Ignition and early flame development in stratified-charge mixtures

Saumyadip Mukhopadhyay, Purdue University

Abstract

Stratified-charge combustion is important in internal combustion engines and other combustion applications when there is fuel injection and insufficient time for the fuel and air to completely premix before start of combustion. In this work, numerical studies are carried out to understand the influence of stratification on laminar and turbulent compression-ignited stratified-charge mixtures. It is shown that in laminar compression-ignited fuel/air mixtures, the ignition behavior of stratified mixtures is controlled by two phenomena: diffusive losses of species and heat from the ignition location, and ignition front propagation from the ignition initiation location to the location where the flame finally stabilizes. When diffusive losses are controlling, increasing gradients retards ignition. When ignition front propagation is controlling, increasing gradients accelerates ignition. The controlling phenomenon is dependent on the fuel, initial temperature, and temperature gradients in the mixture. The influence of turbulence on autoignition characteristics of n-heptane/air mixtures is studied using two-dimensional (2-D) direct numerical simulations (DNS). Turbulence, due to its stochastic nature, is observed to modify the initially specified gradients, i.e. the initial scalar dissipation rate χ, to generate a range of values of χ around the initial value. Ignition is observed in the regions with low χ. The presence of these low χ regions accelerates the onset of ignition relative to the laminar mixing layer. An increase in turbulence intensity and/or a decrease in turbulence integral length scale increases the χ scatter in mixture fraction (Z) space, i.e the χ-Z spread in the domain and accelerates ignition. While ignition delay times are altered, the influence of gradients on ignition, identified in the laminar mixing layers, is not altered in the presence of turbulence. Using the DNS database, the accuracy of a promising sub-grid scale combustion model, the unsteady flamelet progress variable (UFPV) model employed in Reynoldsaveraged simulations (RAS) and large eddy simulation (LES) applications is indirectly assessed. Differences are observed between the ignition trends obtained from DNS, and those predicted by the UFPV model. The differences are greater when the compositional gradients are smaller suggesting that the flamelet assumption is questionable in such situations. In fact, when the gradients are small, heat release significantly alters the χ-Z dependence which is assumed á priori in flamelet models.

Degree

Ph.D.

Advisors

Abraham, Purdue University.

Subject Area

Chemical engineering|Mechanical engineering

COinS