Evaluation of ethanol-gasoline blends in SI engines using experimental and ANN techniques

Abstract Fuel combustion has become a major global concern, with much research focusing on the various emissions resulting from different types of fuels. Due to the harmful pollutant emissions from fossil fuels, the world has turned to renewable and alternative fuels to limit toxic emissions and greenhouse effects. Ethanol is a biofuel that, when used in spark ignition engines with gasoline can improve the octane number, combustion efficiency, and produce less emissions. The current research studies the effect of different ethanol blends E0, E5, E10, and E15 with gasoline 92 on engine performance parameters and emissions of a GX35 four-stroke engine at different engine speeds. The results along the speed range reveal that increasing ethanol amount leads to an average increase of 2.7%, 1%, and 1.1% in brake power (BP), brake thermal efficiency (BTE), and CO 2 emissions, respectively. Meanwhile, it causes an average decrease of 28 °C, 3%, 15 ppm, and 0.18% in exhaust gas temperature (EGT), brake-specific fuel consumption (BSFC), HC, and CO emissions respectively. Moreover, the current study develops an Artificial Neural Networks (ANN) model for predicting the performance and emissions of spark ignition (SI) engines. Python programming language is used for ANN coding to train and validate the ANN model with E15. Regression plots were generated to visualize the correlation between the target and predicted data, indicating outstanding performance. The results confirmed the model’s reliability for BP, EGT, CO, CO 2 , and HC parameters with R 2 values more than 0.99 and with acceptable performance for BSFC and BTE with R 2 of 0.9339, and 0.9708, respectively. To ensure that the is no overfitting during the ANN study, we used different statistical methods, such as validation set, cross-validation, and learning curves.