Prediction of Spark-Ignition Internal Combustion Engine Performance Using Artificial Neural Network

This study introduces an innovative approach utilizing Artificial Neural Network (ANN) modeling to predict different performance metrics in a gasoline engine. Key parameters such as power, torque, brake specific fuel consumption (bsfc), fuel consumption, manifold air pressure, intake air temperature, and engine coolant temperature are the focus of investigation. The study further conducts a comparative analysis of performance outcomes by incorporating both wide-band and narrow-band oxygen sensors, predicting gas emissions (HC, NOx, CO) alongside engine power. Experimental data for training and testing the ANN have been obtained through controlled experiments on a four-cylinder engine, encompassing varied throttle loading conditions. The ANN model has been trained by using a combination of standard back-propagation and Bayesian regularization learning algorithms. Another subset of experimental data has been reserved for performance verification by comparing the predicted data from the ANN model with actual results. The study highlights exceptional regression values of 1 for training and 0.99996 for testing, underscoring the reliability and precision of the ANN model in predicting engine performance. These outcomes emphasize the myriad advantages of employing ANN for accurate and dependable predictions of engine performance parameters, highlighting its potential to enhance understanding and optimization in this domain.