An Investigation on Turbulent Burning Velocity of Premixed Combustion and Quasidimensional Simulation of a Gasoline Direct Injection Engine

Abstract

This thesis is composed of two parts. Part I deals with validation of the new asymptotic expressions of turbulent burning velocity, S_T, of premixed combustion at the leading edge and Part II discusses quasidimensional analysis of combustion, emissions and knocking in a homogeneous gasoline direct injection (GDI) engine. S_T has been the critical issue of numerous experimental and theoretical researches for turbulent premixed combustion due to its potential effect to determine the mean reaction rate. Despite these continuing efforts, predictive correlations of S_T do not reproduce proper trends for wide variation of the conditions of combustion with measured S_T showing substantial sensitivity to flame geometry and experimental methods. This thesis presents validation of new analytical expressions for the turbulent burning velocity, S_T, based on asymptotic behavior at the leading edge (LE) in turbulent premixed combustion. Reaction and density variation are assumed to be negligible at the LE to avoid the cold boundary difficulty in the statistically steady state. Good agreement is shown for the slopes, dS_T⁄du', with respect to L_c⁄δ_f at low turbulence, with both normalized by those of the reference cases. δ_f is the inverse of the maximum gradient of reaction progress variable through an unstretched laminar flame and L_c is the characteristic length scale given as burner diameter or measured integral length scale. Comparison is made for thirty-five datasets involving different fuels, equivalence ratios, H2 fractions in fuel, pressures and integral length scales from eight references (Aldredge et al., 1998; Lawes et al., 2012; Kido et al., 2002; Wang et al., 2013; Kobayashi et al., 1998; Chiu et al., 2012; Venkateswaran et al., 2013; Fairweather et al., 2009). The turbulent burning velocity is shown to increase as the flamelet thickness, δ_f, decreases at a high pressure, for an equivalence ratio slightly rich or close to stoichiometric and for mixture of a high H2 fraction. Two constants involved are C to scale turbulent diffusivity as a product of turbulent intensity and characteristic length scale and C_s to relate δ_f with the mean effective L_m. L_m (=D_mu⁄(S_Lu^0 )) is the scale of exponential decay at the LE of an unstretched laminar flame. The combined constant, K C⁄C_s , is adjusted to match measured turbulent burning velocities at low turbulence in each of the eight different experimental setups. All measured S_T⁄(S_Lu^0 ) values follow the line, K D_tu⁄D_mu +1, at low turbulent intensities and show bending below the line due to positive mean curvature and broadened flamelet thickness at high turbulent intensities. Further work is required to determine the constants, C_s and K, and the factor, (L_m⁄(L_m^* )-L_m 〈∇∙n〉_f ), that is responsible for bending in different conditions of laminar flamelet and incoming turbulence. Turbulent premixed combustion is used in quite a few industrial combustion systems. A representative example may be a homogeneous charge spark ignition (SI) engine. In this study a quasidimensional model is developed with the surrogate mechanism of iso-octane and n-heptane to predict knock and emissions of a homogeneous GDI engine. With early injection a homogeneous GDI engine goes through spark ignition and turbulent premixed flame propagation as in port fuel injection (PFI) engines. It is composed of unburned and burned zone with the latter divided into multiple zones of equal mass to resolve temperature stratification. Combustion is based on turbulent entrainment and burning in a spherically propagating flame with the entrainment rate interpolated between laminar and turbulent flame speed. Validation is performed against measured pressure traces, NOx and CO emissions at different load and rpm conditions. Comparison is made between predictions by the empirical knock model and the chemistry model in this work. There is good agreement for pressure, NOx, CO and knock for the test engine. Promising results are obtained through parametric study with respect to octane number and engine load by the chemistry knock model.