WELCOME VÄLKOMMEN BIENVENIDO DOBRO DOSLI WELKOM SELAMAT DATANG CROESO NAMASTE WILLKOMMEN BENVIDO VELKOMMEN KARIBU BENVENUTO BIENVENUE SOO DHAWOW 2016 Convergent Science. All Rights Reserved
Combustion Modeling Development in CONVERGE Eric Pomraning, PhD 2018 CONVERGE User Conference Bologna, Italy 2016 Convergent Science. All Rights Reserved
Overview Combustion model development (3D) - Species based ECFM - SAGE-RANS Triple-Delta PDF - LES-SAGE Non-Premixed - LES-TFM SAGE Premixed Chemistry utilizes (0D/1D) - Improved mechanism reduction - Rapid Compression Machine model - HCCI engine model - RON/MON calculator - Laminar flame speed calculations
Species Based ECFM/ECFM3Z Model Original ECFM approach uses a passive based progress variable - Incoherence in transport/diffusion of species and passives - Need to use first order upwind - Progress variable not zero in unburned gases before combustion Species based progress variable - Solves the issues above - Allows the detailed chemistry/emission coupling - Allows multi-component fuel 2 nd order with SB-ECFM3Z DC3 Engine Sector 1 st order with passive-ecfm3z 5 c YF 1, Y FT
SAGE-RANS Triple-Delta PDF The RANS species transport equation is given by If we solve for the chemical reactions directly we can evaluate As the chemical source term is non-linear in temperature and species mass fractions, there is an error when assuming that the RANS averaging commutes. The RANS commutation error does not reduce to zero with reduced cell size. The commutation error is unimportant for small fluctuations In a simulation, if a RANS turbulence model does not recover the ensemble average, it is not clear if the commutation error is important or even how it should be evaluated! Frequently, a RANS turbulence model does not recover the ensemble average!
Triple-Delta PDF Model To evaluate the commutation error, we need to consider a distribution in species (non-premixed) or temperature (premixed) Triple Delta PDF is developed to consider the fluctuations in turbulence reacting flow. Transport mean species mass fraction : small variance unimodal distribution Probability of three states are based on the mean equivalence ratio and its variance: large variance bimodal distribution
Triple-Delta PDF Model: Sandia Flame D Flame D: Non-premixed methane-air mixture Jet flame with burning pilot-stabilizer Turbulence model: Standard k-epsilon model Combustion model: Single-Delta PDF (finite-rate chemistry) and Triple- Delta PDF Mechanism: 30-species skeletal mechanism based on GRI3.0 The Flame D simulated flow field with k-epsilon is steady and recovers the experimentally averaged flow field Sandia National Laboratories, TNF Workshop website
Triple-Delta PDF Model: Sandia Flame D x/d=15.0 x/d=30.0
LES SAGE Non-Premixed CSI believes LES is better for predictive combustion simulations There are outstanding questions for non-premixed combustion with LES - Grid resolution required - The importance of the commutation error To investigate these issues, an LES combustion study of Flame D was conducted LES Simulation of Diesel Engine (courtesy of Aramco)
LES SAGE Non-Premixed (Flame D) LES Turbulence model: Dynamic Structure Model Combustion model: Single-Delta PDF (finite-rate chemistry) Inflow: Mean inflow velocity are directly from experimental measurements, synthetic isotropic turbulence perturbations are added to the inflow velocity Finest resolution (Case a) is 0.25 mm. Resolution is clearly not sufficient for a DNS simulation! *Sandia National Laboratories, TNF Workshop website
Instantaneous Temperature Distribution (LES) E (2.0 mm) D (1.0 mm) C (0.5 mm) B (0.375 mm) A (0.25 mm) Large Cell Size Small
Mean Temperature Distribution (LES) E (2.0 mm) D (1.0 mm) C (0.5 mm) B (0.375 mm) A (0.25 mm) Large Cell Size Small
Case A: 0.25mm Velocity Temperature Mixture fraction CO mass fraction Near field temperature
Case A: 0.25mm, Centerline profile
LES Engine Combustion Simulation Compression ignition simulation with combustion Simulation with approximately 100 Million cells! Aramco Designed Piston With Compression Ratio 20.5
LES Thickened Flame Model (TFM) SAGE Premixed The thickness of a premixed flame is about 0.01-1mm. Generally, it is computationally too expensive to resolve flame front (5 to 10 grid points across the flame front). Thickening the flame is an effective way to resolve the turbulence premixed flame in LES simulation. O. Colin et al., Phys. Fluids A 12 (7) (2000) 1843 1863 Turbulent Premixed V-Flame *Yuki Minamoto (Sandia National Laboratories) and Mamoru Tanahashi (Tokyo Institute of Technology)
1D Laminar Flame with TFM Laminar flame speed didn't change too much with thickening factor from 1 to 20.
2D Flame Growth with Decaying Isotropic Turbulence with TFM flame front resolved Well resolved results DNS results With thickened flame model, simulation of turbulent premixed flame with relative coarse grid can give reasonable results.
2D Flame Growth with Decaying Isotropic Turbulence with TFM Temperature Vorticity
Improved Mechanism Reduction Originally chemical mechanism is reduced by looking at the ignition delay. - Error can be introduced to laminar flame speed. Improved method allows user to consider laminar flame speed as targets, beyond the ignition delays. 0D only 0D& 1D
Temperature, K Zero-D: Rapid Compression Machine Closed homogeneous reactor. Temperature profiles of 0D RCM simulations Users can specify volume evolution profiles and heat transfer at the boundary. Useful tool for reactor simulations of HCCI, RCM etc. Time, s http://ibnox.com/read/7eapnzt/engines-advancements-advantages-and-disadvantages
Zero-D: HCCI Engine Model Internal Combustion Homogenous Charge Compression Ignition Model - Study engine knock, auto-ignition, and HCCI engines - Engine parameters are user specified (compression ratio, crank radius, RPM, displaced volume, etc ) and used in Heywood (1988)* volume equations. - Wall Heat Transfer Models (Woschni, Hohenberg, Annand, and user modified) provide further accuracy of engine simulation. *J. B. Heywood, Internal Combustion Engines Fundamentals, McGraw-Hill Science/Engineering/Math, New York, 1988.
Zero-D: Engine RON/MON Based on HCCI Engine Model Calculates RON or MON of a fuel composition - Finds critical compression ratio (CCR) where autoignition occurs - Correlation based on the CCR of PRF fuel composition (i.e., PRF 80 equates to RON/MON of 80) using the LLNL Gasoline Mechanism - Recommend for gasoline fuels only - Engine parameters are based on ASTM D2699 (RON) and ASTM D2700 methods (MON)
One-D Premixed Flame Sparse Solver A new option to the one-d premixed flame Newton solver Up to an order of magnitude faster for larger mechanisms Different treatment of species diffusion yields a sparser matrix - Nitrogen is used to account for inconsistencies introduced by Fick s law instead of correction velocity
Summary ECFM species based has been added to CONVERGE in collaboration with IFPEN SAGE RANS Triple PDF model has been added to account for commutation error - Works well for steady solutions that give average result (e.g., Flame D) - Does not work well for solutions that do not give average result (e.g., engine simulations) We are actively researching LES combustion for non-premixed and premixed combustion - Implementing TFM model in collaboration with IFPEN CSI is continuing to improve our 0D and 1D chemistry modeling capabilities
Thank you pomraning@convergecfd.com THANK YOU! CONVERGECFD.COM If you are interested in joining C3, please visit fuelmech.org 2016 Convergent Science. All Rights Reserved