Integrated Aftertreatment System
Modeling Using GT-POWER
Syed Wahiduzzaman, Weiyong Tang and Seth Wenzel
Gamma Technologies
10th DOE Crosscut Workshop on Lean Emissions Reduction Simulation
May 1st - 3rd, 2007
Motivations
• Motivations
– Tremendous need for the ability to simulate AT
systems
– Design/development process is increasingly a
collaborative endeavor
– Variety of domain experts (e.g. test, performance,
AT systems, control/calibration) frequently using
disparate tools
– Overall system optimization is ultimately the goal
– A common simulation environment is a necessity
Goals and Approach
• Goals
– Provide an integrated tool
– Flexible environment capable of handling customized kinetics
– Facilitate communication and collaboration between various
domain experts
– Minimize learning curve
– Eliminate the need of users written code to input kinetics
• Approach
– Specialized AT object (DPF, CatalystBrick) as flow component
– Chemistry library templates to input customized kinetics
– Built directly into GT-Power library sharing existing
capabilities, look and feel
Concept: Chemistry Library
Templates
Burner, hydrolysis
Catalyst
ChemKin
Catalyst with storage
Geometric and Thermal
Specifications
Catalyst Brick
Diesel Particulate
Filter
0D or 1D
Kinetics Specification
GlobalReactions
Template
Principle
• Chemistry component turns a flow
component into a reactor when connected
to it
Collaborative Simulations
• Test cell data: Calibration of engine
models, generation maps and fits (e.g.
emissions, mean value, neural network
models)
• Generation of boundary conditions for
standalone simulation of AT unit models
(e.g. simulated test cycle data)
• Validation of unit models
• Integration of unit models into system
model
• Control system modeling
Generation of BC for Standalone
AT Models
EngCylEmisMaps
SpeciesSampler
(from Test cell)
To AT Group
SpeciesSampler Output:
Pressure Concentrations
Temperature
Flow Rate
An standalone Model for
Catalyst Domain Experts
Kinetics
SpeciesSampler or CatalystBrick
Experimental Data
Examples
and
Applications
DPF Model
Lumped model P, T, fm
Uniform temperature at gas and solid phases
Uniform wall flow distributions
Computational efficiency
P1, T1, fm1
1D quasi-steady model Ts
P2, T2, fm2
Axially different temperature and pressure
Non-uniform wall flow distributions
Detailed information
Deep Bed Filtration/Pressure Drop
Deep-Bed Filtration Model
wall flow
Schematic of Pressure Drop
Soot Deposit Inlet Channel
wall flow Wall
min,1
1 mc,1 = E1 x min,1 ∆P1 ∆P2
Plug
∆P3
min,2 = min,1 – mc,1 Wall
Substrate Wall Thickness
2 mc,2 = E2 x min,2
min,3 = min,2 – mc,2 ∆P4
∆P5 ∆P6
• •
• •
• •
n-1 mc,n-1 = En-1 x min,n-1
Outlet Channel
min,n = min,n-1 – mc,n-1
n mc,n = En x min,n
∆Ptotal = ∆P1 + ∆P2 + ∆P3 + ∆P4 + ∆P5 + ∆P6
mout = min,n – mc,n
Channel Length
Study deep bed filtration using theories
of uniform spherical collectors
and Brownian diffusion.
Experimental Validations
Table 1: DPF Specifications for Simulation
Type EX-80* CCRT**
Trap Diameter (mm) 229 267
Channel Length (mm) 305 305
Channel Width (mm) 1.5 1.49
Wall Thickness 0.31 0.305
Number of Inlet Channels 6013 8659
Wall Porosity 0.48 0.52
Pore Diameter (mm) 0.0125 0.013
Bulk Density (kg/m3) 1400 1400
Specific Heat (J/kg-K) 1120 1120
Thermal Conductivity (W/m-K) 2.092 2.092
* Uncatalyzed, made by Corning;
** Catalyzed, made by Johnson Matthey.
Experimental Validation:
with an Uncoated EX-80
7 10
Experimental
6 Lumped model
PM Loading Density (g/L)
8
1-D model
Pressure Drop (kPa)
5
6
4
4
3
2
2
1 0
0 1 2 3 4
Time (hr)
Symbols: measurements as Fig. 5 in SAE 2002-01-1019;
Lines: predictions.
Experimental Validations:
with an Coated CCRT
Experimental, Lumped model, 1-D model
14 140 16
14
12 120
Pressure Drop (kPa)
12
Pressure Drop (kPa)
10 100
10
Mass (g)
8 80
8
6 60 experimental
retained mass 6 Lumped model
4 40 1-D model
4
oxidized mass
2 20 2
0 0 0
0 1 2 3 4 5 6 7 8 0 10 20 30 40 50 60 70 80 90
Time (hr) Mass Deposited (g)
Symbols: measurements as Figs. 5&6 in SAE 2006-01-0879;
Lines: predictions.
Experimental Validation:
Active Regeneration by Fuel Injection
22 1100
Experimental , inlet X/L = 0/14
20 , X/L = 2/14
1-D model
, X/L = 7/14
Pressure Drop (kPa)
1000
18 , X/L = 10/14
Temperature (K)
16 900
14
800
12
10 700
8
600
6
4 500
0 50 100 150 200 0 50 100 150 200
Time (sec) Time (sec)
Symbols: measurements as Figs. 8&A.2 in SAE 2003-01-0842;
Lines: predictions.
Kinetics Templates
GaseousReactions Template
• Elementary gas-phase kinetic mechanisms
• Reactions can be directly typed into the
template or imported from a file
• Advanced feature set
– Reversible and irreversible reactions
– Generalized Arrhenius formulation (A*f(T)*g(P)*Exp(-Ta/T))
– Pressure-Dependent reactions, including:
• Third-body reactions
• Reactions with low pressure limit
• Troe falloff
• SRI falloff
GlobalReactions Template
• Global kinetic mechanisms
• Arbitrary reaction order and
concentration expression
• Generalized inhibition functions
including Langmuir and Hinshelwood
type
• Support various types of rate
specifications
• Diffusion from and to surface
SurfaceReactions Template
• All relevant features of GlobalReactions
Template
• Coverage calculation
• Supports storage and arbitrary functions
of coverage in the rate specification
Building a SCR Model
Step 2:
1: Link
Placeparts
necessary
to create
flow
a flow
objects
model
into project map
Note reaction kinetics has not yet been specified to the SCR.
Building a SCR Model (Cont.)
Step 3: Input global kinetics to a chemistry template and link
it to flow object SCR-01
Global kinetics
Building a SCR Model (Cont.)
Step 4: Construct associated control and sensor parts
Reaction kinetics
Catalytic device
NH3 injection controller
Species sensors and
NOx conversion cal.
Building a SCR Model (Cont.)
Step 5: Unit model testing, calibration and validation
Building a SCR Model (Cont.)
Step 6: Add subsytems
Building a SCR Model - Advanced
• User can optionally account for NH3 storage using
‘SurfaceReactions’ template:
Absorption: NH3 + S -> S(NH3)
Desorption: S(NH3) -> NH3 + S
• Where ‘S’ represents a general surface site.
• Specific sites can also be studied: WO3, Ti2O, V205
Building a SCR Model - Advanced
• User can optionally account for NH3 storage using
‘SurfaceReactions’ template:
Building a SCR Model - Advanced
Desorption: S(NH3) -> NH3 + S
Steady State
Absorption: NH3 + S -> S(NH3)
Building a SCR Model - Advanced
User can optionally account for NH3 storage using
‘SurfaceReactions’ template:
DOC + LNT Example
Lean Nox Trap (LNT) Reactions:
NOx storage forming nitrite:
BaCO3 + 2NO2 + .5O2 => Ba(NO3)2 + CO2
BaCO3 + 2NO + 1.5O2 => Ba(NO3)2 + CO2
NOx regeneration:
NOx release:
Ba(NO3)2 + 3CO => BaCO3 +2NO + 2CO2
Ba(NO3)2 + H2 + CO2 => BaCO3 +2NO2 + H2O
NOx reduction:
CO + NO => 0.5N2 + 2CO2
NO2 <=> NO + 0.5O2
DOC + LNT Example
Diesel Oxidation Catalyst (DOC) Reactions:
CO + 0.5O2 => CO2
C3H6 + 4.5O2 => 3CO2 + 3H2O
H2 + 0.5O2 => H2O
NO+0.5O2 => NO2
NO2 => NO+0.5O2
HC + 20.8O2 => 14.6CO2 + 12.4H2O
Note HC stands for diesel fuel (C14.6H24.8).
DOC + LNT Example
DOC + LNT Example
DOC + LNT Example
Lean Operation 0-10s Rich Operation 10-20s
Kinetic Parameter
Identification
GlobalReactions Template
Application: Three-Way Catalyst
CO oxidation:
CO + 0.5O2 => CO2
HC (unburned and partially burned) oxidations:
CH4 + 2O2 => CO2 + 2H2O
C3H6 + 4.5O2 => 3CO2 + 3H2O
C3H8 + 5O2 => 3CO2 + 4H2O
NO oxidation:
CO + NO => CO2 + 0.5N2
NO + 0.5O2 => NO2
NO2 => NO + 0.5O2
H2O formation:
H2 + 0.5O2 => H2O
Kinetics of 3-Way Catalyst
Inhibition Terms:
Global Reaction Scheme
K expressions for near
equilibrium behavior:
Individual Production Rates:
Parameter Identification
• Original kinetic set is not applicable and has to
be calibrated
• Parameter optimization are often carried out on:
1. Pre-exponential factor Ai
2. Activation temperature/energy Ei
• Optimization of a total of 16 parameters
8 reactions x 2 values = 16 parameters
Design of Experiments (DoE)
TWC: Experimental Inlet Conditions
Minor species inlet mole fractions
Inlet exhaust temperature and mass flow rate
Test Rig Model
Global Reactions
used to define
chemical kinetics
Test Rig Model (Cont.)
Error = Sum{(measured-predicted)^2}
User Defined
Error Function
Input for
Measured Data
DOE Sensitivity Study
• Built-in DOE tools is used to perform sensitivity study
• Latin Hypercube partial factorial
DOE Sensitivity Results
Sensitivities
Error Function value
of
200 experiments
DOE Sensitivity Study
• Discovered that the following 3 equations most
significantly affected results:
R1
– R1 and R4 most sensitive
– R2 also sensitive
R4
R2
Final Optimized Parameters
Pre-exponential factor, Activation temperature,
Ai (mole•K/cm3•S) Ei (K)
R#
Optimum Original Optimum Original
1
1.04E+11 1.0E+11 9828.612 9622
2
1.39E+12 1.39E+13 11000 12500
3
3.66E5 3.66E+6 16000 16000
4
3.37E+7 2.2E+7 11967.34 10171
5
8.00E+6 6.7E+7 2100 2000
6
1.39E+12 1.39E+13 11000 12500
7
1.43E+09 4.5E+9 8419.25 8419.25
8
7.71E+7 1.0E+9 8419.25 8419.25
System Integration and
Simulation
Integrated Model of
Engine, Vehicle and DPF Systems
Typical Results of System
Simulation
Summary I
• Highly comprehensive and flexible solution:
not hardwired
• Specialized models for DPF
• General tools for chemical/catalytic reactions
• Solves user supplied reaction sets
(mechanisms)
• Can be encrypted into a subassembly or
compound object (black-box)
• Integrated tool: multiple catalysts in series
(+engine + vehicle)
Summary II
• Single simulation environment (no third
party tool is needed)
• Built-in DOE optimization for parameter
identification
• Can be run as standalone unit model
• or as integrated simulation in GT-SUITE
New Developments
• Faster Solvers
– Method of lines (Cavendish, Oh, Bissett)
– Quasi-steady hybrid-boundary solution
– Up to two orders of magnitude faster
• 2-D Conduction in
CatalystBrick and DPF
• Kinetics-based combustion
and emissions models
Thank You!
10th DOE Crosscut Workshop on Lean
Emissions Reduction Simulation
May 1st - 3rd, 2007
