Designing a Deep Learning Based SI Engine Model With DOE Response

Abstract

This research introduces an innovative approach to modeling Spark Ignition (SI) engines by combining deep learning techniques with the Design of Experiments (DOE) framework. Conventional SI engine models often depend on complex physical equations or empirical correlations, which are computationally demanding and may fall short in fully capturing engine dynamics across diverse operating conditions. To address these challenges, this study develops a data-driven SI engine model designed to accurately predict both performance and emission characteristics. A structured DOE methodology will be applied to generate a comprehensive dataset covering a broad spectrum of input parameters, such as engine speed, load, spark timing, and air–fuel ratio. This carefully designed dataset will serve as the foundation for training and validating advanced deep learning models, including Artificial Neural Networks (ANNs), Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs), selected based on the nature of the input features.