Application of Extreme Learning Machine (ELM) for Predicting Combustion Pressure-Related Parameters in a Dual Ignition Gasoline Engine

Abstract

The study focused on the development of Extreme Learning Machine (ELM) based Artificial Neural Network (ANN) model. The developed model then used for predicting combustion related cylinder pressure parameters of a spark ignition (SI) engine. A widely used back propagation (BP) based ANN model also developed for the prediction performance comparison. For training and testing the model, set of data has been collected by conducting the experiment on twin spark ignited SI engine. The experiment was carried out under different load, ethanol-gasoline blend, compression ratio and spark timing. The modelling results showed that ELM based ANN model gives minimum MSE and MAPE (%) compared to the BP based ANN model. It is also found that the ELM algorithm is faster as it takes only one epoch with added advantages of good generalization performance and compact network architecture.