Upload
buikien
View
224
Download
0
Embed Size (px)
Citation preview
Fuel Efficiency and EmissionsOptimization of Heavy-Duty Diesel
Engines using Model-BasedTransient Calibration
Chris Atkinson Atkinson LLC
& Marc Allain, Houshun Zhang, Yury Kalish & Craig Savonen
Detroit Diesel Corporation
Atkinson LLC
Presentation Topics
Background – Diesel Engine Industry Trends Motivation – Increasing Control and Calibration Burden Solution – Model-Based Calibration and Rapid Transient Calibration Optimization Process Results – Validation of Fuel Efficiency Improvements and Emissions Compliance Next Steps – Further Improvements using Model-Based Control and Diagnostics Summary
Atkinson LLC DEER Conference August 2007
2
Background – Industry Trends
Ever Decreasing Emissions Levels Increasing Demands for Fuel Efficiency Rapidly Increasing Complexity of Engine Control Close Integration of Aftertreatment Systems New Combustion Regimes under Investigation Emissions Regulations – FTP, In-Use, NTE HD OBD Requirements ⇒ All Of These Lead To A Significantly Increased
Calibration and Optimization Burden. Atkinson LLC
DEER Conference August 2007 3
Diesel Engine Control Complexity
Year Engine Technologies Number of Calibrateable Control Parameters
1998 Injection Timing Injection Pressure Smoke Control
3
2004 EGR 4 - 5 VGT
2007 2010+
DPF Regeneration Injection Timing (multiple) Injection Pressure or Rate Shaping EGR
6 - 7 11 - 15+
Full Air Path Management Active Aftertreatment Control
Atkinson LLC DEER Conference August 2007
4
Full Factorial Calibration Space
Year Engine Control Parameters Number of Discrete Test Setpoints
1998 Speed, Load, Injection Timing & Pressure 10,000 2002 Speed, Load, Injection Timing & Pressure, EGR 100,000 2004 Speed, Load, Injection Timing & Pressure,
EGR, Turbocharger Control 1,000,000
2007 Speed, Load, Injection Timing, Injection Pressure, EGR, Turbocharger Control, Particulate Filter Regeneration
10,000,000
2010 Speed, Load, Multiple Injection Timing, Injection Pressure, EGR, Turbocharger Control, Aftertreatment Controls
10,000,000,000
Atkinson LLC DEER Conference August 2007
5
The Challenge for Calibration
The Curse of Dimensionality – a 2007 specification diesel engine has 107 test points if tested in a full factorial experiment Calibration requirements will increase by 2-3 orders of magnitude by 2010-2014 due to new engine technologies Design of Experiments can reduce the overall engine steady-state mapping burden significantly (perhaps by a factor of 100), but this still results in huge experimental matrices Most importantly, steady state engine mapping not well suited to TRANSIENT emissions regulations, fuel consumption reduction, performance and aftertreatment regeneration.
Atkinson LLCDEER Conference August 2007
6
Motivation – The Engine Development Dilemma
How can engine calibration be performed quicker, better and cheaper?
How do we improve real-world fuel efficiency while maintaining very low
regulated exhaust emissions?
How do we maximize the brake thermal efficiency of any given engine and
aftertreatment hardware set?
How can we further increase fuel efficiency through smart calibration and
control?
How can we do all of this without further overloading transient engine test
facilities and without incurring significant extra expense?
Atkinson LLCDEER Conference August 2007
7
Solution – Model-Based Calibration
The Problem:
How can engine calibration be performed quicker, better and
cheaper?
The Solution:
Transfer the majority of the calibration effort out of the
transient test cell and onto the engineer’s desktop, using a
systematic MODEL-BASED Rapid Transient Calibration
approach. Atkinson LLC DEER Conference August 2007
8
The Model-Based Calibration Process
Design of Transient Experiments Transient cycles Control and calibration perturbations
Data Collection Dynamic Engine Modeling
Real-time transient engine performance, emissions and operating states – with full inertial, thermal, EGR and air path dynamics included
Optimization Off-line optimization of engine calibration Multiple simulations with varying optimization weights
Verification and Validation of Calibration Data Sets
Atkinson LLCDEER Conference August 2007
9
First Application of this Technology
Utilized a 2007-specification DDC Series 60 engine with EGR, VGT and DPF. 15 custom transient cycles developed and implemented (~10 dynamometer days). Dynamic engine models developed off-line. Calibration optimization performed off-line. Simulated engine performance performed off-line. Re-visited transient engine test cell for validation and verification (~10 dynamometer days).
Atkinson LLCDEER Conference August 2007
1
Engine Control Unit(ECU)
DEER Conference August 2007
Atkinson LLC
Data Collection
Engine Dynamic test
cycle speed and
load requirements
Engine Control Unit
Calibration Maps and Lookup Tables
TRANSIENT DATA Emissions
Performance Fuel Economy
Map and Lookup Table Perturbations using Rapid Prototyping
Engine Control System
1
Transient Engine Data Coverage
Atkinson LLC DEER Conference August 2007
1
Air-EGR Perturbations (all cycles)
Atkinson LLC DEER Conference August 2007
1
EGR-Injection Timing Perturbations
Atkinson LLC DEER Conference August 2007
1
Injection Pressure-Timing Perturbations
Atkinson LLC DEER Conference August 2007
1
Dynamic Engine Modeling Approach
Dynamic (transient) models use a combination of Physical Modeling
First principles Equation-based Phenomenological
Heuristic ModelingData-driven Learning Data from actual engine operation (real time emissions, performance, fuel consumption and operating states).
Atkinson LLCDEER Conference August 2007
1
DEER Conference August 2007
Atkinson LLC
Diesel Engine Modeling
Equation-Based and
Data-Driven Engine Model
INPUTS
Engine Speed Fuel Quantity
Injection Timing Injection Pressure
Exhaust Gas Recirculation Variable Geometry Turbocharger Setting
OUTPUTS Torque
Emissions •HC •CO •NOx •CO2 •PM
Exhaust Temperature Exhaust Backpressure
Turbo Speed Combustion Pressure
DDC S60 Diesel Engine Model
1
1DEER Conference August 2007
Atkinson LLC
Transient Engine Modeling Results -Torque
Transient Engine Modeling Results -NOx
Atkinson LLC DEER Conference August 2007
1
Transient Engine Modeling Results -PM
Atkinson LLC DEER Conference August 2007
2
DEER Conference August 2007
Atkinson LLC
Rapid Transient Calibration Optimization
Equation-Based and Data-Driven
Engine Model
PREDICTED OUTPUTS Emissions
Performance Fuel Economy
Calibration Optimizer (variable weights)
COMMANDS Engine cycle
dynamic requirements
Engine Control Unit (ECU)
Calibration Maps and Lookup Tables
2
Calibration Optimization Functions
Cost function – minimize fuel consumption, Subject to the constraint of meeting NOx and PM emissions integrated across a transient cycle, While not exceeding certain engine operating state parameter levels, such as turbocharger speed, peak cylinder pressure, exhaust temperature and peak injection pressure, While also meeting NTE and steady-state exhaust emissions levels.
Atkinson LLCDEER Conference August 2007
2
Typical Calibration Optimization Progress – cell by cell
Atkinson LLCDEER Conference August 2007
2
DEER Conference August 2007
Atkinson LLC
Simulated Emissions Results – I (~100 FTPs with different calibration optimization weights)
CALIBRATION SIMULATION RESULTS
0
0.25
0.5
0.75
1
1.25
0 0.25 0.5 0.75 1 1.25
N Ox [ nor m a l i z e d]
S88 S88
S88
S88
S87
S87 S87
S87
S86S86
S86 S86
S8S85
S85S85
S84
S84
S84 S84
S83
S83
S83
S83 S8S82
S82 S82
S81 S81
S81S81S80
S80
S80
S80
S79S79 S79
S79 S7S78
S78S78
S02-7
S02-7
S02-S02-7
S02-6S02-6
S02-6S02-6
S02-5
S02-5
S02-5S02-5
S74 S74
S74S74
S03 S03
S03
S03
S02S02
S02S02
S03
S03
S02 S02
S01
S01 S01
S01
2
Simulated Emissions Results – II
1.200
1.150
1.100
1.050
1.000
0.950
0.900
0.850
0.800 0.000 0.250 0.500 0.750 1.000 1.250
N O x E M IS S IO N S [ n orma lize d ]
Atkinson LLC DEER Conference August 2007
2
Validation and Verification – Actual FTP Emissions Results - I
PM
[nor
mal
ized
] 1.25
0.00 0.25 0.50 0.75 1.00
NOx [normalized]
Baseline S01M S02M 1.00
S03M S04M S02N
S02O S02P S02Q 0.75
S05 S06 S07
S02R S02S S02T 0.50
S01-R S07-R S03N
S04N S08N S09N
0.25 S10N S08N S11
S12 S13 S14
0.00 S15
Atkinson LLC DEER Conference August 2007
2
Validation and Verification – Actual FTP Emissions Results - II
HC
[nor
mal
ized
] 1.20
1.00
0.00 0.25 0.50 0.75 1.00
NOx [normalized]
Baseline S01M S02M
S03M S04M S02N
0.80 S02O S02P S02Q
S05 S06 S07 0.60
S02R S02S S02T
S01-R S07-R S03N 0.40
S04N S08N S09N
0.20 S10N S08N S11
S12 S13 S14
0.00 S15
Atkinson LLC DEER Conference August 2007
2
Validation and Verification – Actual FTP Emissions Results - III
1.20
1.15
0.00 0.25 0.50 0.75 1.00
NOx [normalized]
Baseline S01M S02M
1.10 S03M S04M S02N
S02O S02P S02Q 1.05
S05 S06 S07 1.00
S02R S02S S02T
0.95 S01-R S07-R S03N
S04N S08N S09N 0.90
S10N S08N S11 0.85
S12 S13 S14
0.80 S15
Atkinson LLC DEER Conference August 2007
Bra
ke T
herm
al E
ffici
ency
[nor
mal
ized
]
2
Project Results
Rapid Transient Calibration process requires two periods ofabout 10 dynamometer days each in the transient test cell
Data collection – 15 custom transient cycles Verification and validation
4 calendar months beginning to end Improved real-world fuel efficiency (~5%) across the FTPwhile meeting the same regulated emissions levels, with thesame levels of engine performance and protection. Shifted a significant portion of the calibration burden out ofthe high cost, high demand transient test cell.Process is scalable – adding an extra control or calibration variable (such as urea dosing for SCR) will add effort, butnot in a geometric fashion.
Atkinson LLC DEER Conference August 2007
2
Future Applications
Control Strategy Development Using Point-by-Point calibration optimization (determining the optimum set of control parameters on a time-based basis, rather than on a table-based basis) gives insight into transient control strategies.
Model-Based Control System Development Utilize dynamic engine modeling techniques for real-time on-line, on-engine application, rather than off-line use.
Model-Based Diagnostics Dynamic modeling for real-time Virtual Sensing.
Atkinson LLCDEER Conference August 2007
3
Summary
Model-Based Rapid Transient Calibration is able to Meet prevailing low emissions standards With improved fuel economy With significant reductions in the time and effort required to calibrate.
The key enabling technologies are dynamic engine modelingand off-line optimization.Reductions in engineering time, cost and effort have beenshown with this beginning-to-end model-based calibrationoptimization process.Process is scalable – encouraging for even more complex future engine and aftertreatment systems.Further gains in on-line engine control, calibration,optimization and diagnostics are possible using the samemodel-based methods. Atkinson LLC
DEER Conference August 2007 3
Acknowledgements
• U.S. Department of Energy • Gurpreet Singh • Roland Gravel
• National Energy Technology Laboratory• Carl Maronde • Jeffrey Kooser
• Atkinson LLC •Gregory Mott
Atkinson LLCDEER Conference August 2007
3
Atkinson LLC DEER Conference August 2007
3
DEER Conference August 2007
Atkinson LLC
Design of Transient Experiments
BTRQ NOx HC CO PM
RPM Fresh air setpoint EGR setpoint Injection parameters (timing & pressure)
Offline Equation and Data-based Transient Engine
Model
Transient Engine Model
BTRQ NOx HC CO PM BSFC
Optimization scheme
RPM Fueling rate Fresh air setpoint EGR setpoint Injection parameters (timing & pressure)
Air
EGR
Inj.
3
Transient Engine Test CycleEn
gine
Spe
ed (r
pm) a
nd T
orqu
e (N
m)
2250
1750
1250
750
250
-250
0 200 400 600 800 1000
Transient Engine Test Cycle
Time (s)
Atkinson LLCDEER Conference August 2007
3
3000
Typical Table-Based Output EGR Table
10
5
0
EGR
[kg/
min
] EG
R [k
g/m
in]
30002000
200 300 400
Speed [rpm] EGR Table Original
Fuel [mg/stroke]
10
5
0
1000 0 0 100
2000 1000
0 0 100 200 300 400
Speed [rpm] Fuel [mg/stroke]
Atkinson LLCDEER Conference August 2007
3
Simulated Emissions Results – III (Integrated FTP emissions as a function of optimization weights)
NOx_Weight CO_Weight PM_Weight CO2_Weight NOx CO PM CO2 EMISSIONS NOx -0.62 0.31 0.26 -0.06 1 CO 0.43 -0.61 -0.36 -0.05 -0.67 1 PM 0.36 -0.36 -0.58 0.19 -0.51 0.81 1 CO2 -0.05 0.09 0.07 -0.37 0.05 0.00 0.07 1
Increasing NOx, CO, PM or CO2 weights in the optimization function results in a reduction in those emissions NOx and PM emissions are anti-correlated, as expected, while CO and PM are co-correlated.
Atkinson LLCDEER Conference August 2007
3
Background – Industry Trends
Ever Decreasing Emissions Requirements Ever Increasing Fuel Economy and Brake Thermal Efficiency Demands Increasing Mechanical and Electronic Complexity of Engines and Aftertreatment Systems Cost Reduction Demands for Engine Development Reduced Product Engineering Development Cycles Increasing Demands on Transient Engine and Emissions Test Facilities
Atkinson LLCDEER Conference August 2007
3