Forecasting of Engine Performance for Gasoline-Ethanol Blends using Machine Learning

The incorporation of alternative fuels in the automotive domain has brought a new paradigm to tackle the environmental and energy crises. Therefore, it is of interest to test and forecast engine performance with blended fuels. This paper presents an experimental study on gasoline-ethanol blends to test and forecast engine behavior due to changes in the fuel. This study employed a machine learning (ML) technique called TOPSIS to forecast the performance of a slightly higher blend fuelled engine based on experimental data obtained from the same engine running on 0% ethanol blend (E0) and E10 fuels under full load conditions. The engine performance predictions of this ML model were validated for 15% ethanol blend (E15) and further used to predict the engine performance of 20% ethanol blend fuel. The prediction R2 score for the ML model was found to be greater than 0.95 and the MAPE range was 1% to 5% for all observed engine performance attributes. Thus, this paper presents the potential of TOPSIS methodology-based ML predictions on blended fuel engine performance to shorten the testing efforts of blended fuel engines. This methodology may help to faster incorporate higher blended fuels in the automotive sector.