Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Lawrence Livermore National LaboratoryLawrence Livermore National Laboratory
Modeling of high efficiency clean combustion engines Daniel Flowers Salvador Aceves Tom PiggottDaniel Flowers, Salvador Aceves, Tom Piggott,
Nick Killingsworth, Jonas Edman, Randy Hessel (University of Wisconsin)
DOE National Laboratory Advanced Combustion Engine R&D Merit Review and Peer EvaluationMerit Review and Peer Evaluation
Washington, DC This presentation does not contain any proprietary or confidential information
This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344
t
00
Goals/Objectives: collaborate with industry, academia and national labs in the development of analysis tools enabling clean, efficient engines
Our Artificial Neural Networkk-based KIVA-ANN provides
� We are demonstrating our � O A tifi i l N l N modeling tools for partially
computationally inexpensivestratified combustion analysis capability for HCCI/PCCI
T
P τ
……
Ignition Input Delay
Variables φ Time
� We are working with ORNL in analyzing HCCI-SI transitionexperiments
13%13% Internal EGRInternal EGR 58%58%1515
Conventional
2500 NOx (ppm)
2500NOx (ppm)
0
500
1000
1500
2000
NO
x, p
pm
0
COV IMEP (%)COV IM
500
1000
1500
2000
NO
x, p
pm
EP (%)1212 Discontinuity
between transition range and spark assisted HCCI
CO
V IM
EP, %
CO
V IM
EP, %
99
66
33
00
Exhaust Valve Closing, ATDCExhaust Valve Closing, ATDC34360 34360 320320 300300 280280 260260 240240 220220
Transition Spark Assisted HCCI HCCI
Requires Spark
Lawrence Livermore National Laboratory
Requires Spark No Spark
EGR
……
Input Hidden OutputLayer Layers Layer
� We are working with LANL on KIVA4MZ, a multi-zone version ofKIVA4
2
t t
FY2007 Reviewer’s Comments and Response
� Relevance is good Relevance is good, but the scope should include a broad view� but the scope should include a broad view of low temperature combustion, so it is not narrowly focused on HCCI. Most of our activity is now focused on partially stratified combustion and SI-HCCI transition.
� Multiple injections are a “fact of life” and will eventually have to be fully incorporated in modeling and lab work. We are currently dedicating most of our effort to direct-injected, partially stratified engine analysis
� It h b diffi lt d l i l IP ithIt has been difficult to develop commercial IP agreements with the LLNL industrial partnership office. Dan et al. are excellent to work with at the technical level, but at the “commercial level” we have struggled. Codes and models have been level we have struggled. Codes and models have been transferred to industry and they are available to industrial collaborators and researchers.
3 Lawrence Livermore National Laboratory
FCVT Barriers: we are contributing to an improved understanding of high efficiency engine regimesunderstanding of high efficiency engine regimes � Barrier: Increases in engine efficiency and decreases in
enggine emissions are beingg inhibited byy an inadequate abilityy to simulate in-cylinder combustion and emission formation processes • Our Contribution: We focus on improving understanding of
advanced combustion processes by developing detailed advanced combustion processes by developing detailed models of chemical kinetic-fluid mechanic interactions and their effect on engine combustion
4 Lawrence Livermore National Laboratory
Approach: Fundamental and practical understanding of advanced engine combustion regimes by judicious use of modeling and experiments
Fundamental UnderstandingUnderstanding
+ Design
Guidance
H3C O.O CH2 OH CH O.O CH O HO
H2C H2C Additive 1
H3C C CH3 O
H3C C CH3 H2C HC H3C
O Additive 2 O
H3C H2 HO CH
C H2 Additive 103 C
CH O. O. H3C O
O O O
OH CH2 O Additive 3 H2C H3C
H3C C CH3 CHC
CH3 O. H2C CH2 H3C H3C
C O O O
OH H2C HO Additive 122
H3C C CH3 Additive 4
H2C CH3 OH
O Additive 5 H3C
C O HO H2C Additive 204
H3C C CH3
H3C
Chemical kineticsHigh Resolution CFD
Advanced diagnostics Advanced diagnostics Engine Experiments
5 Lawrence Livermore National Laboratory
00
Technical accomplishment summary: we are working at demonstrating advanced analysis tools for challenging engine combustion problems � KIVA3V-MZ-MPI is running in direct � KIVA-ANN has demonstrated
injected mode and producing good performance in PCCI results analysis
T
P τ Ignition Input Delay
Variables φ Time ……
……
� Our chemical kinetic model matches ORNL spark-ignitedpressure traces
13%13% Internal EGRInternal EGR 58%58%1515
Conventional
2500 NOx (ppm)
2500NOx (ppm)
0
500
1000
1500
2000
NO
x, p
pm
0
COV IMEP (%)COV IM
500
1000
1500
2000
NO
x, p
pm
EP (%)1212 Discontinuity
between transition range and spark assisted HCCI
CO
V IM
EP, %
CO
V IM
EP, %
99
66
33
00
EGR
Input Hidden OutputLayer Layers Layer
� We are validating KIVA4Z againstSandia data
6 Lawrence Livermore National Laboratory
Requires SparkRequires Spark No Spark
Exhaust Valve Closing, ATDCExhaust Valve Closing, ATDC360360 3203434 320 300300 280280 260260 240240 220220
Transition HCCISpark Assisted HCCI
In the last few years we have been able to demonstrate high fidelity modeling tools for homogeneous (HCCI) combustion
9000
Solid lines: experimental
8000 Φ=0.26 Dashed lines: numerical5.88 mm 5.23 mm 9.01 mm 0.305 mmΦ=0.24 (0.2314 in) (0.206 in) (0.355 in) (0.0120 in)
Φ=0.22 7000 Φ=0.20
Φ=0.18Φ=0.16
6000 Φ=0.14Φ=0.12Φ=0.105000 Φ=0.08Φ=0.06
4000 4000 Φ=0.04 102 mm (4.02 in)
3000
2000-30 -20 -10 0 10 20 30
crank angle, degrees.
Pre
ssur
ee, k
Pa.
Lawrence Livermore National Laboratory 7
We have demonstrated unprecedented modeling fidelity accurately predictingg exhaust composition of 50 sppeciesy p p
Isooctane Formaldehyde 2-Methyl-1-propene
Acetone Methane Ethene
8 Lawrence Livermore National Laboratory
Accurate prediction of specific exhaust components possible due to synergies in collaborators’ unique capabilities
Analytical chemistry forAnalytical chemistry for detailed exhaust speciation
G
High quality HCCI engine experiments
(S di ) Extensively validated chemical kinetic models
Gasket Volume
(Sandia) chemical kinetic models Piston/Liner Crevice
9 Lawrence Livermore National Laboratory
High fidelity engine analysis
Speciation experiments are optimum for model validation & mechanism tuning
- -
We are now conducting an all-out effort on partially stratified combustion
1. KIVA3V with CHEMKIN calculations 2. KIVA multi-zone (KIVA3V-in every cell: months in 100 processor MZ-MPI, Kiva4-MZ): 1 week in computer. For benchmarking only 100 processor machine
T
P τ
……
……
Ignition Input Delay
Variables φ Time
EGR EGR
Input Hidden OutputLayer Layers Layer
3. KIVA artificial neural network (KIVA ANN) 4 h i i l 4. KIVA sequential multi zone(KIVA-ANN): 4 hours in single cloud model: 1 day in single
4 KIVA-sequential multi-zone
10 Lawrence Livermore National Laboratory
processor computer processor computer
We are now modeling Sandia PCCI experiments with KIVA3V-MZ-MPI
11 Lawrence Livermore National Laboratory
We are now modeling Sandia PCCI experiments with KIVA3V-MZ-MPI
12 Lawrence Livermore National Laboratory
Our artificial neural network model (KIVA-ANN) can predict partiallystratified PCCI combustion with 1000 times less computational effort
T
…
τ T
P
EGR
Ignition Delay Time φ
Input Variables …
Output Layer
Hidden Layers
Input Layer
KIVA-ANN ppredictions of the effect of stratification on PCCI combustion agree well with KIVA3V-MZ-MPI
13 Lawrence Livermore National Laboratory
We are analyzing ORNL gasoline SI-HCCI transition experiments with a 1-D flame propagation/autoignition code
13%13%13%13% Internal EGRInternal EGRInternal EGRInternal EGR 58%58%58%58%
2000
2500
12
15
%
NOx (ppm)
COV IMEP (%) 2000
2500
12
15
%
NOx (ppm)
COV IMEP (%)Conventional Discontinuity
between transition range and spark
1000
1500
NO
x, p
pm
6
9 C
OV
IMEP
, %
1000
1500
NO
x, p
pm
6
9C
OV
IMEP
, %
range and spark assisted HCCI
0
500
220240260280300320340360
E h t V l Cl i ATDC
0
3
0
500
220240260280300320340360
E h t V l Cl i ATDC
0
3
Exhaust Valve Closing, ATDCExhaust Valve Closing, ATDC
Transition HCCISpark Assisted HCCIRequires Spark Requires Spark No Spark
Heat Ignition transfer
to wallsFlame propagation
14 Lawrence Livermore National Laboratory
Our 1-dimensional kinetics-based flame propagation model has been able to predict pressure and burn fraction as a function of crank angle
Experiment Experiment
1D-flame model 1D-flame model
In progress are simulations of sequentially running multiple cycles to determine combustion stabilityy
15 Lawrence Livermore National Laboratory
We have been working with David Torres at LANL to integrate our multizone model into KIVA4 (KIVA4-MZ)
KIVA-3 MZ KIVA-4 MZ
16 Lawrence Livermore National Laboratory
Kiva4MZ is significantly faster and more accurate because of unstructured grids and parallel computing
� Kiva4MZ enable analysis of ultra-high resolution grids (1 million to 10 million cells)(1million to 10 million cells)
We are validating KIVA4MZ with Sandia HCCI Engine ExperimentsWe are validating KIVA4MZ with Sandia HCCI Engine Experiments
17 Lawrence Livermore National Laboratory
Technical Publications during FY07-08 1. Analysis of Homogeneous Charge Compression Ignition (HCCI) Engines for Cogeneration Applications, Salvador M. Aceves, Joel Martinez-
Frias, Gordon M. Reistad, Journal of Energy Resources Technology, Vol. 129, No. 4, pp. 332-337, 2007.
22. Gaseous Fuel Injection Modeling using a Gaseous Sphere Injection Methodology Randy P Hessel Salvador M Aceves Daniel L Flowers SAEGaseous Fuel Injection Modeling using a Gaseous Sphere Injection Methodology, Randy P. Hessel, Salvador M. Aceves, Daniel L. Flowers, SAE Paper 2006-01-3298, 2006.
3. Fast Prediction of HCCI Combustion with an Artificial Neural Network Linked to a Fluid Mechanics Code, Salvador M. Aceves, Daniel L. Flowers, J.Y. Chen, Aristotelis Babajimopoulos, SAE Paper 2006-01-3298, 2006.
4. A Simple HCCI Engine Model for Control, Nick Killingsworth, Salvador Aceves, Daniel Flowers, Miroslav Krstic, Proceedings of the IEEE I t ti l C f C t l A li ti M i h G 2006International Conference on Control Applications, Munich, Germany, 2006.
5. A Comparison on the Effect of Combustion Chamber Surface Area and In-Cylinder Turbulence on the Evolution of Gas Temperature Distribution from IVC to SOC: A Numerical and Fundamental Study, Randy P. Hessel, Salvador M. Aceves, Daniel L. Flowers, SAE Paper 200601-0869.
6. Effect of Charge Non-uniformity on Heat Release and Emissions in PCCI Engine Combustion, Daniel L. Flowers, Salvador M. Aceves, Aristotelis BBabbajijimopoullos, SAE Paper 20062006 -0101-13631363, 2006. SAE P 2006
7. Overview of Modeling Techniques and their application to HCCI/CAI Engines, Salvador M. Aceves, Daniel L. Flowers, Robert W. Dibble, Aristotelis Babajimopoulos, in HCCI and CAI Engines for the Automotive Industry, Edited by Hua Zhao, CRC Press, Woodhead Publishing Limited, Chapter 18, pp. 456-474, 2007.
8. Improving Ethanol Life Cycle Energy Efficiency by Direct Utilization of Wet Ethanol in HCCI Engines, Joel Martinez-Frias, Salvador M. Aceves, D iDaniell L. Fl P IMECE2005 79432 P di f th ASME I t ti l M h i l E i i C d E hibiti 2005L Flowers, Paper IMECE2005-79432, Proceedings of the ASME International Mechanical Engineering Congress and Exhibition, 2005, Accepted for Publication, Journal of Energy Resources Technology.
9. Effect of Laser-induced Excitation of Oxygen on Ignition in HCCI Engines Analyzed by Numerical Simulations, Daniel L. Flowers, Salvador M. Aceves and Robert W. Dibble, Combustion Theory and Modelling, Vol. 11, No. 3, pp. 455-468, 2007.
10. Extremum Seeking Tuning of an Experimental HCCI Engine Combustion Timing Controller, Nick Killingsworth, Dan Flowers, Salvador Aceves, Mi l K i P di f h A i C l C f N Y k NY J l 200Miroslav Krstic, Proceedings of the American Control Conference, New York, NY, July 2007.
11. In Pursuit of New Engine Dynamics, Daniel Flowers, Nick Killingsworth and Robert Dibble, Mechanical Engineering, pp. 20-21, July 2007. 12. A Numerical Investigation into the Anomalous Slight NOx Increase when Burning Biodiesel: A New (Old) Theory, George A. Ban-Weiss, J.Y.
Chen, Bruce A. Buchholz, Robert W. Dibble, Fuel Processing Technology, Vol. 88, pp. 659-667, 2007. 13. Pathline Analyysis of Full-cyycle Four-stroke HCCI Enggine Combustion Usingg CFD and Multi-Zone Modelingg, Randyy P. Hessel, David E. Foster,
Richard R. Steeper, Salvador M. Aceves, Daniel L. Flowers, SAE Paper 2008-01-0048.
18 Lawrence Livermore National Laboratory
14. Modeling Iso-octane HCCI using CFD with Multi-Zone Detailed Chemistry; Comparison to Detailed Speciation Data over a Range of Lean Equivalence Ratios, Randy P. Hessel, David E. Foster, Salvador M. Aceves, M. Lee Davisson, Francisco Espinosa-Loza, Daniel L. Flowers, William J. Pitz, John E. Dec, Magnus Sjöberg, Aristotelis Babajimopoulos, SAE Paper 2008-01-0047.
t t
Collaboration: We have long standing partnerships with industry, national labs, and academia
� I dIndustry PPartners: • Collaborative modeling and analysis of experiments
� National Labs and Universities: • Modeling tools, experiments, analysis
19 Lawrence Livermore National Laboratory
te at o a Jou a o e esea c
Other collaborative activities � Ongoing participation at MOU meetings with vehicle and engine
manufacturers
� FACE working group
� HCCI symposium in Lund Sweden (2 invited presentation)HCCI symposium in Lund Sweden (2 invited presentation)
� Several collaborative publications involving National Laboratories and Universities in US and abroad: • Participation at SAE technical meetings and symposia • The Combustion Institute • International Journal of Enggine Research • Combustion Theory and Modeling • Journal of Energy Resources Technologies • IEEE Control Systems Technology IEEE Control Systems Technology
20 Lawrence Livermore National Laboratory
• 8 PhD and >15 MS through collaboration and direct support
- -
Future plans: We will complete validation of our PCCI codes by comparison with Sandia iso-octane results and exhaust speciation values
1. KIVA3V with CHEMKIN calculations in every cell: months in 100 processor computer. For benchmarking only
T
P
……
Ignition Input τ Delay
Variables φ Time
EGR EGR
……
Input Hidden OutputLayer Layers Layer
3. KIVA artificial neural network (KIVA ANN) 4 h i i l(KIVA-ANN): 4 hours in single
2. KIVA multi-zone (KIVA3V-MZ-MPI): 1 week in 100 processor machine
4 KIVA-sequential multi-zone4. KIVA sequential multi zone cloud model: 1 day in single
21 Lawrence Livermore National Laboratory
processor computer processor computer
Summary: We continue to develop high fidelity HCCI and PCCI analysis techniques with greatly improved computational efficiency
• Detailed exhaust species simulations demonstrate a powerful tool for validationdemonstrate a powerful tool for validation and chemical mechanism development
• We are now conducting an all-out effort on partially stratified combustion
22 Lawrence Livermore National Laboratory
parallel computing
• Kiva4MZ is significantly faster and more accurate because of unstructured grids and accurate because of unstructured grids and