can be calculated and train services can be scheduled to minimize fuel consumption. The remaining useful life can be calculated by studying the fuel consumption behavior, and fault detection can be performed in accordance with the fuel consumption pattern.