Validation of engine combustion models against detailed in-cylinder optical diagnostics data for a heavy-duty compression-ignition engine

Abstract

Three different approaches for modelling in-cylinder compression-ignition engine processes for partially premixed combustion modes are compared with experimentally observed cylinder pressure and in-cylinder images of liquid- and vapour-fuel penetration, ignition, combustion, and soot formation in an optically accessible heavy-duty direct injection engine. A multi-dimensional computational fluid dynamics model for engine combustion, KIVA-3V, served as a common platform into which three different combustion submodels were integrated: (1) characteristic time combustion (KIVA-CTC); (2) representative interactive flamelet (KIVA-RIF); and (3) direct integration using detailed chemistry (KIVA-CHEMKIN). Three different engine operating strategies with significant premixing of fuel and air prior to ignition were investigated: Low-temperature combustion achieved by charge dilution, with fuel injection either (1) early, or (2) late in the engine cycle, and (3) long ignition delay, high-temperature combustion (i.e. no charge dilution) with fuel injection near top dead centre of the piston stroke. Comparison of simulated cylinder pressure and heat-release rates with the experimental results shows that all the combustion submodels predict the cylinder pressures and heat-release rates reasonably well, but predictions of in-cylinder phenomena were significantly different among the submodels. The KIVA-CHEMKIN submodel predictions agree best with experimental observations of the location of ignition sites and the spatial distribution of soot and OH. The KIVA-RIF model, which uses global quantities to account for turbulence-chemistry interactions, under-predicts the flame lift-off, while ignition sites and species distributions are broader than observed experimentally. The KIVA-CTC submodel greatly over-predicts the spatial extent and total amount of in-cylinder soot.