Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
40 Gábor Janiga<br />
For all computations presented in this paper, the complete set <strong>of</strong> chemical<br />
species <strong>and</strong> elementary reactions, with their Arrhenius coefficients Ai (preexponential<br />
factor), β (temperature exponent) <strong>and</strong> Ei (activation energy), is<br />
taken for methane/air flames from [50]. This detailed reaction scheme involves<br />
29 species <strong>and</strong> 141 elementary reactions. It thus involves 28 supplementary<br />
transport equations in the form <strong>of</strong> Eq. (2.8) plus Eq. (2.12), leading to a<br />
tremendous increase <strong>of</strong> the requested computing time compared to Case A.<br />
2.4.2 Numerical Solution<br />
The numerical simulation is performed using UGC + . This code has been<br />
optimized for the computation <strong>of</strong> steady laminar low-Mach number flows<br />
with chemical reactions [7, 68]. It is designed as an application <strong>of</strong> the multipurpose<br />
UG library [8]. UG is a modular numerical toolbox originally aimed<br />
at investigations <strong>of</strong> multigrid methods on various model problems described<br />
by sets <strong>of</strong> partial differential equations.<br />
UGC + is based on two main modules: a low-Mach Navier-Stokes solver<br />
<strong>and</strong> a thermo-reactive solver. A joint module has been developed to achieve<br />
the full coupling <strong>of</strong> the two sub-modules into one single Partial Differential<br />
Equations (PDE) system. The two solvers are in charge <strong>of</strong> their own diagonal<br />
block <strong>of</strong> the Jacobian matrix <strong>and</strong> there is an information interchange between<br />
them (mass fluxes, density <strong>and</strong> viscosity).<br />
The optimization procedure varies the mass flows <strong>of</strong> the fuel <strong>and</strong> air at the<br />
two inlets. The corresponding velocity values are used at the inlets for the<br />
Navier-Stokes equations. On both sides <strong>of</strong> the numerical domain, symmetry<br />
conditions are applied. The temperature <strong>of</strong> the fresh gas <strong>and</strong> walls at the<br />
inlets is 298 K. The total quantity <strong>of</strong> methane <strong>and</strong> air injected in the system<br />
(primary + secondary inlet) is kept constant, so that in principle the same<br />
energy is always available. At the outlet atmospheric pressure is imposed.<br />
The UGC + code attempts to find steady solutions through time-marching.<br />
Time discretization is <strong>of</strong> first-order implicit type. The value <strong>of</strong> the time-step<br />
can be adapted at each iteration, according to convergence or any user-defined<br />
condition. The unsteady equations are solved by fixed-point or approximate-<br />
Newton iterations <strong>and</strong> the user can freely specify how <strong>of</strong>ten the Jacobian<br />
matrix has to be assembled. The linearized equations are then solved by a<br />
Bi-CGSTAB [77] algorithm, preconditioned by multigrid V- or W-cycles with<br />
an ILU smoother [69]. A dynamic adaptive grid is used to increase resolution<br />
for such multi-scale problems (thin reaction zones, large geometries).<br />
The numerical simulation <strong>of</strong> a physical problem can be performed using<br />
various geometries <strong>and</strong>/or boundary conditions. For the present burner, the<br />
computational geometry is fixed for all computations, but the boundary conditions<br />
(composition <strong>of</strong> the mixture for the primary <strong>and</strong> the secondary inlets)<br />
are varied during the optimization procedure.