Reduction in the nitrogen oxide and soot emissions in a diesel engine combustion system using an approximate optimization method

Abstract

The present study explores how to optimize a common-rail diesel engine to minimize nitrogen oxide and soot emissions with constraints on hydrocarbon and carbon monoxide emissions. A numerical diesel engine combustion and emission analysis is conducted using the zero-dimensional model. The simulation data are approximated, which accommodates the combustion characteristics of common-rail diesel engines and exhaust. The back-propagation neural network is used not only as a global function approximation tool for predicting non-linear characteristics but also for a global sensitivity analysis to examine causality between the input variables and the output responses. Given an engine operating condition, constrained bi-objective Pareto-optimal solutions are identified by the non-dominated sorting genetic algorithm II. Optimal designs to minimize the nitrogen oxide and soot emissions subject to the hydrocarbon and carbon monoxide emissions requirements are identified. By comparing the number of Pareto solutions with a baseline design, a reduction in both nitrogen oxide emissions and soot emissions is achieved and these back-propagation-neural-network-based approximate designs are verified using the actual zero-dimensional model data.