Mack James 2016 Thesis

Calibration of Automotive Aftertreatment Models through Co-Simulation with MATLAB
Optimization Routines
THESIS
Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in
the Graduate School of The Ohio State University
By
James Mack
Graduate Program in Mechanical Engineering
The Ohio State University
2016
Master's Examination Committee:
Dr. Shawn Midlam-Mohler, Advisor
Prof. Giorgio Rizzoni

Copyright by
James Mack
2016
Abstract
New and existing government regulations mandate limits on various automotive
exhaust tailpipe-out species including but not limited to Oxides of Nitrogen (𝑁𝑂𝑥 ),
Carbon Monoxide (CO), Unburned Hydrocarbons (THC), and Particulate Matter (PM).
Automotive aftertreatment systems allow for the mitigation of harmful engine-out species
by converting pollutants into less harmful products by the time they reach the tailpipe.
Traditionally, catalytic converters have been used in stoichiometric Gasoline Port Fuel
Injected (PFI) applications for reduction of gaseous emissions while particulate filters
have been used in diesel applications to reduce PM. Gasoline Direct Injected (GDI)
engines pose potential regulatory difficulties since unlike PFI, GDI vehicles emit PM at
levels near regulatory limits. If manufactures cannot improve GDI engine control
strategies to reduce PM levels, a Gasoline Particulate Filter (GPF) may be a required
addition to GDI aftertreatment systems.
GDI aftertreatment systems consisting of Three-Way Catalytic Converters (TWC)
and GPFs can be developed in commercial automotive powertrain modelling packages.
The performance of candidate component configurations can be virtually tested and
evaluated; however before this can occur individual components must first be calibrated
to insure modelled performance is close to reality. In this work 1-D models for both a
TWC and a GPF were modelled in the powertrain modelling package GT-Power (GT-P).
ii
A simplified version of the kinetic set proposed by Ramathan and Sharma [42] was
utilized within the TWC while the GPF was modelled as a non-catalyzed filter with
thermal PM oxidation kinetics.
Calibration was accomplished utilizing a series of optimization routines
developed in MATLAB. These routines tuned system parameters until simulation values
matched experimental results. GT-P models were coupled to MATLAB using a
communications block within Simulink. Simulation values were passed from GT-P to
Simulink while tuning parameters in GT-P were adjusted in MATLAB. In total, 17
parameters in the TWC and 6 parameters in the GPF were adjusted. Calibrated models
show reasonable agreement in terms of species conversion efficiencies, filtration
efficiency, and pressure drop. Details regarding the data analysis, model development,
communications coupling, optimization routines, and results are presented herein.
iii
In dedication to those who have become family
iv
Acknowledgments
I would like to thank Fiat Chrysler Automobiles for their generous support, Dr. Midlam-
Mohler and the rest of SIMCenter for their guidance, and my Alma Mater, the University
of North Dakota, for providing the foundation of knowledge for which I begin my career.
v
Vita
March 1992 ....................................................Born- Madison Wisconsin
2010-2014 ......................................................B.S. Mechanical Engineering, University of
North Dakota (UND)
2014 to present ..............................................Graduate Research Associate, Department
of Mechanical and Aerospace Engineering,
The Ohio State University
Fields of Study
Major Field: Mechanical Engineering
vi
Table of Contents
Abstract ................................................................................................................................ ii
Acknowledgments................................................................................................................v
Vita...................................................................................................................................... vi
List of Tables .......................................................................................................................x
List of Figures ..................................................................................................................... xi
Chapter 1: Introduction ...................................................................................................... 1
Chapter 2: Literature Review .............................................................................................. 5
Beginnings of Emission Regulation ................................................................................ 6
Vehicle Emissions and Regulation.................................................................................. 6
Vehicle Emissions Production ...................................................................................... 10
Aftertreatment Devices ................................................................................................. 12
Catalytic Converters ...................................................................................................... 13
Catalytic Converter Operation Criteria/ Failure ........................................................ 14
Substrate Technology ................................................................................................ 15
2-Way Catalysts ......................................................................................................... 16
vii
3-Way Catalysts ......................................................................................................... 17
Lean 𝑁𝑂𝑥 Traps ........................................................................................................ 18
Urea Based SCR ........................................................................................................ 19
Passive SCR ............................................................................................................... 20
Particulate Filters........................................................................................................... 20
Particulate Filter Materials ........................................................................................ 21
Particulate Filter Operation and Performance ........................................................... 23
Aftertreatment System Modelling ................................................................................. 27
Particulate Filter Modelling ....................................................................................... 27
Catalytic Converter Kinetic Modelling ..................................................................... 30
Chapter 3: Tools and Methods .......................................................................................... 33
Experimental Data collection ........................................................................................ 33
GT-Power Aftertreatment Models ................................................................................ 38
MATLAB ...................................................................................................................... 44
Chapter 4: TWC Model Development and Calibration .................................................... 48
Data Analysis ................................................................................................................ 49
Data Analysis: Engine Map........................................................................................... 50
Fitting FTIR Measured Emissions ............................................................................. 50
Catalytic Converter Conversion Efficiencies ................................................................ 55
viii
Oxygen Storage Capacity .............................................................................................. 56
Calibration of Kinetic Values........................................................................................ 57
Chapter 5: GPF Model Development and Calibration...................................................... 61
Data analysis ................................................................................................................. 63
Simulation and Optimization ........................................................................................ 68
Chapter 6: Results and Discussion.................................................................................... 71
TWC Data Analysis ...................................................................................................... 71
TWC Calibration ........................................................................................................... 76
GPF Calibration............................................................................................................. 81
Chapter 7: Conclusions and Future Work......................................................................... 86
References ......................................................................................................................... 89
ix
List of Tables
Table 1: TWC Test setup .................................................................................................. 35
Table 2: GPF Test Setup ................................................................................................... 38
Table 3: GT-Power TWC Model Parameters ................................................................... 48
Table 4: Optimizer Initial Conditions and Bounds ........................................................... 59
Table 5: GPF Parameters .................................................................................................. 62
Table 6: GPF Tuning Parameters...................................................................................... 63
Table 7: TWC Optimization Results................................................................................. 78
Table 8: GPF Optimization Parameter Results ................................................................. 83
x
List of Figures
Figure 1: List of Emissions Standards [3]........................................................................... 7
Figure 2: Emissions Tables [3] ........................................................................................... 7
Figure 3: New European Drive Cycle [3] ........................................................................... 8
Figure 4: Particulate Filter Flow [22] ............................................................................... 24
Figure 5: PM loading in Filter Substrate [23] ................................................................... 24
Figure 6: Active Regeneration of Particulate Filters [26]................................................. 26
Figure 7: Single Channel Model Proposed by Bisset [38]................................................ 28
Figure 8: 2-D aftertreatment Model [40] .......................................................................... 29
Figure 9: Aftertreatment System Used for TWC Testing ................................................. 34
Figure 10: GPF Testing Experimental Setup .................................................................... 36
Figure 11: GT-Power Model of TWC .............................................................................. 39
Figure 12: TWC Graphical User Interface........................................................................ 40
Figure 13: TWC Kinetic Reaction Set[42] ....................................................................... 41
Figure 14: DPF GT-Power Model .................................................................................... 42
Figure 15: GT-Power S-Block GUI .................................................................................. 43
Figure 16: GT-P Case Setup, Reading Excel Files ........................................................... 44
Figure 17: Example of Comparison Between Simulation and Experimental Values ....... 46
Figure 18: Lambda Signal over Filter Loading Cycle ...................................................... 47
Figure 19: Engine-out NO vs. Lambda ............................................................................. 51

xi
Figure 20: Inlet and Outlet 𝐶𝐻4 ....................................................................................... 52
Figure 21: Engine-out Oxygen Concentration .................................................................. 53
Figure 22: Estimated Hydrogen Concentrations ............................................................... 54
Figure 23: Exhaust Gas Temperature Fitting .................................................................... 55
Figure 24: Catalyst Conversion Efficiency Curves .......................................................... 56
Figure 25: Pre and Post Lambda Signals for Oxygen Storage Simulations ..................... 57
Figure 26: GPF Loading Cycle, Lambda Signal............................................................... 65
Figure 27: GPF Loading Cycle, GPF Inlet Temperature .................................................. 66
Figure 28: GPF Loading Cycle, Differential Pressure Signals ......................................... 67
Figure 29: GPF Loading Cycle, Measured Filtration Efficiency...................................... 68
Figure 30: Calculated Engine-Out Species ....................................................................... 72
Figure 31: Engine-Out Species Map From Literature[15]................................................ 73
Figure 33: Calculated TWC Species Conversion Efficiencies ......................................... 74
Figure 34: TWC Conversion Efficiency Values from literature [15] ............................... 75
Figure 35: Final Optimization Curves .............................................................................. 79
Figure 36: Simulated Conversion Efficiencies Utilizing Full Reaction Set from literature
........................................................................................................................................... 81
Figure 37: Experimental vs simulation pressure drop ...................................................... 82
Figure 38: Experimental and Simulated GPF Collection Efficacies ................................ 83
xii
Chapter 1: Introduction
Stringent emission standards, diagnostic requirements, and growing customers’
consideration of fuel economy continually forces the automotive industry to improve
performance and implement novel new technologies. One such instance is the
implementation of Gasoline Direct Injected (GDI) engines in consumer automobiles. GDI
engines are being used to increase powertrain efficiency and reduce vehicle fuel
consumption. However, a side-effect of the operation of GDI engines is increased
Particulate Matter (PM) emissions as compared to engines utilizing Port Fuel Injection
(PFI). At current levels, unmitigated PM emissions may be above current regulatory
levels.
The primary governing organizations for emissions regulation include the
European Union and their European Emission Standards (Euro 6), Unites States
Environmental Protection Agency (US EPA), and California Air Resources Board
(CARB). These organizations define the acceptable limits for various exhaust emissions
species. Historically regulated species for gasoline fueled consumer vehicles have
included total hydrocarbons (THC), carbon monoxide (CO), and oxides of nitrogen
(𝑁𝑂𝑥 ). According to the latest emissions standards, the allowable levels of species
regulated in the past will be reduced and new limitations on toxic pollutants including
PM will be implemented on gasoline vehicles. To accomplish the new regulatory
1
requirements manufacturers must optimize the design of existing technologies and may
be required to utilize additional technologies that are new to gasoline applications.
Three-way catalytic converters (TWC) are the primary exhaust aftertreatment device used
in stoichiometric gasoline vehicles that are sold today. TWC oxidizes carbon monoxide
(CO) and unburned hydrocarbons to CO 2 and H2 O, and reduce oxides nitrogen (NOx).
However with new regulations on PM, the use of a Gasoline Particulate Filter (GPF) may
be required to remove PM from the exhaust gas. GPFs are an iteration of Diesel
Particulate Filters (DPF), which have been used for the past two decades in commercial
Diesel applications. The filters trap carbon based PM by flowing exhaust gas through a
porous ceramic membrane. Under rich or low temperature exhaust conditions, PM slowly
accumulates in the filter until such time as the particulate trap is filled. Onboard controls
are then required to sense or estimate the filter state and request the powertrain system to
start a regeneration strategy, emptying the filter though oxidation of the trapped PM.
In addition to designing control algorithms, manufactures must also insure that exhaust
aftertreatment devices function throughout the useful vehicle life. Manufactures must
comply with On Board Diagnostic (OBD) regulations by creating algorithms to detect
possible failures in the exhaust gas aftertreatment system. The designed algorithms utilize
signals from various sensors in the exhaust system by comparing measured values to
those which would be expected under normal conditions.
Diagnostic and control algorithms could feasibly be developed utilizing an
experimental engine test cell. However, the operation of experimental equipment is often
cost prohibitive in terms of both time and money. As an alternative option, algorithms
2
can and often are developed utilizing calibrated models of the engine and exhaust
aftertreatment systems. Aftertreatment components including catalysts, particulate traps,
sensors, and fluid pipes are modelled. Once calibrated, the models are able to capture a
majority of operation phenomena and simulations of a variety of operating conditions and
faults can be accomplished. Simulation results are then used both to develop the basis of
OBD and control algorithms.
Consequently, the development of calibrated models is of paramount importance
to developing actuate control and fault diagnostics. In this work, 1-Dimenstional models
for both TWC and GPF aftertreatment devices were developed using the commercially
available powertrain modelling software, GT-Power. Equations which govern the system
dynamics including heat transfer, fluid flow, and chemical kinetics are built into the
software and are based both on published literature and proprietary models. Users can
select from a variety of modelling options ranging from 0-D models to full 3-D models
based on imported CAD files. Desired model options are selected using an intuitive
Graphical User interface (GUI). Models can then be calibrated by tuning system
parameters that are either unknown or in which have a large uncertainty.
Calibrating the kinetic reaction set in the TWC model and the pressure drop and
PM collection efficiency in the GPF model was the primary objective of this work.
Typically industry experts utilize their detailed understanding of the physical systems to
calibrate models by hand. Even for the experienced professional, this process can be time
consuming and tedious. Utilizing optimization routines in MATLAB, the calibration
process was automated in effort to remove the need for hand-tuning. The developed GT-
3
Power models were connected and co-simulated with MATLAB /Simulink. The process
of connecting the two software packages allowed for automated tuning of 17 TWC
kinetic parameters and 6 unknown GPF physical properties.
4
Chapter 2: Literature Review
Prior to the commencement of data analysis, sensor algorithm development, and
system modeling, a literature review was conducted. Given that the scope of the project
was to model, calibrate, and create diagnostics for the aftertreatment system of a GDI
vehicle, the range of reviewed topics was broad. This chapter, while not complete with
every piece of information that could be possibly listed, intends to provide the reader
with an introduction to topics that are relevant to the conducted work while providing
references so the curious reader may learn more if they feel inclined.
Starting at the motivation for the conducted work, an introduction to the
regulations placed on vehicle emissions is presented along with a brief history of how the
regulations were first established. From here, the emissions themselves are discussed,
answering the questions of “why are these species regulated?” and “How are pollutants
made by a vehicle?” However, there are devices and sensors which can be placed in the
exhaust stream of a vehicle to reduce vehicle emissions. These devices and their use are
known as the exhaust aftertreatment. There are several types of aftertreatment devices for
gasoline vehicles including catalytic converters, particulate filters, and LNT’s. The
operation and methods to model of each of which are discussed within this review. Also
discussed are the sensors which are commonly used to monitor the state of the exhaust
aftertreatment system. These include oxygen, particulate matter, temperature, and
pressure sensors. The operation, signal interpretation, and modeling of each is discussed.
5
Beginnings of Emission Regulation
Following the Second World War, the population of the state of California was
increasing [1]. With the increased population density came problems; In the Las Angeles
valley region a thick fog of chemicals known as smog started forming at frequent
intervals. Recognizing the potential health risks, through the 1940’s, 50’s, and 60’s, the
state of California formed committees and panels tasked with finding a solution to the
smog issue. These committees determined that one of the major contributors was
automobiles and their tailpipe emissions. Soon several small regulatory boards were
made across California and the beginnings of automotive emissions regulations were
culminated. In 1967 the regulatory bodies were unified to form The California Air
Resource Board (CARB). CARB passed air quality standards in 1969 for several of the
species found in automotive exhaust [2]. In 1970 The United States Congress followed
California’s example and signed The Clean Air act of 1970 which regulated automotive
tailpipe emissions across the country.
Vehicle Emissions and Regulation
Several organizations regulate automotive emissions species in various countries
throughout the world. The primary regulators include the United States Environmental
Protection Agency (EPA), the California air Resource Board (CARB), and the European
Union [3]. Each organization specifies allowable levels of vehicle emissions based on
6
vehicle type. Shown in Figure 1 is a list of the current and recent regulations standards.
Within each standard, Figure 2, the cumulative tailpipe emissions for each regulated
species is specified based on a mandated drive cycle, Figure 3. The vehicle is subjected
to a speed-time trace that is intended to imitate the driving habits of a typical consumer.
Tested automobiles must meet or exceed regulated levels when the vehicle is new and
maintain the high standard throughout the duration of the vehicle life.
Figure 1: List of Emissions Standards [3]
Figure 2: Emissions Tables [3]
7
Figure 3: New European Drive Cycle [3]
Selected exhaust species are regulated in effort to reduce levels of species which
participate in atmospheric reactions, contribute to global warming, or present potential
health effects to human and animals. The primary regulated exhaust species include
oxides of nitrogen (𝑁𝑂𝑥 ), carbon monoxide (𝐶𝑂) and hydrocarbons. Other species,
which are regulated on select vehicles or in which regulation has been proposed, include:
carbonaceous particulate matter (PM), Ammonia (𝑁𝐻3 ), and Nitrous Oxide (𝑁2 𝑂) [4, 5].
𝑵𝑶𝒙 : Oxides of nitrogen is the combination of nitrogen dioxide (𝑁𝑂2 ), a reddish
brown gas, and nitric oxide (𝑁𝑂), a clear gas. Both species are toxic and actively
participate in the atmospheric reactions, causing smog and acid rain.
𝑪𝑶: Carbon monoxide is a colorless, odorless, and tasteless gas which is toxic to
humans and animals in significant concentrations. In addition to the potential
adverse health effects, CO also plays a part in the formation of atmospheric smog.
8
Hydrocarbons : There are numerous hydrocarbons emitted by internal
combustion engines. To simplify regulations and to ensure compliance,
hydrocarbon limits are presented as a single values that represents the summation
of all species. Depending on the regulation hydrocarbons are defined in terms of
THC or NMOG. Total Hydrocarbons (THC) is the concentration of the both the
oxygenated and non-oxygenated gaseous hydrocarbon species in the exhaust
while Non-Methane Organic Gas (NMOG) is equivalent to THC minus the
portion of methane.
PM: Particulate matter, colloquially referred to as soot, are aerosols with mean
diameters ranging from 60 to 100 nm [6]. Primarily emitted by diesel and gasoline
direct injected vehicles, PM is comprised of complex carbon chains imbedded
with soluble organics fractions (SOF) of volatile species. A smaller but still
significant portion of PM, 0.5 to 1% [7], is what is referred to as ash. Ash is
composed of metal oxides, sulfates and Phosphates, but unlike the carbon portion
of PM, cannot be easily oxidized [8, 9, 10,11]. The small particle size of PM
makes it a primary health concerns since inhaled PM can have negative effects on
the repertory system of humans [10].
𝑵𝟐 𝑶: Nitrous oxide is a green-house gas which has significant potential to effect
global warming. The potential greenhouse gas effects of 𝑁2 𝑂 is 298 times that of
𝐶𝑂2 [12].
𝑵𝑯𝟑: Ammonia is a colorless gas which acts as a strong oxidizer in atmospheric
reactions.
9
Particulate matter proposes and new challenge to automakers. Limits have been
applied to allowable PM levels of Diesel Vehicle since the implementation of the Euro 1
standard, however regulation for PM on gasoline vehicles is new as of the Euro 5
standard. In the past a majority of gasoline vehicles have utilized Port Fuel Injection
(PFI) technology; PFI vehicles emit near zero levels of PM. With increased consumer
demand for efficient powertrains, automakers have started utilizing Gasoline Direct
Injection engines in their vehicles. This technology has proven beneficial in terms of
powertrain efficacy, however has also been show to emit significant levels of PM. To
mitigate the increased PM levels, PM is now a regulated species of GDI vehicles, starting
with the Euro 5 Emissions and LEV II standards [13,14]. Through the use of either in-
cylinder methods or aftertreatment devices, Euro 6 standards specify that tailpipe PM
levels must be at or less than 4.5mg/km when the tested vehicle is driven over an NEDC
drive cycle [3, 6].
Vehicle Emissions Production
Internal combustion engines are, at their essence, energy converters. Combing
fuel and air, chemical reactions occur allowing for the release the fuel’s energy into the
systems working fluids. Then, through the use of an engines mechanical mechanism,
whether it be a turbine, rotor, or reciprocating piston, energy from the working fluid can
be extracted into useable work. Regardless if an engine is run on gasoline, diesel, natural
gas, or any other hydrocarbon, the chemical reactions which allow for the engines
operation will produce byproducts. Some byproducts, such as carbon dioxide (𝐶𝑂2 ) are
10
relatively harmless, while others like particulate matter, can cause significant health
concerns.
The simplest form of combustion is the ideal case were an oxidizer and fuel are
reacted in the perfect ratio (stoichiometry) to form the expected products. In this perfect
case the reaction has enough time to reach its equilibrium and trace species do not
participate. The equation for the oxidation of a hydrocarbon in air is shown by Eq.1 [15].
𝑏 𝑏 𝑏
𝐶𝑎 𝐻𝑏 + (𝑎 + 4 ) (𝑂2 + 3.773𝑁2 ) = 𝑎𝐶𝑂2 + 2 𝐻2 𝑂 + 3.773 (𝑎 + 4 ) 𝑁2 Eq. 1
As vehicle operation is highly transient, maintaining perfect stoichiometry in all
the cylinders is difficult and often cannot be accomplished. Departures from
stoichiometry lead to increased emissions of regulated species. During rich operation
there is deficient oxygen for the reactants, causing portions of the fuel to not react or
react partially to form carbon monoxide, hydrogen, and ammonia. Under lean conditions
there is extra oxygen in the fuel charge. Due to the high in-cylinder temperatures,
nitrogen, which is non-reactive at room temperatures, actively participates in reactions
forming 𝑁𝑂𝑥 and trace quantities of 𝑁2 𝑂.
Additional emissions can be caused by possible inhomogeneity in the combustion
chamber caused by improper mixing or as an artifact of direct injection. Fuel is sprayed
directly into the cylinder at high pressures to form a cloud of small droplets. Droplets
vaporize and are mixed with the incoming air charge to form the combustible charge.
However, a small portion of fuel droplets are unable to vaporize prior to combustion,
11
forming local pockets of rich mixture. Within the localized rich pockets, PM forms due to
lack of oxygen and high temperature [14].
In addition to emissions caused by the fuel-air ratio, a smaller but significant
portion of emissions is an artifact of the operating characteristics of reciprocating piston
engines. Depending on the method and timing of fuel injection, a portion of the fuel can
adhere to the cylinder wall. While the piston moves upward during the compression
stroke, fuel that is wetted to the cylinder wall can become trapped between the piston and
cylinder wall. Once released, the portion of trapped fuel, combined with residual oil from
the cylinder wall, is either partially burned forming PM or released into the exhaust
stream.
Emissions formation can be partially mitigated by optimizing various controllable
engine parameters. Exhaust Gas Recirculation (EGR) has been utilized in gasoline and
Diesel applications to reduce in-cylinder temperatures, significantly reducing 𝑁𝑂𝑥
formation. While for GDI engines, one study showed that a significant improvement in
PM emissions could be realized by adjusting fuel rail pressure, spark timing, and engine
coolant temperature [16]; increased engine body temperatures were found to reduce PM
levels. Similarly, a study by Environment Canada showed that decreased ambient
temperatures correlated to increased emissions [17].
Aftertreatment Devices
The efforts of in-cylinder emissions reduction can only go so far, and at some
point, aftertreatment devices must be used. Aftertreatment devices are used to mitigate
12
harmful exhaust species by taking the species leaving the engine and converting them to
less harmful byproducts by the time they reach the tailpipe. For stoichiometric gasoline
applications, Three Way Catalytic Converters (TWC) are primary aftertreatment device
used, however with new government regulation, the use of a Gasoline Particulate Filter
(GPF) may be required. Within this section, the operation of various types of catalytic
converters and particulate filters is discussed.
Catalytic Converters
As their name implies, the functionality of catalytic converters is enabled by the
use of a catalyst. A catalyst is a substance that increases the rate of chemical reactions
without being permanently consumed. Precious metals groupings (PGM) of platinum,
palladium, and rubidium are primarily used in automotive applications. These catalysts
have proven their ability to convert harmful exhaust species of interest to species such as
carbon dioxide (𝐶𝑂2 ), water (𝐻2 𝑂), and nitrogen (𝑁2 ).
Catalytic materials are thinly dispersed in a layer called a washcoat. Serving as a
binder and providing a rough surface that provides a large surface area for reactions, the
washcoat helps maximize the potential of the precious metals. This is critical for
production applications so that the total PGM loading can be reduced to minimize costs.
Additionally, in certain types of catalysts, the washcoat can be impregnated with
substances that can trap and store exhaust species, and then release them at a different
stage of operation.
13
The washcoat is applied to a support structure that is referred to a substrate. The
first generation of catalysts were packed bed reactors utilizing used small beads to fulfill
this purpose. However, this method proved inefficient at utilizing the catalysts. Ceramic
honeycomb structures were then developed and were shown to perform better in
automotive applications. Bead type converters were phased out and today honeycomb
monolith are the preferred choice for automotive applications. Utilizing the basic
honeycomb concept, physical properties have been modified and continued development
has occurred in the field of substrates. Special attention has been given to reducing
backpressure and decreasing the thermal mass to reduce the time need to achieve the
light-off temperature.
Catalytic Converter Operation Criteria/ Failure
Catalysts are sensitive to extreme temperatures, both high and low. The lower extreme
temperature is referred to as the light-off temperature. When the temperature catalyst is
below this value, or in more general sense, if the substrate temperature is below the value,
the catalyst does not assist in reactions. This occurs if exhaust temperatures are too low
and during cold start situations. During which periods, harmful exhaust species are
allowed to slip by the Catalytic converter and out the tailpipe. The other extreme of high
temperature operation sensitivity is more catastrophic. If the substrate temperature
achieves values above 900 C for extended durations, the thinly dispersed PGM loading
will start to sinter; an irreversible process where finely dispersed PGM metals form
globules and reduce the available catalytic surface area available for reactions. If a
14
catalyst is allowed to operate under sintering conditions for extend durations,
performance of a catalyst will be significantly reduced.
Lead and sulfur poisoning are an additional concern for all types of catalytic
converters. Both lead and sulfur form chemically stable compounds with the PGM
materials, preventing catalytic activity. Lead poisoning is irreversible, while sulfur
poisoning can be reversed at a cost. By operating sulfur poisoned catalyst in a rich
exhaust stream at elevated temperatures, the sulfur can be removed, however these
conditions can lead to sintering failure. To avoid these situations, the use of unleaded
fuels with low sulfur content is desired.
Substrate Technology
The original catalytic converters that appeared on U.S. automobiles utilized
pellets as the surface for catalytic reactions. This technology was prone to high PGM
losses and was replaced by honeycomb structures. Originally developed by Corning, the
honeycomb structure provides a large surface area for reactions which allow for a better
conversion efficacies.
The light-off temperature of a catalyst is the minimal temperature that is required
for the catalyst to effectively participate in reactions. Traditionally the increased
temperature is achieved by contact with the heated exhaust gasses. However, the process
of heating can take an extended period of time, during which, the catalyst is ineffective.
To reduce the period of time it takes the catalyst heat up, several options have been
developed and are implemented on production vehicles.
15
One strategy for reducing catalyst light off time is to minimize the specific heat of
the substrate, reducing the amount of energy that must be transferred to the structure to
increase its temperature. Another strategy is to increase the rate at which energy is
transferred to the structure. By increasing the monolith cell density, the surface area
available to for heat transfer is increased.
Heated catalysts are one active technology for light-off time reduction. The
catalyst is heated before any exhaust species reach it and thus eliminating the time to
light-off temperature. These systems however take time to heat up and consumers are
unlikely to be happy if they have to wait to start their cars. Hybrid vehicles do present a
solution to this however since the driver can start vehicle operation using the electric
powertrain. The engine start can then be delayed until after the catalyst heats up.
While thermal and conversion considerations are important from an emissions
standpoint, catalyst are also known for negatively effecting engine performance by
increasing backpressure in the exhaust [18]. When new substrate designs are considered,
the conversion efficacy and backpressure are often competing parameters and a
compromise between the two must be achieved.
2-Way Catalysts
Two way catalysts, also referred to as oxidation catalyst, were the first catalyst
systems developed for automotive applications. These catalyst utilize Pt/Pd oxidation
washcoats. They are referred to 2-way catalyst for their ability to oxidize two primary
emission species, THC and CO. Conversion efficacies are greatest when operated in lean
16
exhaust. While still used in parts of the world for small engine application such as
motorcycles, 2 way catalysts are no longer used for stoichiometric gasoline automobiles.
The technology has however been utilized as part of some diesel particulate filter (DPF)
systems.
3-Way Catalysts
Three-way-catalysts (TWC) are most effective when used on engines operating
under stoichiometric conditions. These catalysts utilize a combination of platinum,
palladium, and rubidium to convert harmful exhaust gasses to species such as 𝑁2 , 𝐶𝑂2 ,
and 𝐻2 𝑂. The platinum and palladium are utilized for oxidation of THC and CO while
the Rubidium is used for reduction of 𝑁𝑂𝑥 . The conversion efficiencies of for both the
oxidation and reduction components are high near stoichiometric conditions however;
𝑁𝑂𝑥 conversion efficiency during lean operation and CO/THC conversion efficiency
during rich operations is poor. Therefore, a high degree of engine fuel mixture control is
required to keep the exhaust gas composition constant. However, even with good fueling
control transients in engine load and speed tend to cause small perturbations from
stoichiometric conditions.
A key advancement that assists a TWC in keeping stoichiometric conditions at the
reaction sites is the addition of ceria to the washcoat. Ceria has the ability to capture and
store oxygen when the exhaust is lean then release the stored oxygen during rich
conditions. This allows for brief transients in the catalyst inlet gas while keeping the
effective stoichiometric concentration at the catalyst surface.
17
The oxygen storage is also used with a fuel control strategy known as AFR
switching. Additional efficiency can be realized by allowing the exhaust stream entering
the catalyst to oscillate between lean and rich conditions.
Lean 𝑁𝑂𝑥 Traps
During lean engine operation traditional TWC have high conversion efficiencies
for THC and CO but very poor conversion efficacies for 𝑁𝑂𝑥 . This presents a problem
for engines that run lean mixtures since low 𝑁𝑂𝑥 conversion efficacies can cause non-
compliance issues regarding 𝑁𝑂𝑥 emissions regulation. Lean 𝑁𝑂𝑥 traps, commonly
known as LNTs, are one of the proven solutions to this dilemma. When coupled with an
upstream oxidation catalyst, moderately large conversion efficiencies can be achieved.
The engine is controlled to switch between running lean and rich. 𝑁𝑂𝑥 is absorbed by the
LNT during lean burn operation and rapidly desorbed during rich operations. During the
rich periods THC and CO conversion rates rapidly diminish over the oxidation catalyst
making those substances available for reactions on the LNT. The available CO and THC
in the exhaust stream in the vicinity then reacts with the stored NOx to create less
harmful species.
While effective and used in many applications, LNTs are by no means the perfect
solution for all lean applications. High cost, low efficiency, and thermal instability are
commonly mentioned in literature as technological issues. In the current stage of
development LNTs require very large precious metal loads to operate effectively which
makes their implementation expensive. In an industry where per vehicle profits are low,
18
the additional cost will likely be passed onto the consumer, which would increase vehicle
prices.
While effective for use in both light duty diesel and GDI engines where the switch
between lean and rich mixtures can be easily controlled, Future technologies such as
stratified GDI and HCCI engines may have difficulties rapidly switch between operation
modes.
Urea Based SCR
Selective catalyst reduction (SCR) has become a common technology for NOx
reduction in both large commercial and consumer vehicles. SCR is typically used on
Diesel vehicles but could be used with other engine operations with lean exhaust species.
In this type of catalyst, NOx emissions are reduced over the catalyst bed with ammonia
(NH3) [19]. The ammonia originates from a urea solution that is injected upstream of the
SCR. Commercially this solution is sold as AdBlue, which is 67.5% water and 32.5 %
urea. The urea is then thermally decomposed to 𝑁𝐻3 in transit to the SCR. Once in
proximity to the SCR, the NH3 is adsorbed and reacted with NOx. Given that the NH3
storage capacity is finite, attention is given to urea injection control to insure minimal
NH3 slip.
The overall conversion efficiency of SCR systems is very good, which has helped
the technology gain favor in the diesel community. However, the system consumes urea
and requires an additional storage tank for the urea solution. In addition to the difficulty
of getting individuals to remember to fill their urea tanks is the inability to find urea
19
supplies. A large infrastructure has not been built to accommodate individuals. These
problems would be even lager if a urea based SCR system was used on a gasoline engine.
The high in cylinder temperatures of gasoline engines lead to greater NOx production,
which in turn, would require a greater consumption of Urea.
Passive SCR
While not fully developed, a passive SCR system has been proposed that would
eliminate the need for urea injection [20]. In this system NH3 would instead be created
on an upstream TWC during brief rich operation periods. During these periods the altered
precious metal catalyst on the TWC would create NH3 from reactions with NOx and H2.
The NH3 would then flow downstream where it would be adsorbed by the SCR and then
later used to reduce NOx.
Different manufactures have suggested slightly different layouts and different
names for this technology but they all operate under similar premises. If development
proceeds, this technology could partially replace current LNT catalyst. This system is
advantageous over an LNT because it requires less PGM which would allow manufacture
to produce lower cost units.
Particulate Filters
Particulate filters are a type of emissions control device which have been used on diesel
vehicles beginning in the 1980’s and show promise for use of GDI vehicles in the near
future. When equipped to a gasoline vehicles, the particulate filter is referred to as a
20
Gasoline Particulate Filter (GPF), while when a filter is equipped to a diesel vehicle, it is
referred to as a Diesel Particulate Filter (DPF). In either application the emissions
control device is used to mitigate tailpipe particulate matter emissions by filtering the
engine-out exhaust stream. Trapped particulate matter is continuously trapped until such
point when the filter backpressure is significantly increased from material loading. At
such time, the filter must be regenerated, a process of emptying the filter by oxidizing the
trapped carbonaceous particulate matter. Details on the materials used for particulate
filters, operation, and filter regeneration are discussed in detail below.
Particulate Filter Materials
Particulate filters used in automotive applications consist of a steel canning which
contains a filtration substrate and a matting material separating the substrate from the
canning. In some diesel applications metallic foils or foams are used, while in gasoline
applications and the majority of production light duty diesel applications, porous ceramic
substrates are used as the filtration substrate. Most published papers reference cordierite
and silicon carbide as the primary material choices in diesel applications while cordierite
is used almost exclusively in gasoline applications. One paper by Hyundai Kia describes
the benefits and drawbacks to potential substrate materials [21]; Cordierite, aluminum
titanate and silicon carbide where evaluated for their use in particulate filters.
Silicon carbide was found to have higher thermal conductivity and higher thermal
mass compared to the other evaluated materials. These properties make controlling
substrate temperature during regeneration easier. It has been however found that the soot
21
cake layer tends to split away from the silicon carbide filter surface when the engine is
turned off and the filter cools. This is due to the large thermal gradient and can cause an
increase in filter backpressure in the proceeding cycles of operation. This presents a
problem in terms of control because the pressure difference across the filter is assumed to
be a linear relationship, and that relationship is currently used in production control
strategies which use differential pressure sensors.
Aluminum Titanate was found to have a high thermal durability due to a
structural expansion joint in the material that mitigates thermal expansion. However these
properties were found to be negated when a washcoat was applied, the different thermal
expansion between raw material and coating can lead to overall thermal expansion of the
filter.
Cordierite was found to be more conducive to catalytic washcoats and was found
to have a lower thermal mass as compared to other material options. These properties
allow for a reduction in light-off time; the time required for the filter substrate to reach
temperatures concussive to species conversion. These characteristics enabled better
conversion rates of CO and HC during cold start testing. However, due to lower thermal
mass and lower conductivity, temperature control during active regeneration presents a
challenge. If an excessive amount of soot is oxidized in a short duration, the filter
monolith could be damaged by substrate-ash interaction above 1100 deg. Celsius.
22
Particulate Filter Operation and Performance
Particulate filters are one of the potential devices that can be used to mitigate
exhaust species. Primarily, particulate filters are used to trap carbon based particulate
matter from the exhaust stream, but can also be utilized to covert gaseous exhaust species
when equipped with a catalytic washcoat.
Filtration is accomplished by flowing exhaust gas through a porous ceramic
membrane. As exhaust flows into a particulate filter, it flows through a number of
channels that run the length of the filter, Figure 4. At the end of each channel is a plug
that does not allow exhaust to flow though. Instead, the exhaust must pass through the
porous ceramic walls of the channel before reaching the plug at the end. By the process
of flowing exhaust through the walls, a certain percentage of the particulate matter is left
behind in the wall. Assuming the trapped particulate matter is not oxidized, the wall will
eventually saturate with PM. At this point the PM is trapped at the surface, building a thin
layer that is referred to as the soot cake layer, Figure 5.
23
Figure 4: Particulate Filter Flow [22]
Figure 5: PM loading in Filter Substrate [23]
As the filter is loaded with particulate matter, the pressure drop will slowly
increase starting from the clean wall pressure drop; the pressure drop across the filter
when there is no trapped particulate matter. At this point, pressure quickly and non-
linearly increases until the porous wall is saturated. Once in the soot cake loading
24
criterion, pressure drop will increase in a linear manor in relationship to trapped PM. This
characteristic has been utilized in diesel regeneration and diagnostics control strategies.
Based on the differential pressure across a DPF, the mass of trapped PM can be
estimated. Once the estimated mass is at an appropriate level, the system will regenerate
the filter through either passive or active means.
Particulate filters must be periodically emptied of collected PM though a process
known as regeneration. PM that has been collected in the filter is oxidized by either 𝑁𝑂2
or 𝑂2 [24, 25]. In diesel applications the exhaust temperature must be raised above the
nominal EGT for the reactions to proceed. One method to induce regeneration is to
increase the filter temperature by burning additional hydrocarbons over a Diesel
oxidation catalyst located in front of the particulate filter or with a catalytic washcoat
applied to the filter, Figure 6 [26]. Fuel is sprayed into the exhaust stream ahead of the
oxidation catalyst. Once in the vicinity of the Catalyst, the vaporized fuel oxidizes over
the catalyst, resulting in energy release that raises the exhaust temperature entering the
DPF to the required regeneration temperature. In gasoline applications, the use of fuel
injectors has been found unnecessary due to the higher exhaust gas temperatures, as
compared to diesel systems.
25
Figure 6: Active Regeneration of Particulate Filters [26]
Potential Gasoline Particulate Filter (GPF) designs are based on DPF technology,
however are optimized to the unique operating conditions and requirements of gasoline
applications. In addition to PM filtration efficiency, of primary concern is potential losses
in performance and efficiency caused by increased backpressure. Hyundai Kia Motors
reported that by optimizing geometric parameters of a particulate filter, a 25% reduction
in backpressure could be achieved [27]. In another article by Hyundai Kia Motors, the
filtration performance was evaluated by fitting a GPF device to a production GDI vehicle.
Driven over US06 and FTP drive cycles, the filter was determined to significantly reduce
PM emission. Similar results have also been achieved in several other studies, finding
that particulate filters are a viable option for gasoline applications [28, 29, 30, 31].
Concerns over efficiency loss have been tested by and increased fuel consumption caused
by fitted GPF devices was determined to be minimal over tested conditions [32].
26
Aftertreatment System Modelling
An exhaust aftertreatment system can be mathematically modelled using a series
of coupled differential equations. Aspects in which can modelled include but are not
limited to fluid flow, wave dynamics, heat transfer, material stresses and chemical
kinetics. Equations can be written by hand and coded using programming languages like
C++, FORTRAN, or MATLAB. Several papers have been published by researchers who
have gone through the process of writing specialized codes [21,33,34,35,36,37]. However
it is not required for most users to write their own code because flexible development
software have been written that allow users to easily model complex automotive systems
using graphical user interfaces. AVL Boost, Gamma Technologies’ GT-Power, and
Axisuite are packages that are used among major automotive suppliers to model
automotive aftertreatment systems.
Particulate Filter Modelling
Several models have been proposed in the literature for Diesel Particulate filters.
One of the earlier works is that of Bisset in 1984 [38], where the mathematical
expressions regarding conservation of mass and energy were derived in order to model
thermal regeneration of particulate matter. By assuming that the flow entering the filter
was spatially uniform, it was proposed that the bulk filter characteristics could be
27
modelled by obtaining the relevant equations for a single channel, Figure 7. Key
simplifications made included assuming an adiabatic system and ideal gas properties.
Figure 7: Single Channel Model Proposed by Bisset [38]
Utilizing the conservation equations, later works added complexity by modelling
the filtration properties of the GPF devices. Using a spherical unit collector concept, a
packed bed model utilizing multiple unit collectors, can be employed to model the
filtration in filter substrate microstructure [37]. Changes in the filtration efficiency and
pressure drop can be expressed in relation to parameters of the PM loaded particulate
filter wall; particulate deposit thickness, density, porosity and permeability. Allowing for
velocity, particulate loading, and filtration efficiency distributions to be simulated more
acutely.
A key assumption of the prior models is that a single channel can accurately
capture the operating phenomenon of the entire filter. Utilizing this simplif ication,
28
reasonable results can be achieved and are presented by several authors [33,34,36,37]. If
required, model accuracy can be improved utilizing higher dimensional models. A 2-
Dimensional model was proposed by Jørgensen [39]. Unlike published 1-D models, inlet
flow distribution were not assumed to be uniform. To account for axially differences, it
was proposed that radial sections of the device contain flow uniformities, Figure 8.
Utilizing this model with AxiSuite, it was shown that simulation results show better
agreement to experimental results, as compared to a 1-D model [40]
Figure 8: 2-D aftertreatment Model [40]
An important phenomena of which modelling is required is the oxidation of
particulate matter through both thermal and catalytic mechanisms. In Diesel applications,
the primary reaction mechanisms for PM oxidation include O2 and NO2 [40]. The rate at
which each reaction occurs is calculated using an Arrhenius expression based on the mass
of PM, the activation energy of the trapped PM, and an associated pre-exponential factor
29
[37]. PM trapped in the soot cake layer and the PM trapped within the wall are typically
assigned different values. The pre-exponential factors are often not reported in literature,
while activation energies ranging from 100 to 160 KJ/mol have been published [24]. Fuel
characteristics, the Soluble Organic Fraction (SOF), and possible catalytic ash content of
PM are all believed to contribute to discrepancies in reported values [41].
Catalytic Converter Kinetic Modelling
A complex series of chemical reactions occur over the catalyzed surfaces found in
many exhaust aftertreatment components. The ability to accurately capture the
characteristics of these reactions is of great importance and can be accomplished either
by using map or physics based approaches. Utilizing map based models is the more
simplistic approach and is often done for use with controls based models. However, the
use of physics based chemical kinetic reaction sets is required to more accurately capture
operating phenomena.
Chemical kinetics is a field of study concerned with chemical reaction pathways
and the rates at which reactions occur. To model a system, a series of chemical pathways
can be defined along with associated rates for each reaction. The complexity of reactions
sets ranges from micro-kinetics sets that model hundreds of reaction pathways and aim to
capture all intermediate reaction steps, to macro-kinetic sets that lump quick intermediate
reactions into with rate limiting steps.
30
For use with aftertreatment modelling, macro-kinetics sets provide reasonable
accuracy when calibrated for the narrow operating range of operating conditions found in
the exhaust systems of automobiles. One such kinetic set was proposed by Ramanathan
and Sharma of General Motor’s Indian Research and Development Laboratory[42];15
kinetic reactions were utilized to model catalytic behavior on precious metal and oxygen
storage sites. Ramanathan and Sharma assumed that hydrocarbons could be modelled by
two species, propane and propylene. Propylene represents fast reacting hydrocarbons
while propane represents the slow reacting hydrocarbon. The rate at which each reactions
occurs is specified by an Arrhenius rate expression based on the catalyst temperature and
the concentration of reactants, Eq. 2. In addition to the concentration of reactants, and
addition term is included to account for the competition between species for available
oxygen. This term is referred to as an inhibition term and is based on Langmuir
Hinshelwood kinetics, Eq. 3.
𝐸𝑎 ,𝑖
𝐾𝑖 [𝑅𝑒𝑎𝑐𝑡𝑎𝑛𝑡𝑎][𝑅𝑒𝑎𝑐𝑡𝑎𝑛𝑡𝑏 ]exp⁡( − )
𝑅𝑇𝑆
𝑅𝑎𝑡𝑒 = Eq.2
𝐺
𝐸𝑎 ,1,𝐺 𝐸𝑎 ,2,𝐺
𝐺 = (1 + 𝑘1,𝐺 exp⁡(− 𝑇𝑠
)[𝐶𝑂] + 𝑘2,𝐺 exp (− 𝑇𝑠
) [𝐶3 𝐻6 )2 (1 +
𝐸𝑎 ,3,𝐺 𝐸𝑎 ,4,𝐺
𝑘3,𝐺 ⁡exp (− ) [𝐶𝑂]2 [𝐶3 𝐻6 ])(1 + 𝑘4,𝐺 exp (− ) [𝑁𝑂])⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡
𝑇𝑠 𝑇𝑠
Eq. 3
Once calibrated using a global optimizer and a transient data set, it was found that
the model could accurately the tailpipe-out species over an FTP drive cycle. That drive
31
cycle was chosen since the authors wanted to develop a model which could predict cold
start conditions. Additional work was then conducted by GM research laboratories and at
the University of Wisconsin to develop kinetic reaction sets, based on the work of
Ramanathan and Sharma, which could additional model the formation of key regulated
trace species [43,44].
32
Chapter 3: Tools and Methods
Simulation and modelling work was conducting at The Ohio State University’s
SIMCenter using commercially available tools. The powertrain modelling package GT-
Power was used to develop after treatment models of both a TWC and GPF. GT-Power
models were connected with Simulink models utilizing a communication block within
each software package. Model co-simulation allowed for the calibration of GT-Power
models through the use of optimization routines within MATLAB. The optimization
routines compared experimental data provided by Fiat Chrysler Automobiles (FCA), to
simulations values and adjusted system parameters to reduce differences in values.
Details regarding the experimental data, modelling methods, co-software
communications, and optimization methods are discussed within this chapter.
Experimental Data collection
Two sets of data were provided by FCA to assist with model development and
calibration. The first data set contained emissions measurements taken from a duel TWC
aftertreatment system while the second set of data consisted of a GPF being loaded with
PM under steady state rich conditions. Both sets of testing were conducted using engine
dynamometers with instrumented aftertreatment systems.
33
The provided TWC dataset included emissions measurements collected from one
bank of the aftertreatment system of a Chrysler Pentastar V6 PFI gasoline engine, Figure
9. The aftertreatment system consisted of 1 close coupled TWC with a high precious
metal loading and an underfloor TWC with a lower precious metal loading. The engine
was run with E0 and E10 fuel under steady-state conditions at a variety of engine
operating points. During operation, bench emissions measurements were taken at the
engine out and tailpipe locations while FTIR measurements were collected from a single
point in the exhaust stream; either the engine-out, between catalysts, or tailpipe location.
Figure 9: Aftertreatment System Used for TWC Testing
34
Table 1: TWC Test setup
Sensors Manufacture/Model # Location

UEGO ECM/ FA1000 1 Engine-out
FTIR 1 Engine-out, Between Catalysts, Tailpipe
Bench Emissions Analyzer 2 Engine-out, Tailpipe
Type K Thermocouples 12 Engine-out, Within Catalysts
Engine
3.6L PFI V6 Chrysler/ Pentastar
Fuel E0 and E10
35
Figure 10: GPF Testing Experimental Setup
GPF Testing utilized an instrumented MY2010 2.0L GM Ecotec DISI
turbocharged engine with an aftertreatment system containing a GPF, Figure 10. The
engine was run with winter grade 93 octane premium unleaded E10 at operating
conditions that induced large amounts of PM production while maintaining low exhaust
temperatures and flowrates. With either a Bare GPF, GPF with a catalytic washcoat, or
straight pipe, the engine was run over a cycle designed to load the particulate filter. The
36
cycle started by running lean conditions to clean any residual particulate matter from the
test filter, followed by an extensive loading duration at lambda= 0.9, Figure 18.
Several measurements were taken during the test procedure using both production
and laboratory sensors, Table 2. A majority of the measurements could be continuously
recorded using data acquisition equipment. However, PM mass collection efficiency was
estimated over discrete periods by using gravimetric sampling techniques. A Sierra
Instruments BG3 partial flow exhaust sampling system was used to sample the post filter
exhaust stream. The sampled exhaust flow passed through a small paper filter that
collected PM. PM concentration in the exhaust stream was estimated by measuring the
mass of PM on the paper filter and dividing by the measured sampled exhaust flowrate.
The engine-out PM levels were calculated by sampling the exhaust stream with a straight
through pipe fitted, while post particulate filter PM concentrations were calculated when
a filter was affixed. Collection efficiency was then calculated using Eq. 4.
[𝑃𝑀𝑖𝑛 ]−[𝑃𝑀𝑜𝑢𝑡 ]
𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛⁡% = ⁡ Eq. 4
[𝑃𝑀𝑖𝑛 ]
37
Table 2: GPF Test Setup
Sensors Manufacture/Model # Location

UEGO ECM/ FA1000 1 Pre-GPF
Differential Pressure Proprietary 4 Pre/Post GPF
Partial Flow Exhasut Sampling Seirra Instruments/ BG3 1 Post-GPF
Scanning Mobility Particle Sizer 1 Post-GPF
Type K Thermocouples 10 Engine-out, Pre-GPF, Within GPF
Particulate Matter Proprietary 6 Pre and Post GPF
Engine
2L Turbocharbed DISI I4 GM/ Ecotec (MY2010)
Turbo Borg Warner/ K04
Fuel 93 Octane E10
Oil Mobile 1 Synthetic/ 5W-30
GT-Power Aftertreatment Models
Automotive modelling software packages are available that allow engineers with
development platforms to digitally model and test potential designs. Some of the more
popular packages for automotive aftertreatment systems include AVL Boost, Gamma
Technologies GT-Power, and Exothermia’s Axisuite. At the request of FCA, GT-Power
was used in this work to model both a TWC and a GPF.
GT-Power’s base template for a 1-D catalytic converter was utilized to create a
simple model for TWC calibration, Figure 11. Known physical parameters including
wall thickness, catalyst volume, inlet cell density, and precious metal loading were
entered into the GUI, Figure 12. Properties that were either not physical or unknown
were chosen with the assistance of the GT-Power help manual or selected by choosing
values that are published in the literature. Chemical kinetics were then modelled in the
TWC system by attaching a surface reactions template to the catalyst block. For this
38
work, a simplified version of the reaction set proposed by Ramathan and Sharma was
utilized.
Figure 11: GT-Power Model of TWC
39
Figure 12: TWC Graphical User Interface
The Ramanathan and Sharma kinetics set utilizes 15 kinetics equations, Figure
13. 9 of these equations occur at the precious metal (PGM) sites while 6 of the equations
occur at the cerium sites. The cerium reactions are responsible the oxygen storage
dynamics where oxygen is either, adsorbed by the catalyst, or desorbed and participates
in reactions with gaseous species. In contrast the reactions occurring at PGM sites deal
exclusively with gaseous reactants. The reaction set was simplified by assuming
hydrocarbons species could be modelled as a single species, 𝐶3 𝐻6 ; Equations utilizing
𝐶3 𝐻8 were removed.
40
Figure 13: TWC Kinetic Reaction Set[42]
A simple aftertreatment system of a GPF was developed in GT-Power utilizing
the 1-D Diesel Particulate Filter (DPF) block, Figure 14. The model was used for
calibration purposes, contained the minimum number of flow components. Parameters
41
that were either known or which could be reasonable estimated from values in the
literature were entered into the model using the GUI.
Figure 14: DPF GT-Power Model
In addition to the required flow components, a variety of sensors were also placed
in the models. Values including conversion efficiencies, lambda, exhaust temperature,
and catalyst temperature were recorded during simulation. The sensed values were then
passed through the S-block to Simulink for calibration and controls, Figure 15.
42
Figure 15: GT-Power S-Block GUI
Both the TWC and GPF GT-Power models were designed to be co-simulated with
dedicated Simulink models. Inlet conditions were passed from Simulink into GT-Power
while sensed signals were passed from GT-Power into Simulink. This was done by
placing a communications block in both models referred to as an S-block. The Block
allowed for communication between the two models at a specified rate. For calibration
work, models inputs were changed at slow rates under the assumption that conditions
were quasi-steady state. As many model iterations were run, a communication rate of
1Hz was selected for calibration since increased rates were found to increase computation
time.
Calibration parameters that were changed in the GT-Power model between
iterations but not within iterations were passed from MATLAB through Excel
spreadsheets. Using the command ‘xlswrite’ in MATLAB, desired calibration parameters
43
were written into a file in the working directory. The GT-Power model then read values
from the spreadsheet by specifying said values within the case setup, Figure 16.
Figure 16: GT-P Case Setup, Reading Excel Files
MATLAB
MATLAB is high level coding package sold by Mathworks and is commonly
used by engineers in industry. The latest version, 2015b, was used to import and process
data, generate figures, run Simulink models, write kinetic parameters to Excel files, and
to conduct optimizations.
The MATLAB optimizer “fmincon” was used in this work for all optimizations
routines. The selected optimizer calibrates a set of user defined parameters, starting at a
set of user defined initial conditions. The optimizer tests the sensitivity of each parameter
44
to correlate the effect of reducing a defined objective function. Parameters that are found
to reduce the defined function are tuned by the optimizer until adjustment no longer
reduces the objective function. Within this work the objective functions were defined as
the sum of root mean squared difference between sampled experimental values and the
simulation values, Eq. 5. Graphically this can be explained by Fig. 17; the goal would be
to minimize the difference between the experimental curve and simulated curve. Each is
sampled, and the summation of sampled squared differences is the objective function. By
minimizing the objective function, the two curves will converge.
𝑦 = ∑(𝑥 𝑖,𝑠𝑎𝑚𝑝𝑙𝑒 − 𝑥 𝑖,𝑒𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡𝑎𝑙 )2 Eq. 5
45
Figure 17: Example of Comparison Between Simulation and Experimental Values
46
Figure 18: Lambda Signal over Filter Loading Cycle
47
Chapter 4: TWC Model Development and Calibration
A three-way catalytic converter (TWC) was modelled in GT-Power (GT-P)
utilizing the 1-D catalyst and reaction templates, Figures 11 and 14. Catalyst parameters
were specified within templates using a combination of known and assumed values.
Selected values are listed below in Table 3.
Table 3: GT-Power TWC Model Parameters
Catalyst Volume 0.8L

Cell Density 900 cspi
Channel length 84 mm
Wall thickness 15 mil
Substrate Material Cordierite
The active site density of Platinum Group Metals (PGM) within the catalyst is a
critical parameter which effects the reaction rates occurring on all PGM reaction sites.
This parameter was calculated using Eq. 6, under the assumption that all precious metals
on the catalyst were actively participating in reactions.
𝑃𝑡𝑙𝑜𝑎𝑑𝑖𝑛𝑔 𝑃𝑑𝑙𝑜𝑎𝑑𝑖𝑛𝑔 𝑅ℎ𝑙𝑜𝑎𝑑𝑖𝑛𝑔

𝑆𝑖𝑡𝑒⁡𝑑𝑒𝑛𝑠𝑖𝑡𝑦 = 𝑉𝑐𝑎𝑡𝑎𝑙𝑦𝑠𝑡 ∗ ( + + ) Eq. 6
𝜌𝑃𝑡 𝜌𝑃𝑑 𝜌𝑅ℎ
48
Once basic parameters of the TWC were determined, the model required that the
kinetic reactions be tuned so the model could predict the conversion efficiencies of
various exhaust species. Possible tuning parameter options included the activation
energies and pre-exponential coefficients for each of the 13 kinetic reactions, the 8
coefficients of the Langmuir-Hinshelwood inhibition expression, and the cerium loading.
To fully calibrate all the possible tunable coefficients, it was hypothesized that both
transient and steady state data would be required. As the mass of datasets was not
available, and under the advisement of individuals at FCA, it was determined that the
majority of effort should be concerted to calibrating the coefficients of the 8 PGM site
kinetic reactions, and to setting the oxygen storage quantity. The inhibition terms and
cerium site reactions were fixed at the values reported by Ramathan and Sharma.
The calibration process occurred in several steps. The first was to analyze the
experimental data and determine the inlet conditions and conversion efficiencies over the
range of inlet air-fuel ratios. The next step was to calibrate the cerium loading to set the
oxygen storage capacity of the catalyst. Once set, the final step was to calibrate the
kinetic coefficients using a series of three optimizers.
Data Analysis
The TWC testing conducted at FCA involved an exhaust aftertreatment system
containing a close coupled TWC and an underfloor TWC. The close coupled catalyst
contained a relatively high PGM loading and an unknown oxygen storage capacity. The
engine was run at various steady-state operating conditions and emissions measurements
49
were taken; gas emission bench measurements were taken at the engine-out and tailpipe
locations while FTIR measurements were taken at either the engine-out, between
catalysts, or tailpipe locations.
From the various engine operating conditions which TWC testing was conducted,
the idle conditions were selected to be used for kinetic calibration. This selection was
made since the idle test conditions occurred at similar flowrates and exhaust gas
temperatures, but at a range of air-fuel ratios.
Data Analysis: Engine Map
An engine emissions map was generated utilizing the engine-out emissions
measurements taken during the idle test cases. This map was used for GT-Power inlet
conditions and contained the following species:
𝐶𝑂2 , 𝐻2 𝑂, 𝑂2 , 𝑁𝑂, 𝐻2 , 𝐶𝑂, 𝐶3 𝐻6 , 𝑁𝑂2 , 𝑁2 𝑂, 𝑁𝐻3 , 𝐶𝐻4 . FTIR measured species
concentrations were used to generate spline-fit curves for each on the non-diatomic
species, 𝑂2 concentrations from the bench emission analyzer were curve-fit, and 𝐻2 was
estimated using a water-gas shift equilibrium expression. Estimated emissions values
achieved using the fitted functions were then utilized to build an engine map with
Simulink that was utilized for simulation inlet conditions.
Fitting FTIR Measured Emissions
For each of the non-diatomic exhaust species, steady state test case points were
plotted against the air-fuel ratio at which testing occurred. The resulting points were then
50
fitted using smoothing splines, resulting in functions relating species concentrations to
AFR. As an example, shown in Figure 19 is the plot of engine out 𝑁𝑂 vs. Lambda.
Figure 19: Engine-out NO vs. Lambda
Several hydrocarbon species were measured in wet concentrations by the FTIR
instrumentation including 𝐶12 𝐻26 , 𝐶2 𝐻4 , 𝐶2 𝐻6 𝑂, 𝐶3 𝐻6 , 𝐶3 𝐻8 , 𝐶7 𝐻8 , 𝐶𝐻2 𝑂, 𝑎𝑛𝑑⁡𝑇𝐻𝐶⁡ (𝑐3).
Within the specified reaction set, only one hydrocarbon is required. However, in addition
to the hydrocarbon actively participating in kinetic reactions, methane was also
considered as an inlet species representative of non-reactive hydrocarbons. From
observation of the data, it was apparent that the reactivity of 𝐶𝐻4 was minimal, Figure
51
20. The FTIR measured concentration of CH4 was utilized directly as an inlet species,
while the model reactive hydrocarbon, 𝐶3 𝐻6 , was calculated using Eq. 7.
[𝐶3 𝐻6 ,𝑖𝑛𝑙𝑒𝑡 ] = [𝑇𝐻𝐶𝐹𝑇𝐼𝑅 ] − 3[𝐶𝐻4 𝐹𝑇𝐼𝑅 ] Eq. 7
Figure 20: Inlet and Outlet 𝐶𝐻4
The emission bench measurement of oxygen was used to create a function for
engine-out oxygen concentration. One benefit to using this sensor was, that unlike the
FTIR measurements, the oxygen measurement was taken at the engine-out location for
52
every test case. Shown in Figure 21 is the engine out measurements for oxygen and the
associated fit curve.
Figure 21: Engine-out Oxygen Concentration
Since an FTIR cannot measure diatomic species and hydrogen was not measured
using the bench emissions equipment, it is required that hydrogen be estimated, Eq. 8
[14]. This equation is derived from the equilibrium expression of the water-gas shift
taking place in-cylinder where K is dependent of in-cylinder conditions. In this work K
was assumed at a value of 3; typical values range from 1.5 to 4.5.
53
𝑥 𝐻 2𝑂 𝑥 𝐶𝑂
𝑥 𝐻2 = Eq. 8
𝐾𝑥 𝐶𝑂2
Figure 22: Estimated Hydrogen Concentrations
In addition to engine-out emissions, a fitted function was generated for the close-
coupled catalyst inlet cone exhaust gas temperature. The thermocouple measurements
taken when the FTIR was located either between catalysts or at the tailpipe locations
were used to fit the function. FTIR measurements require the sampling of a significant
portion of exhaust gas, causing a reduction in temperature downstream of the
54
measurement. The fitted function and associated temperature measurements are displayed
in Figure 23.
Figure 23: Exhaust Gas Temperature Fitting
Catalytic Converter Conversion Efficiencies
Post catalyst exhaust gas species were fitted in the same manner as the pre-
catalyst species were when making the engine map. Conversion efficiencies for
𝐶𝑂, 𝑁𝑂𝑥 ⁡and non-methane hydrocarbons were then calculated at a range of AFR’s using
55
Eq. 9. Shown in Figure 24 is the resulting conversion efficiency curves, which were
utilized in the kinetic parameters optimization objective functions.
𝑃𝑃𝑀𝑠𝑝𝑒𝑐𝑖𝑒𝑠 ,𝑖𝑛⁡−𝑃𝑃𝑀𝑠𝑝𝑒𝑐𝑖𝑒𝑠,𝑜𝑢𝑡⁡
𝐶𝑜𝑛𝑣𝑒𝑟𝑠𝑖𝑜𝑛⁡% = 100 ∗ Eq. 9
𝑃𝑃𝑀𝑠𝑝𝑒𝑐𝑖𝑒𝑠,𝑖𝑛⁡
Figure 24: Catalyst Conversion Efficiency Curves
Oxygen Storage Capacity
The oxygen storage capacity of the tested catalyst was unknown and the transient
data needed to experimentally determine the value was unavailable. Instead, the oxygen
storage capacity was assumed to be 4 grams. Simulations were run where the inlet AFR
was changed in step increments between lean and rich conditions, Figure 25. For each
simulation, the simulated oxygen storage was calculated using Eq. 10, derived from the
work done by Giovanni and Grizzle [45]. The designed optimizer would then compare
56
the calculated value to 4 grams using the objective function Eq. 11, and vary the tuning
parameter “cerium loading” to reduce the difference.
𝜆𝑜𝑢𝑡 −𝜆𝑖𝑛
𝑜𝑥𝑦𝑔𝑒𝑛 𝑠𝑡𝑜𝑟𝑎𝑔𝑒 [𝑔] = 0.23 ∗ 𝑀𝐴𝐹 ∗ ∫ Eq. 10
𝜆𝑖𝑛
2
𝑌 = (4 − 𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦𝑜2 ,𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 ) Eq. 11
Figure 25: Pre and Post Lambda Signals for Oxygen Storage Simulations
Calibration of Kinetic Values

57
A set of 13 kinetic reactions were used to characterize the conversion efficiency
of exhaust species across the TWC. The equations originate from a paper by Sharma et
al. Within this equation set, reactions occur at two locations, PGM and cerium sites. The
PGM sites are locations of precious metal loading where reaction activation energies are
reduced while the cerium sites account for reactions dealing with the adsorbing and
desorption of oxygen from the catalyst. The rate at which each reaction proceeds is
specified by the rate expression Eq.12.
𝐸𝑎 ,𝑖
𝐾𝑖 [𝑅𝑒𝑎𝑐𝑡𝑎𝑛𝑡𝑎][𝑅𝑒𝑎𝑐𝑡𝑎𝑛𝑡𝑏 ]exp⁡( − )
𝑅𝑇𝑆
𝑅𝑎𝑡𝑒 = Eq. 12
𝐺
In this expression a unique pre-exponential factor and activation energy exists for
each reaction. Additionally for the PGM terms there exists a common inhibition term,
“G”, that contains 8 tunable coefficients, Eq. 13. This is the Langmuir Hinshelwood
inhibition term and is present to account for the competition and deactivation of reaction
sites due to the presence of fast hydrocarbons, 𝑁𝑂𝑥 , and C0,
𝐸𝑎 ,1,𝐺
𝐺 = (1 + 𝑘1,𝐺 exp⁡(− )[𝐶𝑂]
𝑇𝑠
𝐸𝑎 ,2,𝐺
+ 𝑘2,𝐺 exp (− ) [𝐶3 𝐻6 )2 (1
𝑇𝑠
𝐸𝑎,3,𝐺 𝐸
+ ⁡⁡⁡𝑘3,𝐺 exp (− ) [𝐶𝑂]2 [𝐶3 𝐻6 ])(1 + 𝑘4,𝐺 exp (− 𝑎,4,𝐺 ) [𝑁𝑂])
𝑇𝑠 𝑇𝑠
Eq. 13
58
The 8 reactions occurring on PGM sites were the primary focus for this work,
making for a total of 16 tuned parameters. The remaining 18 parameters of the inhibition
function and cerium reactions were assumed to be those of reaction set A4 of [42].
The optimizer that was used to identify the kinetic parameters, “fmincon”,
requires that upper and lower limits, in addition to initial values be applied to each of the
tuned parameters, Table 4. The values from kinetic set A4 of Sharma’s paper were used
for the initial values while the upper and lower limits were set providing a margin around
of values found in the literature.
Table 4: Optimizer Initial Conditions and Bounds
Parameter name Optimization # Units Lower Bound Upper Bound Initial Value
Cerium Loading 1 mol/m^3 10 1000 350
Pre-exponential Factor Eq.1 2,3,4 36 1.00E+17 2.54E+14
Activation Energy Eq. 1 2,3,4 k 9478.516 138450 13405.0998
Pre-exponential Factor Eq.2 2,3,4 3.17E+12 1.92E+24 1.92E+14
Activation Energy Eq. 2 2,3,4 k 12953 130530 15579.745
Pre-exponential Factor Eq.4 2,3,4 1.668E+12 1.00E+19 1.81E+15
Pre-exponential Factor Eq.7 2,3,4 500000 78800000000 7880000000
Parameter identification for the kinetic terms was done in 3 optimization steps
using 3 different objective functions, Eq. 14-16. The first optimizer concentrated on CO
59
conversion, the second compared CO and THC, while the third compared CO, THC and
NOx. The result of each optimization was fed into the proceeding optimization as the
initial conditions until the process was complete. The method of chaining optimization
routines was found to be a requirement due to the system complexity. Preliminary
optimizer development utilizing only the last step proved fruitless.
2
𝑌 = ∑(𝐶𝑂𝑠𝑖𝑚,𝑖 − 𝐶𝑂𝑒𝑥𝑝,𝑖 ) Eq. 14
2 2
𝑌 = ∑(𝐶𝑂𝑠𝑖𝑚,𝑖 − 𝐶𝑂𝑒𝑥𝑝,𝑖 ) + ∑(𝑇𝐻𝐶𝑠𝑖𝑚,𝑖 − 𝑇𝐻𝐶𝑒𝑥𝑝,𝑖 ) Eq. 15
2 2 2
𝑌 = ∑(𝐶𝑂𝑠𝑖𝑚,𝑖 − 𝐶𝑂𝑒𝑥𝑝,𝑖 ) + ∑(𝑇𝐻𝐶𝑠𝑖𝑚,𝑖 − 𝑇𝐻𝐶𝑒𝑥𝑝,𝑖 ) + ∑ (NOx sim,i − NOx 𝑒𝑥𝑝,𝑖 )
Eq.16
60
Chapter 5: GPF Model Development and Calibration
A one-dimensional model of a non-catalyzed Gasoline Particulate Filter (GPF)
was developed utilizing the powertrain modelling software GT-Power. Modelling was
accomplished using the 1-D component template for a Diesel Particulate Filter (DPF).
The model was then calibrated to insure simulated pressure drop across the filter and PM
mass collection efficiency matched experimental data. Calibration was accomplished
utilizing 2 sets of optimization routines, calibrating a total of six system parameters.
GPFs are an iterative design of DPFs, which have been developed for Diesel
applications, and been in use for over two decades. Small differences exist between filter
types in order to optimize the design for each application. Given the similarities, it is
assumed that the governing equations utilized within the DPF block would be valid for
GPF devices.
Known parameters including cell density, wall thickness, and filter volume were
specified within the GT-Power Graphical User Interface (GUI). A single set of Kinetic
parameters for PM thermal oxidation were taken from literature and applied to reactions
occurring both inside the filter wall and on the surface in the soot-cake layer [37].
Reactions were limited to oxidation of PM with oxygen, Eq. 17. Passive oxidation
through 𝑁𝑂𝑥 was not included since the exhaust of gasoline engines contains minimal
61
concentrations of 𝑁𝑂𝑥 . Selected values for both physical and kinetic parameters are
presented in Table 5.
0.6⁡𝑂2 + 𝐶 → 0.8⁡𝐶𝑂 + 0.2⁡𝐶𝑂2 Eq. 17
Table 5: GPF Parameters
Filter Volume 1.4 L

Wall Thickness 8 mil
Substrate Material Cordierite
Primary Particticle Diameter 6.6742e-8 m
Particle Diameter 63728e-8 m
Thermal Conductivity of PM Deposit 0.8368 W/(m-k)
PM Thermal Oxidation Activation Energy 150 KJ/mol
PM Thermal Oxidation Pre-Exponential Factor 28 m/k-sec
PM Thermal Oxidation CO selectivity 0.8
Once the known and estimated parameters in the GPF were set, tuning of several
parameters was still needed to ensure that the collection efficiency and pressure drop
across the filter would match experimental data. Six parameters were selected for tuning,
Table 6. In addition, the bounds on reasonable values and the initial conditions for the
optimizers are listed. Limits were selected based on suggested values within the GT-
Power help menu. Values were then adjusted based on the stability of optimization
routines.
62
Table 6: GPF Tuning Parameters
Parameter Name Optimization # Units Lower Bound Upper Bound Initial Condition
Pore Diameter 2 micron 10 40 35.28863902
Filter Porosity 2 fraction [0-1] 0.35 0.55 0.549785213
Clean Filter Wall Permeability 1 mm^2 1.00E-08 1.00E-06 4.0127E-08
Percolation Constant 2 fraction [0-1] 0.84 0.98 0.870811071
Particulate Packing Density Inside the Wall 2 Kg/m^3 6 160 17.73057857
Soot Cake Layer Porosity 2 fraction [0-1] 0.8 0.98 0.881172632
Data analysis
Pre-processed particulate filter data was provided by FCA in the form of
MATLAB data files. Relevant data signals included filter inlet exhaust gas temperature,
pre-filter lambda, and 3 production quality differential pressure signals. To reduce the
level of signal noise, a 100 time-step running average filter was applied to each of the
signals utilizing the MATLAB function “smooth”. The Pressure voltage signals were
converted to pressure in Kpa using the Eq. 18-20. Eq. 18 was provided by the
manufacture while Eq. 19-20 were derived under the assumption that the sensors operate
in a linear manor within the manufactures specified voltage and pressure ranges.
17
𝐷𝑃1 = (𝑣 − 0.5) ∗ Eq. 18
4
50
𝐷𝑃3 = (4.5−1.35 ∗ (𝑣 − 1.35)) − 5 Eq. 19
5.25
𝐷𝑃4 = ( ∗ (5𝑣 − 0.5)) − 0.25⁡ Eq. 20
4
63
Plots of the inlet temperature, lambda, and pressure signals were generated and
are presented in Figures 26-28. There was a significant difference in the magnitude of
pressure signals, this was assumed to be caused by both the large uncertainty in
production quality sensors, and that the sensors were being operated at conditions near
the lower limit of their advertised sensing ranges. However, while the magnitudes were
different, the change in pressure values were similar among sensors. For calibration
purposes, the pressure sensor signal with the highest magnitude was arbitrarily selected.
64
Figure 26: GPF Loading Cycle, Lambda Signal
65
Figure 27: GPF Loading Cycle, GPF Inlet Temperature
66
Figure 28: GPF Loading Cycle, Differential Pressure Signals
Particulate filter collection efficiency was calculated using gravimetric sampling
techniques. PM mass was filtered from the sampled exhaust stream with a paper filters.
The paper filters were replaced and weighed at discrete time instances and used to
estimate collection efficiency, Figure 29.
67
Figure 29: GPF Loading Cycle, Measured Filtration Efficiency
Simulation and Optimization
GT-Power simulations were run through a GT-Power/MATLAB communications
block within a Simulink model. GPF inlet conditions were defined within the Simulink
model; PM flow-rate, filter inlet exhaust temperature, and exhaust flow-rate measured
during experimental testing were passed into the GPF model as inlet conditions.
Experimental and simulation results were compared using the objective functions of each
optimization routine.
A set of two optimization routines were used to calibrate the pressure drop and
filtration efficiency of the GPF model, both utilizing the MATLAB optimizer “fmincon”.
The first optimization routine was utilized to calibrate the pressure drop across the GPF
when the device contains no particulate matter. This value is known as the clean wall
68
pressure drop and was found to be primarily effected by the GT-Power parameter “clean
wall permeability”. The experimental value for the clean wall pressure drop was assumed
to be the filter pressure drop observed during the start of the loading cycle, 4.65 Kpa.
Simulations were run utilizing the experiential exhaust temperature and flowrate, but with
a PM flow-rate of zero. The simulation value for clean wall pressure drop was assumed
to be the simulated GPF pressure drop once the system reached steady-state conditions.
The simulated pressure drop and experimental value were compared within the
optimization objective function by taking the squared difference between the two values,
Eq. 21. To minimize the objective function value, the value of clean wall permeability
was changed and the simulations were run until reasonable agreement was achieved.
𝑌 = |𝐷𝑃𝑐𝑙𝑒𝑎𝑛, 𝑠𝑖𝑚 − 𝐷𝑃𝑐𝑙𝑒𝑎𝑛, 𝑒𝑥𝑝 | Eq. 21
A second optimization routine was utilized to calibrate the PM mass filtration
efficiency and pressure drop across the GPF as the device is loaded with PM. Simulations
were run utilizing experimental values for exhaust temperature, lambda, exhaust flowrate,
and PM flowrate. The solution for clean wall permeability from the previous optimizer
was utilized, and the 5 remaining parameters were calibrated using through the
optimization script. For each iteration of potential sets of calibration parameters, 19600 of
time was simulated by the GT-Power model. Simulation results for pressure drop and
collection efficiency were sampled and compared to corresponding experimental results
using the optimization function for the second optimizer, Eq. 22. In particular, filtration
69
efficiency was sample in accordance to the discrete instances it was calculated at during
experimental work while pressure drop was sampled every 500 seconds from 2000 to
19,500 seconds.
2 2
𝑌 = 10 ∑ (𝐷𝑃𝑖,𝑠𝑖𝑚 − 𝐷𝑃𝑖,𝑒𝑥𝑝 ) + 20 ∑ (𝐹𝐸𝑖 ,𝑠𝑖𝑚 − 𝐹𝐸𝑖 ,𝑒𝑥𝑝 ) Eq. 22
70
Chapter 6: Results and Discussion
The primary goal of this work was to develop and calibrate aftertreatment models
for both a TWC and a GPF. 1-dimensional models were developed in GT-Power using a
combination of known and assumed component parameters. Inlet conditions, tuning
parameters, and simulation results were passed between aftertreatment models and
Simulink utilizing communications blocks in each software package. Using an engine
map developed with experimental catalyst data, models were calibrated using
optimization routines within MATLAB. Resulting calibrated aftertreatment models
showed reasonable agreement to experimental data. Results from each step of the
calibration process are presented along with a discussion of possible errors and
improvements.
TWC Data Analysis
Exhaust species conversion efficiency curves and an engine map were developed
utilizing a TWC dataset provided by FCA. The developed engine map was used for inlet
conditions for both TWC and GPF simulations. The map was developed by spline fitting
engine-out emissions measurements to correlate engine-out species to AFR. The resulting
71
map, Figure 30, appears reasonable and is similar to results published by John Heywood
in his book regarding internal combustion engines, Figure 31[15].
Figure 30: Calculated Engine-Out Species
72
Figure 31: Engine-Out Species Map From Literature[15]
Similar to the engine-out species, the catalyst-out species concentrations were
correlated to the air-fuel ratio utilizing spline-fits. Conversion efficiencies of CO, THC,
and 𝑁𝑂𝑥 were then calculated using the estimated values for both pre and post catalyst
exhaust species as a function of AFR. Resulting curves for CO, THC, and 𝑁𝑂𝑥 , Figure
32, show reasonable agreement to curves published within literature, Figure 33.
73
Figure 32: Calculated TWC Species Conversion Efficiencies
74
Figure 33: TWC Conversion Efficiency Values from literature [15]
The conversion efficiency for 𝑁𝑂𝑥 is the most suspect out of the three species in
which conversion efficiencies were calculated. From observation of published figures, it
would be expected that the rate of change of CO and THC conversion efficiencies is
relatively gradually as inlet conditions go from rich to lean. However, the conversion
efficiency for NOx would be expected to transition rapidly from nearly 100% to 0%, as
the inlet conditions go from rich to lean. Rapid transitions was not observed in calculated
75
conversion efficiency for NOx; little data was available at values close to stoichiometry
and appear to have caused the slope to be more gradual than what might be expected.
There is additional uncertainty in estimated conversion efficiency values due to
the way testing was conducted. During any 1 test case, FTIR emissions measurements
were taken at 1 of 3 locations; engine-out, between catalyst, or tailpipe. A variety of test
conditions including cam sweeps, idle AFR, and part throttle cases were run repeatedly
over several days. The FTIR measuring probe was then moved and testing was repeated.
Given the time between measuring emissions pre/ post catalyst, variability in fuel,
atmospheric conditions, instrumental calibration, or the component wear may have
skewed results. There is no guarantee that operating conditions were consistent between
all cases. Ideally, measurement would have been taken simultaneously for catalyst inlet
and outlet conditions. However, this would have required an additional FTIR and only
one catalyst would have been able to be tested as opposed to testing the duel-catalyst
aftertreatment system.
TWC Calibration
TWC calibration was accomplished using 4 optimization routines within
MATLAB. The first optimization routine calibrated the oxygen storage capacity of the
modelled catalyst to 4 grams by tuning the cerium loading. Simulations were run where
the inlet gas was changed from lean to rich conditions using a step input. The simulated
oxygen storage capacity was calculated using a function derived from the work of
Giovanni and Grizzle [45]. The next 3 three optimization routines were used to calibrate
76
the chemical kinetics reaction set. A simulation was run with a slow ramp input changing
the AFR from rich to lean. Simulation results for species conversion efficiency were
compared to the calculated experimental values by sampling each set of curves at select
lambda values. The difference between curves was then given a weighted numerical
value using each optimization routine’s objective function. The developed optimization
routines within MATLAB adjusted calibration values in effort to reduce the objective
function value, reducing the error between experimental and simulation results. The first
of three optimizations concentrated on CO, the next compared CO and THC, and the final
compared CO, THC and NOx. Initial conditions and results for each optimization routine
are presented in Table 7.
77
Parameter name Optimization # Units Initial Value Optimization #1 Solution Optimization #2 Solution Optimization #3 Solution Optimization #4 Solution
Cerium Loading 1 mol/m^3 350 459.4224372 459.4224372 459.4224372 459.4224372
Pre-exponential Factor Eq.1 2,3,4 2.54E+14 2.542E+14 1.00E+17 1.00E+17 1.00E+17
Activation Energy Eq. 1 2,3,4 k 13405.1 13405.09983 13405.09983 13405.09983 13405.09983
Activation Energy Eq. 2 2,3,4 k 15579.745 15579.74501 1.56E+04 15579.74501 15579.74501
Pre-exponential Factor Eq.5 2,3,4 6.86E+09 6857000000 6857000000 6857000000 6857000000
78
Pre-exponential Factor Eq.8 2,3,4 180000 180000 180000 180000 180000
Pre-exponential Factor Eq.9 2,3,4 123000 123000 123000 123000 123000
Table 7: TWC Optimization Results
Figure 34: Final Optimization Curves
Comparing the kinetic parameter optimization results, it can be seen that only
small changes were made between the initial values and the final solution, Figure 34.
This confirms the results presented by Ramathan and Sharma in their 2011 paper.
However, the kinetic model used in this work was a simplified version and did require
alterations to achieve the preceding results. The kinetic set proposed by Ramathan
utilized two hydrocarbons, 𝐶3 𝐻6 and 𝐶3 𝐻8 , while the model used in this work was
simplified to use only one hydrocarbon, 𝐶3 𝐻6 . This simplification was made out of
79
necessity as optimizations using two inlet hydrocarbons proved fruitless. Several methods
of scaling experimental FTIR measured hydrocarbon concentrations to the two species
resulted similarly with hydrocarbon conversion efficiencies showing the trend in Figure
34. Running simulations utilizing various combinations of the complete kinetic set, it
found that a majority of hydrocarbon conversion under rich conditions was occurring
through the reaction between 𝐶3 𝐻6 and water. A similar reaction for propylene is not
included by Ramathan, but is included in alternative reaction sets [46] and may be a
required addition to more accurately capture system behavior.
80
Figure 35: Simulated Conversion Efficiencies Utilizing Full Reaction Set from literature
GPF Calibration
GPF calibration was accomplished utilizing two optimization routines. The first
routine adjusted clean wall permeability to calibrate the clean wall pressure drop across
the filter. Using the result from the first optimizer, the five remaining calibration
parameters were tuned by the second optimization routine to calibrate the pressure drop
and collection efficiency as the GPF is loaded with PM under steady state conditions,
Table 8. Simulations of 19,600 seconds of GPF loading under rich conditions were run;
measured pressure drop and collection efficiency were sampled and then compared to
81
experimental data using an objective function. Final results show reasonable correlation
between simulation and experimental data, Figure 36-37.
Figure 36: Experimental vs simulation pressure drop
82
Figure 37: Experimental and Simulated GPF Collection Efficacies
Table 8: GPF Optimization Parameter Results

Parameter Name Optimization # Units Initial Value Optimization #1 Solution Optimization #2 Solution
Pore Diameter 2 micron 35.288639 35.28863902 33.79794859
Filter Porosity 2 fraction [0-1] 0.5497852 0.549785213 0.549728597
Clean Filter Wall Permeability 1 mm^2 4.013E-08 4.0127E-08 4.0127E-08
Percolation Constant 2 fraction [0-1] 0.8708111 0.870811071 0.868271303
Particulate Packing Density Inside the Wall 2 Kg/m^3 17.730579 17.73057857 17.66336404
Soot Cake Layer Porosity 2 fraction [0-1] 0.8811726 0.881172632 0.87879062
The simulation value for pressure drop correlates well to experimental data; both
the magnitude and trend of decreasing slope are captured by the model. However, there is
83
some difference between the two. It is possible that the difference is caused by a sub-
optimal solution of the tuned parameters, and that by finding the true global optimum,
better agreement could be obtained. Given that a local optimization routine was used and
that the initial conditions were random, it is likely that a better solution could be found.
However, while using a global optimizer may improve results, the experimental data
utilized to calibrate the model may also be a cause of error; the signal from a production
differential pressure sensor was used to calibrate the model. The sensor used was sensing
values at the lower spectrum of its sensing range, and also, did not agree with other
production grade pressure sensors that were used during testing. The true pressure drop
across the filter was unknown to the required accuracy needed for calibration. Ideally, a
laboratory grade sensor would have been used to eliminate this concern.
At first glance, the results for collection efficiency are sub-optimal. The
simulation is only able to capture the behavior of half of the data. However, given
uncertainties in the testing procedure, it is reasonable to assume that the three extraneous
points are outliers. In addition, the three outlying points show a decreasing collection
efficiency as the filter is loaded. This trend would only be possible if PM was being
oxidized from the filter; as PM is collected in a filter, the additional material helps
increase the filtration efficiency.
One additional influence of error in the model is in regards to the engine map
used for inlet conditions. The engine used during GPF testing was a gasoline direct
injected 4-cyclindar while the engine used to develop the engine map was a Port Fuel
injected V6. Species emissions were fitted at a range of engine-out lambda values
84
between 0.94 and 1.2. The contributions of both data extrapolation and different engine
operation may introduce error regarding species concentrations in the exhaust gas. Within
the GT-Power models, gas properties were defined using NASA gas tables. Small
differences in species may effect pressure drop across the filter as gasses would behave
differently.
85
Chapter 7: Conclusions and Future Work
Government regulatory bodies around the world implement ever increasingly
stringent emissions standards in effort to reduce harmful pollutants from automobiles.
Vehicle manufactures comply with regulations through optimization of engine controls
and by implementing automotive aftertreatment systems. Historically Three-Way
Catalytic Converters (TWC) have been the aftertreatment device of choice in gasoline
applications, however with the next generation of vehicles utilizing Gasoline Direct
Injected (GDI) engines, additional aftertreatment components may be required. Similar to
Diesel engines, GDI engines emit significant quantities of Particulate Matter (PM). To
comply with current and future regulations, gasoline particulate filters (GPF) may be a
required addition to GDI aftertreatment systems.
The development of control algorithms is required to ensure optimal performance
of aftertreatment components. Regulatory bodies also mandate that On Board Diagnostics
(OBD) be implemented to ensure that aftertreatment systems function throughout the
useful vehicle life. To assist with algorithm development, automotive systems are often
modelled, providing algorithm designers a platform for development.
Models of both a GPF and of a TWC were built in the commercially available
powertrain modeling software GT-Power utilizing 1-D component templates. A
combination of known and assumed system parameters were used to characterize the
aftertreatment devices. Within the TWC, a simplified version of the kinetic reaction set
86
proposed by Ramathan and Sharma was employed, while a single reaction for PM
oxidation was utilized in the GPF model. The developed GT-Power models were
connected to MATLAB /Simulink, where inlet conditions and tuning parameters were
defined. The device models were then calibrated through developed MATLAB
optimization routines utilizing experimental data as the reference. In particular, 17 terms
in the TWC kinetic reaction set, characterizing species conversion and oxygen storage,
and 6 GPF parameters that effect the pressure drop and collection efficiency were tuned.
The developed optimization routines were successful at calibrating both the TWC
and the GPF models. Simulation results show reasonable agreement to experimental data
used for calibration. Small discrepancies between simulation and experimental results can
likely be attributed to uncertainty in inlet conditions and model assumptions. Once
models are validated using additional data sets, they will be used in part, to model one
possible aftertreatment solution for a GDI application; calibrated TWC and GPF models
will be used in conjunction with models of differential pressure, particulate matter,
lambda, and RTD sensors. Simulation studies will be conducted to evaluate system
performance under different drive cycles and candidate control strategies. Component
fault models will then be added for OBD algorithm development.
As diagnostic requirements tighten, the TWC reaction kinetics may require
improvement to reduce discrepancies between simulation and experimental results.
Tuning the kinetic terms of the Langmuir-Hinshelwood expression of the rate equations
may resolve differences, however it is also possible that a more detailed kinetic reaction
set is required. Additionally, to model exhaust species including 𝑁𝑂2 and 𝑁𝐻3 , which
87
regulations have been proposed, an alternative kinetic reaction set will be required.
Candidate reaction sets are currently being developed and will be tested and calibrated
using both steady state and transient testing in the future.
88
References
[1] 2015, “Key Events in the History of Air Quality in California,” California
Environmental Protection Agency, Air Resource Board, From
http://www.arb.ca.gov/html/brochure/history
[2] 1999, “The History of Reducing Tailpipe Emissions,” United States Environme ntal
Protection Agency, Report No. EPA420-F-99-017.
[3] 2015, “Delphi Emissions Handbook, Passenger Cars and Light Duty Vehicles,”
Delphi Automotive PLC, 2015-2016 ed.
[4] 2010, “Preliminary Discussion Paper – Amendments To California’s Low-Emission

Vehicle Regulations For Criteria Pollutants – Lev III,” State of California, Air Resource
Board.
[5] 2010, “Preliminary Discussion Paper – Proposed Amendments To California’s Low

Emission Vehicle Regulations – Particulate Matter Mass, Ultrafine Solid Particle
Number, And Black Carbon Emissions,” State of California, Air Resource Board.
[6] Chan, T. W., Meloche, E., Kubsh, J., Rosenblatt, D., Brezny, R., & Rideout, G., 2012,
“Evaluation of a gasoline particulate filter to reduce particle emissions from a gasoline
direct injection vehicle,” SAE Technical Paper 2012-01-1727.
[7] Sappok, A. and Wong, V., 2010, "Ash Effects on Diesel Particulate Filter Pressure
Drop Sensitivity to Soot and Implications for Regeneration Frequency and DPF Control,"
SAE Technical Paper 2010-01-0811.
[8] Sappok, A. and Wong, V., 2007, "Detailed Chemical and Physical Characterization of
Ash Species in Diesel Exhaust Entering Aftertreatment Systems," SAE Technical Paper
2007-01-0318.
[9] Sappok, A., Wang, Y., Wang, R., Kamp, C. et al., 2014, "Theoretical and
Experimental Analysis of Ash Accumulation and Mobility in Ceramic Exhaust
Particulate Filters and Potential for Improved Ash Management," SAE Technical Paper
2014-01-1517.
89
[10] Mayer, A., Ulrich, A., Czerwinski, J., and Mooney, J., 2010, "Metal-Oxide Particles
in Combustion Engine Exhaust," SAE Technical Paper 2010-01-0792.
[11] Owrang, F., Olsson, J., & Pedersen, J., 2004, “Chemical Analysis of Exhaust
Emissions from a Gasoline Direct Injection SI Engine,” SAE Technical Paper 2004-01-
1445.
[12] US EPA, “Overview of Greenhouse Gases,” United States Environmental Protection

Agency, from www3.epa.gov/climatechange/ghgemissions/gases.html.
[13] Zhang, S., McMahon, W., Toutoundjian, H., Cruz, M., & Frodin, B., 2010,
“Particulate Mass and Number Emissions from Light-duty Low Emission Gasoline
Vehicles,” SAE Technical Paper 2010-01-0795.
[14] Piock, W., Hoffmann, G., Berndorfer, A., Salemi, P., & Fusshoeller, B., 2011,
“Strategies towards meeting future particulate matter emission requirements in
homogeneous gasoline direct injection engines,” SAE Technical Paper 2011-01-1212.
[15] Heywood, J., 1988, “Internal Combustion Engine Fundamentals,” 1 st ed., McGraw-
Hill, New York, NY.
[16] Chan, T., Meloche, E., Kubsh, J., Brezny, R. et al., 2013, “Impact of Ambient
Temperature on Gaseous and Particle Emissions from a Direct Injection Gasoline Vehicle
and its Implications on Particle Filtration,” SAE Technical Paper 2013-01-0527.
[17] Whelan, I., Smith, W., Timoney, D., & Samuel, S., 2012, “The Effect of Engine
Operating Conditions on Engine-out Particulate Matter from a Gasoline Direct-injection
Engine during Cold-start,” SAE Technical Paper 2012-01-1711.
[18] Schmidt, J., Franz, J., Merdes, N., Brady, M. et al., 2002, "Utilization of Advanced
Three-Way Catalyst Formulations on Ceramic Ultra Thin Wall Substrates for Future
Legislation," SAE Technical Paper 2002-01-0349.
[19] Kim, C., Perry, K., Viola, M., Li, W. et al., 2011, "Three-Way Catalyst Design for
Urealess Passive Ammonia SCR: Lean-Burn SIDI Aftertreatment System," SAE
Technical Paper 2011-01-0306
[20] Kodama, Y. and Wong, V., 2010, "Study of On-Board Ammonia (NH3) Generation
for SCR Operation," SAE Technical Paper 2010-01-1071.
90
[21] Gong, J., Rutlan, R., 2014, “Filtration Characteristics of Fuel Neutral Particulates
Using a Heterogeneous Multi-Scale Filtration Model,” ASME, Internal Combustion
Engine Division Fall Technical Conference, Report # ICEF2014-5590.
[22] 2013, “Diesel Particulate Filters- Everything You Need To Know,” Professional
Motor Mechanic, from http://pmmonline.co.uk/technical/diesel-particulate- filters-
everything- you-need-know.
[23] Koltsakis, G., Bollerhoff, T., Samaras, Z., and Markomanolakis, I., 2012, "Modeling
the Interactions Of Soot and SCR Reactions in Advanced DPF Technologies with Non-
homogeneous Wall Structure," SAE Technical Paper 2012-01-1298.
[24] Yang, S., Lee, K., and Chong, H., 2010, “Characterization of Oxidation Behaviors
and Chemical-Kinetics Parameters of Diesel Particulates Relevant to DPF Regeneration,”
[25] Suresh, A., Yezerets, A., Currier, N., and Clerc, J., 2008, “Diesel Particulate Filter
System – Effect of Critical Variables on the Regeneration Strategy Development and
Optimization,” SAE Technical Paper 2008-01-0329.
[26] Haralampous, O., Koltsakis, G., Samaras, Z., Vogt, C. et al., 2004, “Reaction and
Diffusion Phenomena in Catalyzed Diesel Particulate Filters,” SAE Technical Paper
2004-01-0696.
[27] Seo, j., Won Soon, P., Myung, L., 2012, “The Best Choice of Gasoline/Diesel
Particulate Filter to Meet Future Particulate Matter Regulation,” SAE Technical Paper
2012-01-1255.
[28] Rubino, L., Crane, R. I., Shrimpton, J. S., & Arcoumanis, C., 2001, “Experimental
Evaluation of a Wall-Flow Filter for Gasoline Engine Particulate Emission Control,”
[29] Shimoda, T., Ito, Y., Saito, C., Nakatani, T., Shibagaki, Y., Yuuki, K., et. Al., 2012,
“Potential of a Low Pressure Drop Filter Concept for Direct Injection Gasoline Engines
to Reduce Particulate Number Emission,” SAE Technical Paper 2012-01-1241.
[30] Saito, C., Nakatani, T., Miyairi, Y., Yuuki, K. et al., 2011, "New Particulate Filter
Concept to Reduce Particle Number Emissions," SAE Technical Paper 2011-01-0814.
[31] Ito, Y., Shimoda, T., Aoki, T., Shibagaki, Y. et al., 2013, “Advanced Ceramic Wall
Flow Filter for Reduction of Particulate Number Emission of Direct Injection Gasoline
Engines,” SAE Technical Paper 2013-01-0836.
91
[32] Richter, J. M., Klingmann, R., Spiess, S., & Wong, K. F., 2012, “Application of
catalyzed gasoline particulate filters to GDI vehicles,” SAE Technical Paper 2012-01-
1244.
[33] Mohammed, H., Triana, A. P., Yang, S. L., & Johnson, J. H., 2006, “An advanced
1D 2-layer catalyzed diesel particulate filter model to simulate: filtration by the wall and
particulate cake, oxidation in the wall and particulate cake by NO2 and O2, and
regeneration by heat addition,” SAE Technical Paper 2006-01-0467.
[34] Versaevel, P., Colas, H., Rigaudeau, C., Noirot, R., Koltsakis, G. C., & Stamatelos,
A. M. , 2000, “Some empirical observations on diesel particulate filter modeling and
comparison between simulations and experiments,” SAE Technical Paper 2000-01-0477.
[35] Pontikakis, G., & Stamatelos, A. 2006, “3-D Catalytic Regeneration Modeling of
SiC Diesel Particulate Filters”, ASME, Journal of Engineering for Gas Turbines and
Power, Volume 128.
[36] Tang, W., Wahiduzzaman, S., & Leonard, A., 2007, “A Lumped/1-D Combined
Approach for Modeling Wall-Flow Diesel Particulate Filters-Applicable to Integrated
Engine/Aftertreatment Simulations,” SAE Technical Paper 2007-01-3971.
[37] Huynh, C. T., Johnson, J. H., Yang, S. L., Bagley, S. T., & Warner, J. R., 2003, “A
one-dimensional computational model for studying the filtration and regeneration
characteristics of a catalyzed wall-flow diesel particulate filter,” SAE Technical Paper
No. 2003-01-0841.
[38] Bissett, E., Farhang, S., 1985, “Thermal Regeneration of Diesel-Particulate

Monolithic Filters,” American Institute of Chemical Engineers, Volume 31, Issue 5.
[39] Jørgensen, M. W., & Sorenson, S. C. 1997, “A 2-dimensional simulation model for a
diesel particulate filter,” SAE Technical Paper 970471.
[40] Haralampous, O., Dardiotis, C., Koltsakis, G., and Samaras, Z., 2004, “Study of
Catalytic Regeneration Mechanisms in Diesel Particulate Filters Using Coupled
Reaction-Diffusion Modeling,” SAE Technical Paper 2004-01-1941.
[41] Strzelec, A., Toops, T., Daw, C., Foster, D. et al., 2010, "Diesel Particulate
Oxidation Model: Combined Effects of Volatiles and Fixed Carbon Combustion," SAE
Technical Paper 2010-01-2127.
[42] Ramanathan, K., Sharma, C., 2001, “Kinetics Parameter Estimation for Three Way
Catalyst Modeling,” Industrial and Engineering Chemical Research, Volume 50.
92
[43] Ramanathan, K., Sharma, C., 2012, “Global Kinetics for Ammonia Formation and
Oxidation Reactions in Commercial Three-Way Catalyst,” Industrial and Engineering
Chemical Research, Volume 51.
[44] Gong, J. and Rutland, C., 2013, "Three Way Catalyst Modeling with Ammonia and
Nitrous Oxide Kinetics for a Lean Burn Spark Ignition Direct Injection (SIDI) Gasoline
Engine," SAE Technical Paper 2013-01-1572.
[45] Giovanni Fiengo, J. W. Grizzle et. al., 2005, “Dual-UEGO Active Catalyst Control
for Emissions Reduction: Design and Experimental Validation.” IEEE, Transactions on
Control Systems Technology, Volume 13.
[46] Stewart, J., Douglas, R., Goguet, A., and Glover, L., 2012, "Limitations of Global
Kinetic Parameters for Automotive Application," SAE Technical Paper 2012-01-1638.
[47] 2014, Automotive Handbook, 9th ed., Robert Bosch GmbH, Germany.
93

Mack James 2016 Thesis

Uploaded by

Copyright:

Available Formats

Mack James 2016 Thesis

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mack James 2016 Thesis

Uploaded by

Copyright:

Available Formats

Calibration of Automotive Aftertreatment Models through Co-Simulation with MATLAB

Graduate Program in Mechanical Engineering

The Ohio State University

Master's Examination Committee:

Dr. Shawn Midlam-Mohler, Advisor

Prof. Giorgio Rizzoni

New and existing government regulations mandate limits on various automotive

strategies to reduce PM levels, a Gasoline Particulate Filter (GPF) may be a required

addition to GDI aftertreatment systems.

GDI aftertreatment systems consisting of Three-Way Catalytic Converters (TWC)

and GPFs can be developed in commercial automotive powertrain modelling packages.

The performance of candidate component configurations can be virtually tested and

thermal PM oxidation kinetics.

Calibration was accomplished utilizing a series of optimization routines

matched experimental results. GT-P models were coupled to MATLAB using a

Simulink while tuning parameters in GT-P were adjusted in MATLAB. In total, 17

show reasonable agreement in terms of species conversion efficiencies, filtration

communications coupling, optimization routines, and results are presented herein.

March 1992 ....................................................Born- Madison Wisconsin

2010-2014 ......................................................B.S. Mechanical Engineering, University of

North Dakota (UND)

2014 to present ..............................................Graduate Research Associate, Department

of Mechanical and Aerospace Engineering,

The Ohio State University

Major Field: Mechanical Engineering

List of Tables .......................................................................................................................x

List of Figures ..................................................................................................................... xi

Chapter 1: Introduction ...................................................................................................... 1

Chapter 2: Literature Review .............................................................................................. 5

Beginnings of Emission Regulation ................................................................................ 6

Vehicle Emissions and Regulation.................................................................................. 6

Vehicle Emissions Production ...................................................................................... 10

Aftertreatment Devices ................................................................................................. 12

Catalytic Converters ...................................................................................................... 13

Catalytic Converter Operation Criteria/ Failure ........................................................ 14

Substrate Technology ................................................................................................ 15

2-Way Catalysts ......................................................................................................... 16

Lean 𝑁𝑂𝑥 Traps ........................................................................................................ 18

Urea Based SCR ........................................................................................................ 19

Passive SCR ............................................................................................................... 20

Particulate Filter Materials ........................................................................................ 21

Particulate Filter Operation and Performance ........................................................... 23

Aftertreatment System Modelling ................................................................................. 27

Particulate Filter Modelling ....................................................................................... 27

Catalytic Converter Kinetic Modelling ..................................................................... 30

Chapter 3: Tools and Methods .......................................................................................... 33

Experimental Data collection ........................................................................................ 33

GT-Power Aftertreatment Models ................................................................................ 38

Chapter 4: TWC Model Development and Calibration .................................................... 48

Data Analysis ................................................................................................................ 49

Data Analysis: Engine Map........................................................................................... 50

Fitting FTIR Measured Emissions ............................................................................. 50

Catalytic Converter Conversion Efficiencies ................................................................ 55

Calibration of Kinetic Values........................................................................................ 57

Chapter 5: GPF Model Development and Calibration...................................................... 61

Data analysis ................................................................................................................. 63

Simulation and Optimization ........................................................................................ 68

Chapter 6: Results and Discussion.................................................................................... 71

TWC Data Analysis ...................................................................................................... 71

TWC Calibration ........................................................................................................... 76