Data-driven modeling of energy-exergy in marine engines by supervised ANNs based on fuel type and injection angle classification

Abstract

The application of artificial neural networks with the involvement of a modified homogeneity factor to predict exergetic terms from combustive and/or mixing dynamics in a marine engine is considered in this study. This is a significant step since the mathematical formulation of exergy in combustion is complicated and even unconvincing due to the turbulent and highly nonlinear nature of the combustion process. The computational simulations are carried out on a marine CI (compression ignition) engine and the respective data per different fuel types that are used for thermodynamic exergetic computations as well as energetic simulations. A new parameter namely the modified homogeneity factor derived by an artificial neural network (ANN) is considered for the mixing dynamics, i.e. as an input parameter for the availability and irreversibility predictions. This parameter is based on the standard deviation from an ideal air-fuel mixture formed within the combustion chamber of the marine engine. Furthermore, spray and injection quantities along with the combustion process and its heat transfer parameters are served to predict the exergetic terms for two study cases: (a) fuel type and (b) injection orientation. It is shown that using data analytics that consists of neural networks can provide an adequate approach in diesel engines for improving energy efficiency and reducing emissions.