Validation and Prediction of Ultra-Low NOx Burner Performance using Computational Fluid Dynamics

Paper from the AFRC 2015 conference titled Validation and Prediction of Ultra-Low NOx Burner Performance using Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) modelling and simulations were performed to evaluate and; predict the performance of an ultra-low NOx burner design. CFD model validation using; experimental results was performed to evaluate model settings and parameters. The validation; showed a good trend agreement with the experimental results and provided confidence in the; model that it can be used to predict the burner performance.; Geometric changes were made to the model and simulations were performed to predict the; performance and emissions of the burner. The CFD model prediction was performed and testing; was conducted for these data points. Results indicated that the prediction trend was in good; agreement with the follow-on experimental testing verifying the robustness of the CFD model to; predict outcomes in an accurate and meaningful way. These results of the model development for; the ultra-low NOx burner will be presented.

Validation and Prediction of Ultra-Low NOx Burner Performance using Computational Fluid Dynamics > Sandeep Alavandi (GTI), Derek Wissmiller (GTI), Anchal Jatale (Ansys, Inc.) and Muhammad Sami (Ansys, Inc.) > July 15, 2015 Outline > Burner Development Project Overview > Burner Geometry and Validation Parameters > Modelling Approach > Grid Independent Study > Validation and Prediction Results ─ Turbulence - Chemistry Interaction ─ Methane Reaction Mechanisms ─ NOx Validation ─ Prediction with Design Changes > Conclusions © GTI AFRC 2015 2 GTI Overview > Not-for-profit (501c3) RD&D organization with 70 year history > Facilities ─ 18 acre campus near Chicago ─ 200,000 ft2 with 28 specialized labs ─ Other sites in California, D.C., Texas, Pittsburgh, Massachusetts > Staff ─ Approximately 270 ─ 180 engineers, scientists covering a variety of energy fields Industrial and Power Lab Residential & Commercial Lab © GTI AFRC 2015 3 Burner Development Project > GTI is engaged in activities to develop an ultralow NOx burner for boiler applications > The test burner is rated at a capacity of 4 MMBtu/hr. > Testing is being performed on GTI's boiler simulator test facility ─ Exhaust emissions ─ In-situ flame probe measurements ─ Hardware temperatures ─ Flame luminosity imaging © GTI AFRC 2015 4 Model Development and Validation > A Computational Fluid Dynamics (CFD) model (3D) of the burner was developed using Ansys (Fluent) to ─ 1) provide insight into flame characteristics, 2) guide experimental testing activities > Efforts were made to achieve validation of the model with experimental results ─ 1) hardware temperatures, 2) CO concentration, 3) flame structure, 4) NOx emissions > All boundary conditions and domain properties were defined as per real-world conditions > Studies were performed to evaluate model settings ─ 1) grid independence, 2) turbulence-chemistry interaction, 3) inflation layer refinement, 4) shell conduction, 5) reaction mechanism, 6) NOx model settings > Model development efforts were initiated with premixed fuel/air, rather than modeling the actual fuel/air mixing geometry of the test burner ─ the actual fuel/air mixing was modeled subsequently © GTI AFRC 2015 5 Model Development and Validation (contd)… AIR FUEL Mixing Zone Actual Mixing  - Air and fuel  introduced separately as in the rig. Inlet Section Burner/ Combustor EXHAUST Premixed  - Air + Fuel introduced at the same  boundary at the inlet to the combustor to reduce  cell count and for initial validation. Ansys Model Settings > Turbulence - Realizable k-e, Scalable wall function > Species Transport - Eddy Dissipation versus Eddy Dissipation-Finite Rate > Reaction mechanism - 2-step versus 4-step > Radiation - Discrete Ordinates Geometry/Condition Changes > Premixed - Air + Fuel defined at same boundary > Actual Mixing - Air and Fuel defined as in actual system © GTI AFRC 2015 6 Validation: Grid Independence Normalized Burner Hardware Temperature > Burner hardware temperature was found to have the greatest sensitivity to model settings and mesh and was used for validation 1.2 1.1 1 Experimental 0.9 Mesh 1 - 815k Mesh 2 - 1.48MM 0.8 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 1 > A 81% increase in the number of cells provided only a minor change in results > Inflation boundary layers were modeled near the wall of burner hardware for accuracy © GTI AFRC 2015 7 Validation: Turbulence-Chemistry Interaction > Eddy-Dissipation (ED) vs. Eddy Dissipation-Finite Rate (ED-FR) was analyzed 2 Experimental 1.75 Normalized CO Concentration Normalized Burner Hardware Temperature 1.2 1.1 1 Experimental 0.9 ED ED‐FR 0.2 0.4 0.6 Normalized Axial Distance 0.8 ED‐FR 1.25 1 0.75 0.5 0.25 0 0.8 0 ED 1.5 1 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 1 > Temperatures: ED shows better agreement than ED-FR > CO Concentration: ED shows very poor agreement, whereas ED-FR shows very good spatial agreement and proximal agreement in magnitude > Flame Shape: ED showed very poor agreement with visual flame observations, whereas ED-FR showed good proximal agreement > ED-FR appears to more accurately characterize flame behavior © GTI AFRC 2015 8 Validation: Shell Conduction > An error in the Ansys (Fluent) software for shell conduction model was detected Normalized Burner Hardware Temperature 1.2 Ansys R15 1.1 1 0.9 Experimental With Shell Conduction Without Shell Conduction 0.8 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 1 > With the shell conduction model deactivated, the model results agree very well with experimental results © GTI AFRC 2015 9 Validation: Reaction Mechanism 2 1.2 Experimental Experimental 1.75 Methane-Air - 2-step - Mechanism Normalized CO Concentration Normalized Burner Hardware Temperature > Built-in 2-step mechanism vs. 4-step mechanism Methane-Air - 4-step - Mechanism 1.1 1 0.9 Methane-Air - 2-step - Mechanism Methane-Air - 4-step - Mechanism 1.5 1.25 1 0.75 0.5 0.25 0.8 0 0.2 0.4 0.6 0.8 Normalized Axial Distance 1 0 0 0.5 1 Normalized Axial Distance 1.5 > Temperatures: the 2-step and 4-step mechanisms provided good agreement with each other and experimental results > CO concentrations: the 4-step showed very poor agreement with experimental results whereas the 2-step showed much better agreement > The 2-step mechanism appears to give more accurate results, and is more computationally efficient © GTI AFRC 2015 10 Validation: NOx Model Settings > Ansys (Fluent) default NOx model did not provide good agreement > Reactor Network (RN) model provides higher fidelity in chemical reactions leading to NOx formation, and improved accuracy Reactor Network On Patch GRI Mech. 3.0 on the domain  Choose # of Reactors  > The RN is a post processing tool that resolves the CFD model domain into a number of perfectly stirred reactors ─ The user specifies the number of reactors (far fewer than the number of cells) ─ The software automatically breaks the domain into the reactors as based on information from the solution ─ The full GRI Mechanism can be applied to each of the reactors > This provides higher accuracy with greatly reduced computational time in comparison to using detailed chemistry on the full mesh © GTI AFRC 2015 11 Validation: NOx Model Settings > NOx emissions were calculated using the RN model for varying numbers of reactors 1.50 Normalized NOx Emissions Large  variation  in NOx Experimental Normalized Value = 1.0 (Premixed Model) 1.25 Results at 500, 1000, and  1500 reactors were used to  provide an average and  uncertainty 1.00 0.75 Release_16 Release_15 0.50 0.25 0.00 0 500 1000 1500 2000 2500 Number of Reactors 3000 3500 uncertainty =  standard deviation / average > RN provides good agreement with experimental results > Below 500 reactors, there was very high uncertainty in the results > Even with high number of reactors, the NOx results never fully converged ─ There appears to be some intrinsic uncertainty with the Reactor Network model ─ For functional purposes, this uncertainty was characterized by using results for 500, 1000, and 1500 reactors © GTI AFRC 2015 12 Effect of Excess Air - Premixed > The model was used to predict the effect of varying excess air in comparison to experimental results ─ No changes were made to mesh or model settings other than the fuel/air boundary condition SOLID LINE ‐ EXPERIMENTAL  DOTTED LINE ‐ PREMIXED  1.05 2.00 Normalized CO Concentration Normalized Burner Hardware Temperature 1.1 Increasing Excess Air 1 0.95 1.75 1.50 Variable Excess Air 1.25 1.00 0.75 0.50 0.25 0.00 0.9 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 1 > Temperatures: the model under-predicts the absolute temperature values > CO Concentrations: the model under-predicts CO values, but shows good proximal agreement with spatial location of peak CO concentrations © GTI AFRC 2015 13 Effect of Excess Air - Premixed > The model was used to predict the effect of varying excess air in comparison to experimental results ─ No changes were made to mesh or model settings other than the fuel/air boundary condition 2.5 Normalized NOx Emissions Experimental Premixed 2 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 Normalized Excess Air Factor > The model shows good agreement both in relative magnitude and in absolute values in comparison to experimental NOx measurements © GTI AFRC 2015 14 Effect of Excess Air - Actual Mixing > The model geometry was modified upstream of the inlet zone to match the boiler simulator geometry for fuel and air entry and mixing ─ No changes were made to the mesh or model settings downstream of the nozzles 2.00 Variable Excess Air SOLID LINE ‐ EXPERIMENTAL          DOTTED LINE ‐ ACTUAL MIXING  Normalized CO Concentration Normalized Burner Hardware Temperature 1.1 1.05 Increasing Excess  Air 1 0.95 1.75 1.50 Experimental 1.25 1.00 0.75 0.50 0.25 0.00 0.9 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 0 0.2 0.4 0.6 Normalized Axial Distance 0.8 1 > Temperatures: the model under-predicts the absolute temperature values; > CO Concentrations: both the location and magnitude of the CO values are in good proximal agreement with experimental results; the values are somewhat overestimated, but in better agreement than with the premixed model © GTI AFRC 2015 15 Effect of Excess Air - Actual Mixing > The model was modified upstream of the nozzle outlet to provide actual modelling of the fuel and air mixing geometry ─ No changes were made to the mesh or model settings downstream of the nozzles Normalized NOx Emissions 2.5 Experimental 2 Actual Mixing 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 Normalized Excess Air Factor > The model shows fair agreement both in relative magnitude and in absolute values relative to experimental NOx measurements © GTI AFRC 2015 16 NOx Prediction for Design Changes > NOx emissions were evaluated for physical changes in the burner geometry 2 Exp- Baseline Design Normalized NOx Emissions 1.75 CFD - Baseline Design Exp - Design Change 1.5 CFD-Design Change 1.25 1 0.75 0.5 0.25 0 0.5 1 1.5 2 2.5 Normalized Excess Air Factor > NOx emissions showed good trend agreement to the design change © GTI AFRC 2015 17 Conclusions > Meshing, boundary layer inflation and turbulence-chemistry model evaluation were critical to providing validated and meaningful results > Effect of excess air for the two geometry configurations provided good trend agreement > Reactor network model provided good agreement with NOx experimental results > Qualitative trends matched well with the experimental results, however, quantitatively the results were not accurate. ─ This could be attributed to not having a full reaction mechanism ─ Lack of a good NOx model that can provide fast and accurate results > Current results demonstrate that the model can be used to estimate the effect of design changes on burner performance © GTI AFRC 2015 18