Graham M. Goldin

Pdf calculations of turbulent lifted flames of H2/N2 fuel issuing into a vitiated co-flow

This paper presents detailed calculations of the flow, mixing and composition fields of a simple jet of hydrogen–nitrogen mixture issuing into a vitiated co-flowing stream. The co-flow contains oxygen as well as combustion products and is sufficiently hot to provide an ignition source for a flame that stabilizes at some ten diameters downstream of the jet exit plane. This configuration forms a good model problem for studying lifted flames as well as issues of auto-ignition. The calculations employ a composition probability density function (PDF) approach coupled to the commercial CFD package FLUENT. The in situ adaptive tabulation method is adopted to account for detailed chemical kinetics. A simple k–ε model is used for turbulence along with a low Reynolds number model for the walls. Calculations are optimized to obtain a numerically accurate solution and are repeated for two different H2 mechanisms, each consisting of ten species. The flame is found to be largely controlled by chemical rather than mixing processes. The mechanisms used yield different lift-off heights and compositions that straddle the data. Ignition delays are found to be extremely sensitive to the chemical kinetic rates of some reactions in the mechanisms.

A numerical study of auto-ignition in turbulent lifted flames issuing into a vitiated co-flow

This paper presents a numerical study of auto-ignition in simple jets of a hydrogen–nitrogen mixture issuing into a vitiated co-flowing stream. The stabilization region of these flames is complex and, depending on the flow conditions, may undergo a transition from auto-ignition to premixed flame propagation. The objective of this paper is to develop numerical indicators for identifying such behavior, first in well-known simple test cases and then in the lifted turbulent flames. The calculations employ a composition probability density function (PDF) approach coupled to the commercial CFD code, FLUENT. The in-situ-adaptive tabulation (ISAT) method is used to implement detailed chemical kinetics. A simple k–ϵ turbulence model is used for turbulence along with a low Reynolds number model close to the solid walls of the fuel pipe. The first indicator is based on an analysis of the species transport with respect to the budget of convection, diffusion and chemical reaction terms. This is a powerful tool for investigating aspects of turbulent combustion that would otherwise be prohibitive or impossible to examine experimentally. Reaction balanced by convection with minimal axial diffusion is taken as an indicator of auto-ignition while a diffusive–reactive balance, preceded by a convective–diffusive balanced pre-heat zone, is representative of a premixed flame. The second indicator is the relative location of the onset of creation of certain radical species such as HO2 ahead of the flame zone. The buildup of HO2 prior to the creation of H, O and OH is taken as another indicator of autoignition. The paper first confirms the relevance of these indicators with respect to two simple test cases representing clear auto-ignition and premixed flame propagation. Three turbulent lifted flames are then investigated and the presence of auto-ignition is identified. These numerical tools are essential in providing valuable insights into the stabilization behaviour of these flames, and the demarcation between processes of auto-ignition and premixed flame propagation.

A cell agglomeration algorithm for accelerating detailed chemistry in CFD

A cell agglomeration algorithm is proposed to mitigate the computational cost of incorporating detailed chemical kinetics in multi-dimensional Computational Fluid Dynamics (CFD) simulations. Cells that are close in species and energy composition space are agglomerated before calling the reaction integrator, substantially reducing the number of chemistry integrations. The algorithm is generalized and applicable to any reacting flow configuration, and the accuracy is fully controllable. A dynamic hash table is used to efficiently bin cells into high dimensional hyper-cubes in composition space. The method is applied to four different CFD simulations and the speed-up and incurred error are assessed for a range of agglomeration tolerances and table dimensions. The proposed approach exhibits up to an order of magnitude speed-up with a relatively moderate decrease in accuracy.

Reduced description of reactive flows with tabulation of chemistry

The direct use of large chemical mechanisms in multi-dimensional Computational Fluid Dynamics (CFD) is computationally expensive due to the large number of chemical species and the wide range of chemical time scales involved. To meet this challenge, a reduced description of reactive flows in combination with chemistry tabulation is proposed to effectively reduce the computational cost. In the reduced description, the species are partitioned into represented species and unrepresented species; the reactive system is described in terms of a smaller number of represented species instead of the full set of chemical species in the mechanism; and the evolution equations are solved only for the represented species. When required, the unrepresented species are reconstructed assuming that they are in constrained chemical equilibrium. In situ adaptive tabulation (ISAT) is employed to speed the chemistry calculation through tabulating information of the reduced system. The proposed dimension-reduction / tabulation methodology determines and tabulates in situ the necessary information of the nr-dimensional reduced system based on the ns-species detailed mechanism. Compared to the full description with ISAT, the reduced descriptions achieve additional computational speed-up by solving fewer transport equations and faster ISAT retrieving. The approach is validated in both a methane/air premixed flame and a methane/air non-premixed flame. With the GRI 1.2 mechanism consisting of 31 species, the reduced descriptions (with 12 to 16 represented species) achieve a speed-up factor of up to three compared to the full description with ISAT, with a relatively moderate decrease in accuracy compared to the full description.

Coupling Complex Reformer Chemical Kinetics with Three-Dimensional Computational Fluid Dynamics

A new capability is developed that enables the modeling of certain logistics-fuel reformers. The system described in this paper considers a shell-and-tube configuration for which the catalytic reforming chemistry is confined within the tubes. The models are designed to accommodate detailed gas-phase and catalytic reaction kinetics, possibly including hundreds of species and thousands of reactions. The shell flow can be geometrically complex, but does not involve any complex chemistry. An iterative coupling algorithm is developed with which the geometrically complex flow is modeled with FLUENT and the chemically complex reforming is confined to straight tubes. The paper illustrates the model using propane partial oxidation and reforming as an example.

Modeling CO With Flamelet-Generated Manifolds: Part 2 — Application

Conventional Flamelet Generated Manifold (FGM) closure of the mean progress variable reaction rate assumes PDF shapes to account for turbulent fluctuations. The FGM parameters are commonly assumed to be statistically independent, and the marginal PDFs invariably require second moments, which are difficult to model accurately and have limited coefficients that can be adjusted to calibrate the simulation. A new model is presented which locates the flame brush with a turbulent flame speed model, and applies the FGM kinetic rate to model kinetically limited processes, such as CO quenching, behind the flame-front. The model is applied to 3D RANS simulations of an equivalence ratio sweep in the GE Entitlement Rig perfectly premixed combustor experiment. Calculating the mean FGM reaction progress source term with standard assumed shape PDFs leads to a narrow flame brush and equilibrium CO outlet emissions. By limiting the mean FGM reaction progress source term by the turbulent flame speed model, the flame brush is broadened and super-equilibrium CO is predicted at the outlet. Good agreement with measurement is obtained with default model coefficients. Since the majority of the mean reaction progress source term is limited by the turbulent flame speed reaction rate, it is demonstrated that the model is relatively insensitive to assumed shape PDFs for the FGM rate, as well as the parameter used to determine the turbulent flame leading edge.Copyright

Modeling CO With Flamelet-Generated Manifolds: Part 1 — Flamelet Configuration

In laminar flamelet modeling, a laminar flame in a simple 0D or 1D configuration is calculated a-priori and parameterized by a few scalars such as mixture fraction and reaction-progress or strain-rate. Transport equations, or algebraic expressions, for these parameters are then solved in 3D CFD simulations, avoiding computationally expensive in-situ chemical kinetic calculations. Typical configurations for laminar flamelets include, in 1D, opposed flow configurations with either non-premixed or premixed streams, freely propagating premixed flames, premixed flames impinging on a (heated) wall, and burner stabilized premixed flames. In 0D, plug-flow, perfectly-stirred (PSR) and partially-stirred reactors have been employed to build ‘flamelet-like’ ignition and flame-propagation tables. This paper compares 1D strained steady and unsteady non-premixed flamelets, 1D strained premixed flamelets, and 0D PSR simulations at a stochiometric and a lean equivalence ratio. At stochiometric mixtures, all three flamelet configurations show comparable manifolds (i.e. CO and OH mass fractions, and reaction-progress source term distributions). However, at lean equivalence ratios, the different configurations show substantially different manifolds. It is concluded that a unique flamelet configuration cannot be identified for a general partially-premixed model that ranges from the non-premixed to the perfectly premixed limit. Instead, to accurately model CO emissions, it may be necessary to include both premixed and non-premixed flamelets, with a flame-index (e.g. Yamashita et al., 1996, Proc. Combust. Inst., 26) to select the appropriate burning regime.Copyright

A Comparison of RANS and LES of an Industrial Lean Premixed Burner

LES and RANS simulations of a Siemens scaled combustor are compared against comprehensive experimental data. The steady RANS simulation modeled one quarter of the geometry with 8M polyhedral cells using the SST-k-ω model. Unsteady LES simulations were performed on the quarter geometry (90°, 8M cells) as well as the full geometry (360°, 32M cells) using the WALE sub-grid model and dynamic evaluation of model coefficients. Aside from the turbulence model, all other models are identical for the RANS and LES. Combustion was modeled with the Flamelet Generated Manifold (FGM) model, which represents the thermo-chemistry by mixture fraction and reaction progress. RANS simulations are performed using Zimont and Peters turbulent flame speed (TFS) expressions with default model constants, as well as the kinetic rate from the FGM. The flame speed stalls near the wall with the TFS models, predicting a flame brush that extends to the combustor outlet, which is inconsistent with measurements. The FGM kinetic source model shows improved flame position predictions. The LES predictions of mean and rms axial velocity, mixture fraction and temperature do not show improvement over the RANS. All three simulations over-predict the turbulent mixing in the inner recirculation zone, causing flatter profiles than measurements. This over-mixing is exacerbated in the 900 case. The experiments show evidence of heat loss and the adiabatic simulations presented here might be improved by including wall heat-loss and radiation effects.Copyright

LOWER DIMENSIONAL MODEL FOR MODELING THE HEAT TRANSFER AND DETAILED REACTIONS INSIDE LONG CHANNELS

In the current work, a methodology is developed for coupling the one dimensional (1D) solution inside the non-permeable channels with the 3D outer flow in shell and tube type of configurations. In the proposed lower dimensional model, called the channel model, the 1D channels have detailed reactions while the outer 3D flow can be reactive or non-reactive. The channel is discretized into 1D grid points and a parabolic solver is used to solve the species transport and energy equations inside the channel. The channel is assumed to have a plug flow but the flow and the mixture properties inside the channel can have the axial variations. Since, the channel walls are non-permeable; the two zones are coupled only through the heat transfer. The current approach is tested and validated for a series of problems with increasing complexities. First, the study is performed with non reactive flow inside the thin channels as well as in the outer flow. The study is then extended to have reactions inside the channels. The current predictions of the channel model are compared with 3D modeling of the channels. It has been shown the channel model predictions are in excellent match to the fully resolved model with much lesser computational cost