Modeling of Direct Gas Injection in Internal Combustion Engines A. Y. Deshmukh 1, C. Giefer 1, M. Khosravi 2, H. Pitsch 1 1 Institute for Combustion Technology, RWTH Aachen University 2 Ford Research and Advanced Engineering, Ford Werke GmbH
Acknowledgements This work was performed within the Ford RWTH-Aachen Research Alliance Project FA-99 funded by the Ford Motor Company The experimental data used for comparison was provided by Delphi Licenses of CONVERGE software and support were provided by Convergent Science Inc. License of compressible LES solver CHRIS was provided by Cascade Technologies Inc. Simulations were performed with computing resources granted by RWTH Aachen University under project thes382 and JARA117
Contents Motivation and Challenges Simulation Setup Nozzle Flow Model Coupling with CONVERGE Results Summary & Outlook
Motivation: Spark-Ignited Direct Injection Engine with CNG as Fuel Emissions - CNG composition [1] : 7-95% CH 4 - High H/C ratio leads to lower CO 2 emissions (upto 25% reduction [3] ) - Lower particulate emissions [2] Potentially higher thermal efficiency - High resistance to knock => High compression ratio Direct Injection vs Port Injection - Higher volumetric efficiency Exhaust Intake Ø Design and optimization of DI CNG engines using numerical simulations (URANS) However, this is challenging Region ID Z Y X 2 1 Cylinder [1] McTaggart-Cowan G. P., Rogak S. N., Munshi S. R., Hill P. G., Bushe W. K., Fuel 21; 89:752 9. [2] INGAS- Integrated GAS Powertrain, Project Final Report. [3] Preuhs, J-F., Hoffmann, G. and Kirwan, J., International Conference: SIA Powertrain, Versailles, France, May 27-28 215.
Challenges in DI CNG Simulations Direct gas injector: Poppet-type valves - Small and complex gas passages (order of micrometer) Small passages Small mesh size (1-2 µm) required to resolve flow in the gaps Compressible flow in gas passages - High pressure ratios: Supersonic velocities (~1 m/s) Under/Over-expanded configuration leading to expansion waves or compression shocks Small time-step of order of 1 nanoseconds on coarse mesh - Typical duration of injection in an engine: 3-6 ms - Simulation time for a full engine cycle including injection: 3-4 weeks Ø First focus on injection process
Simulation Setup Axi-symmetric cylindrical domain - Diameter: 75 mm, Length 82.5 mm Injected fluid: Helium Fluid in cylinder: Air (77% N 2, 23% O 2 by mass) Boundary conditions: - Inlet total pressure: 15 bar - Inlet temperature: 298 K - Injector walls: Adiabatic Initial conditions: - Pressure: 1.1325 bar - Temperature: 298 K Duration of Injection:.8 ms P = 15 bar T = 298 K Inlet P = 1.1325 bar T = 298 K Helium Air
Mesh ~ 25. µm ~ 2 mm ~ 5. µm Base size: 2 mm Fixed Embedding as shown in Figure No adaptive mesh refinement (AMR) Z X ~ 1 mm
Nozzle Flow Model Scale separation Separate nozzle flow from the full simulation: Develop a fast and accurate model 3D simulation of the downstream region Ø Model the poppet-type valve as a duct with varying area of cross section Poppettype valve Duct with varying area of cross section Area of Cross Section Z
1D Code Development Solve one-dimensional (1D) system of inviscid Euler equations for the duct with varying area 1-D Euler Equations ρa t + ρua = s ρua + (ρu+ + P)A t s ρea ρe + P ua + t s A = A(s, t) = P A s = P A t Geometric centerline s Axis of the Needle Solved using MUSCL scheme (upto 2 nd order spatial accuracy) and forward Euler time integration scheme CFL:.3
1D Code Validation: Comparison with 3D LES* of Nozzle at Steady State Geometric centerline Normalized Area [-] 1.2 1..8.6.4.2 s Axis of the Needle Normalized Area...2.4.6.8 1. s[-] 16 14 12 1 8 6 4 2..2.4.6.8 1. s[-] Helium at nozzle pressure ratio of 15 Reasonable prediction of the nozzle internal flow using 1D code Velocity Magnitude [m/s] Pressure [bar] 16 14 12 1 8 6 4 2 3D LES 1D Velocity Pressure..2.4.6.8 1. s[-] Mach Number [-] Density [kg/m 3 ] 2.5 2. 1.5 1..5...2.4.6.8 1. s[-] 2.5 2. 1.5 1..5 Mach Number Density...2.4.6.8 1. s[-] * LES was performed using CHRIS solver developed by Cascade Technologies Inc.
Coupling with CONVERGE (v2.3.9) Coupling of 1D code to CONVERGE using mass, momentum, and energy sources CONVERGE Point sources USER_SOURCE_SETUP Coupling Function Read Parameters: P inj,t inj, γ, M, cfl, β vel START P cyl, T cyl dt USER_DRIVER_SPECIES Coupling Function v exit, T exit, mfr Solve Equations 1-D Code t = t + dt yes t < t_end no END
Simulation Cases No. Case Remark 1 3D valve with fixed needle lift Reference case 2 1D valve model with fixed needle lift 3 3D valve with needle lift profile Reference case 4 1D valve model with needle lift profile CONVERGE Setup: - URANS with RNG k-epsilon turbulence model - Second order numerical scheme with fully implicit time integration Experimental data using Schlieren imaging provided by Delphi - Injection into open ambient conditions - Injection pressure: 16 bar
Results: Fixed Needle View 3D Valve Geometry 1D Valve Model
Quantitative Parameters AJ VJ APL APL AJ MW MW Experiment Simulation Axial Penetration Length (APL) Maximum Width (MW) Area of Jet (AJ) Volume of Jet (VJ)
Quantitative Comparison Fixed Needle 7 6 5 4 3 2 1 1 2 3 4 5 6 7 Time [µs] AJ Axial Penetration [mm]8 2 15 1 5 1 2 3 4 5 6 7 Time [µs] Area of Jet [mm 2 ] 125 APL Maximum Width [mm] Volume of Jet [cc] 6 5 4 3 2 1 1 2 3 4 5 6 7 Time [µs] 5 VJ 4 3 2 1 Reasonable agreement between Full 3D nozzle and 1D nozzle flow model MW 1 2 3 4 5 6 7 Time [µs] EXP Full 3D Nozzle 1D Nozzle Flow Model
Needle Opening Effects using 1D Code Needle Lift [µm] Needle Lift [µm] Mach Number [-] 5 4 3 2 1 1 2 3 4 5 6 7 Time [µs] 5 4 3 2 1 Needle Lift Vs. Time..2.4.6.8 1. Normalized Area [-] Normalized Area [-] 1..8.6.4.2...2.4.6.8 1. s[-] Time [µs] s [-] s [-] Pressure [bar] 16 14 12 1 8 6 4 2..2.4.6.8 1. s [-]
Results: Needle Lift Profile View 3D Valve Geometry 1D Valve Model
Quantitative Comparison Moving Needle 7 6 5 4 3 2 1 1 2 3 4 5 6 7 Time [µs] AJ Axial Penetration [mm]8 2 15 1 5 APL 1 2 3 4 5 6 7 Time [µs] Area of Jet [mm 2 ] 125 Maximum Width [mm] Volume of Jet [cc] 6 5 4 3 2 1 MW 1 2 3 4 5 6 7 Time [µs] 5 VJ 4 3 2 1 1 2 3 4 5 6 7 Time [µs] Ø This is still a work in progress EXP Full 3D Nozzle 1D Nozzle Flow Model Initial non-linear behavior captured well with 3D valve geometry, but later deviates from experiments
Reduction in Computational Cost Time-step relative to minimum time-step in 3D Relative Computational Cost [-] 6 5 4 3 2 1 3D-FN 1D-FN 3D-MN 1D-MN Cases Computational cost reduced by a factor of 5-6
Summary & Outlook Ø A one dimensional nozzle flow model was developed for a poppet-type gas injector Ø The 1D code is coupled to CONVERGE and the results with fixed needle lift agree with full 3D case and reasonably with experiments Ø Computational time reduced by a factor of 5-6 by increasing the simulation time-step using the nozzle flow model Ø Next Ø Further investigation of effect of needle opening on gas jet evolution Ø Application to full engine simulations Velocity Magnitude [m/s] 16 14 12 1 8 6 4 2..2.4.6.8 1. s[-] 7 6 5 3D LES 1D 4 3 2 1 1 2 3 4 5 6 7 Time [µs] Axial Penetration [mm]8 APL EXP Full 3D Nozzle 1D Nozzle Flow Model
Thank you for your attention Abhishek Y. Deshmukh PhD Student Institute for Combustion Technology RWTH Aachen University Templergraben 64 5256 Aachen Germany Tel: +49-241-894622 Fax: +49-241-892923 EMail: a.deshmukh@itv.rwth-aachen.de WWW: http://www.itv.rwth-aachen.de