Predicting engine performance and exhaust emissions of a spark ignition engine fuelled with 2-butanol-gasoline blends using RSM and ANN models

• Main Author: Muhammad Yusri, Ismail
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://umpir.ump.edu.my/id/eprint/24976/1/Predicting%20engine%20performance%20and%20exhaust%20emissions%20of%20a%20spark%20ignition%20engine%20fuelled.pdf; http://umpir.ump.edu.my/id/eprint/24976/
• Institution: Universiti Malaysia Pahang

Experimental investigation in engine testing using alternative fuels always subjected to more engine operation, time-consuming and require expensive cost of materials. For these reasons, this study is aimed to predict the engine performance and exhaust emissions using 2-butanol-gasoline blended fuels with percentage volume ratios of 5:95 (GBu5), 10:90 (GBu10) and 15:85 (GBu15) of gasoline to 2-butanol, respectively, operated in a four-cylinder, four-stroke port fuel 4G93 Mitsubishi spark ignition engine at 30%, 50% and 70% of throttle position using artificial neural network and response surface methodology techniques. Based on the experimental investigation, at 30%, 50% and 70% of throttle position, 2-butanol–gasoline blended fuels indicated an improvement in engine brake power, brake torque and brake thermal efficiency with increasing 2-butanol content in the gasoline fuels. The engine performance indicated improvement in brake power, brake torque and brake thermal efficiency in the average of 2 to 15% and 0.2% to 1.5%,respectively, for all of the tested throttle position with respect to increasing the 2-butanol content in the gasoline fuel. For exhaust emissions, it was recorded that, a significant decreased of NOx, CO, CO2 and HC for GBu5, GBu10 and GBu15, by an average of 7.1%, 13.7%, and 19.8% than G100, respectively, over a speed range of 1000 to 4000 RPM. Other emission contents indicate lower CO and HC but higher CO2 from 2500 to 4000 RPM for the blended fuels. The engine speeds, 2-butanol blended fuels and engine throttle position and results from the engine performance and exhaust emissions characteristics was then used as the input and output for the for the artificial neural network and response surface methodology. Based on the RSM model, performance characteristics revealed that the increment of 2-butanol in the blended fuels lead to the increasing trends of brake power, brake torque and brake thermal efficiency. Nonetheless, a marginally higher brake specific fuel consumption was observed. Furthermore, the RSM model suggests that the presence of 2-butanol exhibits a decreasing trend of NOx, CO, and HC, however, a higher trend was observed for CO2 exhaust emissions, which are in accordance with the experimental results. Meanwhile, for ANN it was shown that the two hidden layer ANN model trained with the tansig-logsig activation function combination yields the best correlation coefficient, R at a value of 0.9995 against other activation function combinations evaluated. However, to attain a higher fidelity prediction model, all the configurations are further assessed by additional statistical error and correlation metrics, namely Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Theil U2, Nash-Sutcliffe Efficiency (NSE) and Kling–Gupta Efficiency (KGE). Following the evaluation, the best activation function combination for the brake power, BSFC, BTE, NOx, CO, and CO2 ANN predictive models is the tansig-logsig configuration. As for Brake torque and HC, the tansig combination provides a better prediction. It can be conclusively shown from the study that the developed ANN models have a higher predictive accuracy as compared to the RSM model.