Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

SAE TECHNICAL
PAPER SERIES 1999-01-0460

Modeling the Transient Characteristics

of a Three-Way Catalyst
James C. Peyton Jones, J. Brian Roberts, Jinfeng Pan and Richard A. Jackson
University of Sussex

Reprinted From: Gas Direct Injection Aftertreatment and Exhaust Aftertreatment Modeling
(SP-1455)

International Congress and Exposition

Detroit, Michigan
March 1-4, 1999

400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

The appearance of this ISSN code at the bottom of this page indicates SAE’s consent that copies of the
paper may be made for personal or internal use of specific clients. This consent is given on the condition,
however, that the copier pay a $7.00 per article copy fee through the Copyright Clearance Center, Inc.
Operations Center, 222 Rosewood Drive, Danvers, MA 01923 for copying beyond that permitted by Sec-
tions 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying such as
copying for general distribution, for advertising or promotional purposes, for creating new collective works,
or for resale.

SAE routinely stocks printed papers for a period of three years following date of publication. Direct your
orders to SAE Customer Sales and Satisfaction Department.

Quantity reprint rates can be obtained from the Customer Sales and Satisfaction Department.

To request permission to reprint a technical paper or permission to use copyrighted SAE publications in
other works, contact the SAE Publications Group.

All SAE papers, standards, and selected

books are abstracted and indexed in the
Global Mobility Database

No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without the prior written
permission of the publisher.

ISSN 0148-7191
Copyright 1999 Society of Automotive Engineers, Inc.

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely
responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in
SAE Transactions. For permission to publish this paper in full or in part, contact the SAE Publications Group.

Persons wishing to submit papers to be considered for presentation or publication through SAE should send the manuscript or a 300
word abstract of a proposed manuscript to: Secretary, Engineering Meetings Board, SAE.

Printed in USA
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

1999-01-0460

Modeling the Transient Characteristics of a Three-Way Catalyst

James C. Peyton Jones, J. Brian Roberts, Jinfeng Pan and Richard A. Jackson
University of Sussex

Copyright © 1999 Society of Automotive Engineers, Inc.

ABSTRACT inherent computational complexity [3], and because the

dynamic effects are still not fully understood. A number
The dynamic behavior of three-way catalysts has signifi- of authors have therefore sought to measure or identify
cant impact on tailpipe emissions levels, but remains one simpler, semi-empirical models for catalytic dynamics
of the last unknowns in the overall vehicle emissions based on input / output measurements of the actual pro-
model. A simple empirical model (appropriate for use in cess, [4,5,6]. Though initial results are promising, recent
real-time engine control and on-board diagnostic strate- experimental results, [7], highlight some potential defi-
gies) has therefore been identified using fast response ciencies. In this paper a new catalyst model is developed
input / output measurements of the actual process. The which captures many of the dominant features of this
model is able to characterize the (significant) dynamic data, though further work is needed to extend the
behavior which has recently been observed under rich approach to operation at different engine speed and
conditions, as well as the more well known dynamics loads.
which arise from oxygen storage. The results therefore
compare well with measured responses over a wide CATALYST MODELS AND MODEL STRUCTURES
range of air / fuel ratio conditions.
Engine exhaust is composed of many gas constituents,
INTRODUCTION and a complete mathematical model describing all possi-
ble reactions that then occur within a TWC is clearly very
Three-way catalytic converters provide the most impor- difficult, if not impossible, to achieve. Even if it were pos-
tant and now very common means for reducing the sible to formulate such a model, it would be far too com-
tailpipe emissions of spark ignition gasoline engines. The plicated for use in the design and simulation of new
effectiveness of the converter depends critically on the engine controllers and OBD systems.
composition of exhaust gases produced by the engine,
In order to simplify the task, while retaining the dominant
with the optimum time averaged operating condition
dynamic behavior, it is common to consider only a small
occurring in a very narrow region around stoichiometric.
number of key gas components and reactions, and to
The goal of most current engine management systems is
approximate the spatially distributed nature of the cata-
therefore to maintain the air/fuel (A/F) ratio within this
lyst by a lumped parameter model. Within this frame-
window of operation, despite transient engine speed and
work there are then two main approaches. The first is
load disturbances, [1,2].
based on the Arrhenius reaction kinetics of each individ-
However tailpipe emissions depend not only on the time ual gas component, together with the dynamics of gas
averaged "steady state" exhaust conditions, but on the storage on the catalyst surface, while the second
dynamic behavior of the catalyst when subjected to tran- assumes that the latter alone dominates catalyst
sient disturbances in engine exhaust gas. A sufficiently behavior.
simple characterization of such behavior would allow the
catalyst dynamics to be integrated within a real-time, KINETIC MODELING – One advantage of kinetic model-
model-based, engine control strategy. The complete ing is its direct relationship to the underlying chemistry,
engine/catalyst system could then be optimized for both and the global nature of the model which holds across
performance and tailpipe emissions by actively manipu- both rich and lean conditions, and different catalyst tem-
lating, (rather than merely passively regulating), the A/F peratures. This avoids the use of temperature dependent
ratio operating condition. Such a model would also prove look-up tables, or discontinuous switches in model struc-
useful in on-board diagnostic applications. ture which are sometimes used to reflect the fueling con-
dition. The resulting system of coupled ODE's is still
Many existing models used in catalyst design are based
potentially quite complex, and a variety of different model
on the underlying physical processes of heat transfer,
structures have therefore been investigated.
chemical kinetics, and fluid dynamics, but are unsuitable
for real-time control of catalytic operation because of the

1
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

Aimard et al [4] for example modeled the kinetics of the EXPERIMENTAL DATA
four gas components, HC, O2, CO, NOx, and their inter-
action with oxygen stored on the catalyst. They also pre- In practice, the accuracy of many of the models intro-
sented a reduced order model which lumped all the duced above has been hard to assess because of the dif-
reducing and oxidation species together into two equiva- ficulties and costs inherent in obtaining transient
lent components. Both models give good long-term time measurements of all the gas components of interest.
cumulated results, though short-term dynamic responses While upstream and downstream UEGO sensors do pro-
are only presented for upstream and downstream vide a low cost indication of transient catalyst behavior, it
exhaust gas oxygen (EGO), and CO sensors. should noted that such signals are sensitive to the pre-
cise gas composition as well as the A/F ratio itself, [8].
The model of Germann et al [5] is in some ways similar to
The sensor will therefore give different readings
the reduced order model discussed above, since only two
upstream and downstream of the catalyst even under
components of the response, CO and O 2, are consid-
static conditions.
ered. In this case however the catalyst model is capable
of storing CO, (under rich conditions), as well as oxygen Recently however a relatively complete set of transient
when lean. The model also allows stored gas to be catalyst measurements have been obtained, as detailed
released by desorption, rather than only by reaction with in a companion paper [7]. Fast response exhaust gas
the incoming exhaust. Initially this was applied only to analyzers were used to measure the gas concentrations
CO storage, but a small O2 desorption mechanism was of CO, CO2, HC, and NOx upstream and downstream of
also added to resolve some numeric difficulties. The a demountable Pd:Rh catalyst, aged to 80 hours, 850o
stored oxygen level will therefore show a dynamic lean spike. Thin sheathed thermocouples were also
response to lean region changes in A/F ratio, unlike the used to measure upstream, downstream, and brick tem-
Aimard model where no reduction in the stored oxygen perature, and UEGO sensors were used to measure the
level can occur until the mixture returns rich once more. A/F ratio of the gas entering and leaving the catalyst.
Similar behavior, only relating to CO adsorption / desorp-
The dynamic response of the catalyst was then mea-
tion, would be observed in response to rich region
sured in response to step changes in A/F ratio imposed
changes in A/F ratio.
by the engine management system, first within the rich
region alone, then across stoichiometric, and finally in the
STORAGE DOMINATED MODELING – An alternative
lean region alone. Though space precludes a full review,
approach to modeling the catalyst chemical dynamics is
some of the principal conclusions which relate to model-
based on the observation that all the measured gas com-
ing are:
ponents respond to input changes over a similar time-
scale. This suggests that the process is dominated by a. All gas components respond over a similar time-
the relatively slow dynamics of a single process such as scale, suggesting that catalyst behavior is dominated
gas storage and release to / from the catalyst surface, by a single process, (probably the dynamics of gas
and that the other kinetics occur over a much shorter, storage and release).
and less significant, time-scale. The conversion effi- b. The time constant of this response generally
ciency of the catalyst can therefore be modeled using a becomes shorter as the mean A/F ratio operating
single nonlinear dynamic element to represent the stor- point is moved further away from stoichiometric.
age dynamics, (rather than the more complex coupled
c. The catalyst response to transients performed within
dynamic equations of the kinetic approach), followed by a
the rich region still shows significant dynamic behav-
more conventional static map of conversion efficiency.
ior, and it is unlikely that this will be captured by mod-
The latter maps the effective A/F ratio after the absorp-
els which include only oxygen storage mechanisms.
tion / desorption process, into individual tailpipe gas con-
centrations. d. These dynamics may reflect adsorption and desorp-
tion of reductants to/from the catalytic surface as well
The model of Brandt et al, [6], has a structure of this kind, as the re-establishment of equilibria in the steam
and appears remarkably effective, though in some tests reforming and water-gas shift reactions.
the model predicted a response to one side of stoichiom-
e. There appear to be two oxygen storage storage
etry, while the data actually deviated on the other. One
mechanisms, with different rates of storage and
potential difficulty is that only oxygen storage dynamics
release; the more rapid involves formation of M-O
are considered. Under rich operating conditions, the
bonds at the surface, the slower probably involves
model would therefore yield a purely static response
reversible oxidation of Ce2O3 to CeO2.
once any oxygen on the surface had been depleted.
Another possible difficulty is that stored oxygen cannot While a full discussion of the data is given in [7], some
leave the surface if conditions remain lean, since the only evidence for conclusion c) can be seen in the post cata-
removal mechanism is by reaction with a rich input feed- lyst responses to step changes of A/F ratio performed
gas; the lack of a desorption mechanism means that the entirely within the rich region. The “pre-“ and “post-“ plots
catalyst model will again show no dynamic behavior once of Figure 5, for example, show a dynamic response to
the surface has become completely saturated. step changes of l between 0.956 and 0.896, where,

2
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

actual A / F ratio
λ= 5
stoichiometric A / F ratio NOx nox

Under such conditions any stored oxygen would have 4

been completely depleted, so the observed dynamic HC hc
response must be due to other processes such as those
1
suggested in d) above.
Delta 3
One possibility, for example, is that CO is being stored A/F in
Storage CO co
and released, as postulated in the model of Germann et Dynamics
al. In this case however one might expect a pulse of CO 2
generation on rich-to-lean transitions, as stored CO is 2
removed from the catalyst by reaction with excess oxy- CO2 co2
gen. In practice only a very slight increase in CO2 pro-
1
duction is observed, especially when compared to that
produced on lean-to-rich transitions, (see Figure 7b); Delta
A/F out
[The latter results from stored O2 being removed from the
surface by reaction with excess CO]. If reductants are
being stored, the dominant component is therefore more Figure 1. Storage dominated model structure
likely to be hydrogen, producing an (unmeasured) pulse
of H20 generation on rich-to-lean transitions. Alterna- Both upstream and effective A/F ratios are expressed in
tively, the observed dynamics could be the result of the terms of their difference, ∆λ , ∆ λ′
(respectively) from
equilibria in the steam reforming and water-gas shift stoichiometric, ( λ = 1 ). In this way the state of the cata-
reactions readjusting to the new conditions. Whatever lyst can be represented by a single integrator, with posi-
the physical cause, it is important that the dynamic tive values reflecting oxidized states, and negative values
behavior of the catalyst in its reduced state is in some the state under reduced conditions. Such an interpreta-
way characterized by the model. tion is however only implicit in the current model since the
state of the catalyst surface is not represented directly.
AN EMPIRICAL CATALYST MODEL Instead, the focus is on representing the relationship
between the input and effective A/F ratios in the simplest
None of the models introduced above are able to account way possible. This is both easier in terms of system iden-
for all the experimental observations, though each has tification, and more straightforward in terms of future con-
certain desirable properties. In this section an empirical trol objectives, since the lowest emissions are achieved if
catalyst model is developed in order to test the feasibility the effective A/F ratio is maintained at stoichiometric.
of an essentially storage dominated model structure, but The rich region step changes shown in Figure 5, for
one which captures the dynamic behavior observed in example, often give a response which rises sharply in a
the rich region tests as well as the more well known step-like manner, before continuing to change in a more
dynamics which arise from oxygen storage. The model exponential fashion. Such behavior is typical of a first
does not attempt to describe the effects of changing order system with a finite zero in addition to the usual
space velocities, or catalyst temperatures, though such pole. The system transfer function,
variables would not be difficult to incorporate in principle,
[5]. The slow secondary oxygen storage mechanism is b ( s + a)
also ignored for the present. G(s) = a>b (1)
a ( s + b)
The basic model structure, whereby transient responses
of HC, CO, CO2 and output A/F ratio are obtained from for example, can be implemented using a single integra-
the pre-catalyst A/F ratio input, is shown in Figure 1. The tor as shown in Figure 2, and gives the response shown
model assumes that the dominant dynamics can be rep- in Figure 3, for the values b = 1, a = 2.
resented by a single process, represented by the “stor-
age dynamics” element in the figure. This relates the b/a
upstream A/F ratio to the effective A/F ratio experienced
by the other reactions taking place on the catalyst. The +
1 + 1/s b + 1
latter are assumed to occur over a much shorter time- delta
-
Effective
Integrator
scale, and are therefore characterized by static maps of A/F Delta A/F
the tailpipe concentration levels for each gas component,
obtained from steady state tests. The model therefore
has a similar overall structure to that of Brandt et al,
though the underlying dynamics of the storage element
are radically different. Figure 2. First order storage model with finite zero

3
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

under rich conditions, [8], rather than by any model defi-

ciency, especially since the other predictions, (which
depend entirely on the effective A/F ratio), are still quite
good.

lookup b/a

Gain3
+ 1
1 + * 1/s b +
- Delta
Delta
A/F in Gain1 A/F out

Figure 4. First order storage model with finite zero and a

nonlinearly varying time constant
Figure 3. Step Response of first order system with zero The response of the model and catalyst to step changes
performed entirely within the lean region is shown in Fig-
In the case of the catalyst system, rough estimates for ure 6. As one might expect the levels or CO and HC are
the parameters a and b can be made by observation of much lower than in the previous case, while the NOx lev-
the initial step height and settling time from say the HC els have risen. The dynamics of the response are also
response of Figure 5, [9]. This gives, much faster. In spite of these differences, the model still
provides a good fit to the data for all the gas components
Ts ≈ 5 sec ≈ 4 / b ⇒ b ≈ 0.8
considered, and even the measured and predicted A/F
initial step (2) ratio signals are in good agreement.
≈ 0.5 ≈ b / a ⇒ a ≈ 0.6
step amplitude Finally the behavior of the model and catalyst subjected
to step changes across stoichiometric is shown in the
The response to step changes in the lean region how-
responses of Figure 7. Once again the agreement
ever, shows much less of the dynamic exponential curve,
between the simulated and actual response of the down-
and more of a pure step change in the output. Such
stream UEGO sensor, CO, and NOx measurements is
behavior can still be obtained from the first order system,
fairly good. The predicted effective A/F ratio in particular
(1), if the pole position b approaches that of the zero, a.
accurately follows the classic plateau in the data near
The parameter b (which determines the time constant of
stoichiometry, reflecting the mechanism whereby the
the dynamics), therefore varies nonlinearly with the effec-
release of stored gas from the catalyst surface is able to
tive A/F ratio: if there is little gas storage / release, the
maintain stoichiometric conditions for a short period after
system time constant is fast, so b ª a; conversely, signifi-
the upstream A/F ratio has switched.
cant gas storage / release effects moderate any changes
in effective A/F ratio, so a "slow" or small value of b Some discrepancies however can be seen in the CO 2
should be used. and HC results. In part this is believed to be due to the
secondary / slow release of stored oxygen under rich
These effects can be incorporated in the catalyst model
conditions. The large pulse of CO2 production on lean to
by introducing a lookup table to scale the value of b
rich transition is however due to the (much faster) initial
according to the effective A/F ratio operating condition,
release of stored oxygen and its reaction with the CO rich
(see Figure 4). The resulting model is then nonlinear, but
feed gas. To model this effect, the CO 2 output from the
is likely to perform better over a range of different fuelling
static map in Figure 1 needs to be augmented by a quan-
conditions. To assess its performance, the model was
tity proportional to the rate at which oxygen is released
driven with the measured A/F ratio input, and its
from the catalyst surface. The latter can be approxi-
response to a variety of experimental tests compared to
mated by the difference between the upstream and
the actual measured outputs.
downstream (ie. effective) A/F ratios,
The responses of Figure 5, for example, show the perfor-
mance of the model under step changes performed O 2 release rate
entirely within the rich region. These are well character- = − K ( ∆λ − ∆λ ′) for − K ( ∆λ − ∆λ ′) > 0
ized by the model, giving good agreement between mea-
sured and experimental data. However an exception is where K is a constant of proportionality identified for the
seen in the prediction of downstream A/F ratio which data. If this extra component is added to the normal CO 2
should normally follow the effective A/F ratio quite closely. output, then the improved results of Figure 7f are
This is likely to be caused more by the commonly obtained.
observed inaccuracy in UEGO sensor measurements

4
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

Figure 5. Predicted, and measured, catalyst response to step changes within the rich region

5
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

Figure 6. Predicted, and measured, catalyst response to step changes within the lean region

6
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

Figure 7. Predicted, and measured, catalyst response to step changes across stoichiometric

7
 Downloaded from SAE International by Chalmers University of Technology, Monday, January 31, 2022

CONCLUSIONS REFERENCES

The feasibility of using a simple storage dominated model 1. Beaumont A.J., Beauchamp S.W. and Noble A.D., Air Fuel
for characterizing the dynamic behavior of a three-way Ratio Control Technology For Ultra Low Emissions Vehi-
catalytic converter has been investigated. An empirical cles, ISATA, 1993.
model which is able to characterize the (significant) 2. Beaumont A.J. and Frith A.M. Adaptive Control Of Tran-
dynamic behavior which has recently been observed sient Air-Fuel Ratio Using Neural Networks, ISATA, 1994.
under rich conditions, as well as the more well known
3. Vonkeman K.A., Jonkers G. and van Santen R.A., Studies
dynamics which arise from oxygen storage, has therefore
to the Functioning of Automotive Exhaust Catalysts Using
been developed, giving results which compare well with
In-Situ Positron Emission Tomography, SAE paper 910843,
measured responses over a wide range of A/F ratio con-
1991.
ditions.
4. Aimard F., Sorine M., Mathematical Modeling of Automotive
The effects of changing space velocities, catalyst temper- 3-way Catalystic Converters with Oxygen Storage Capac-
ature and slow oxygen storage mechanisms, however, ity, Control Engineering Practice, Vol.4, No.8, pp.1119-
have been ignored for the present. Further work is there- 1124, 1996.
fore necessary to extend the model for operation under
different speed and load conditions, and this is the sub- 5. Germann H.J., Onder C.H. and Geering H.P., Fast Gas
ject of on-going research. Concentration Measurements for Model Validation of Cata-
lytic Converters, SAE paper 950477, 1995.
ACKNOWLEDGMENTS 6. Brandt E.P., Wang Y. and Grizzle J.W., A Simplified 3-way
Catalyst Model for Use in On-Board SI Engine Control and
The authors wish to acknowledge the financial and tech- Diagnostics, ASME International Congress and Exposition,
nical support of the following sponsors: Sixth ASME Symposium on Advanced Automotive Technol-
ogies, 1997.
• Engineering and Physical Sciences Research Coun-
cil (EPSRC), UK 7. Jackson R.A., Peyton Jones J.C., Pan J., and Roberts J.B.,
Chemical Aspects of the Dynamic Performance of a Three-
• Ford Motor Company Ltd, Advanced Vehicle Techon- Way Catalyst, SAE paper 99????, 1999.
logies, UK
8. Germann H.J., Taglaiferri S and Geering H.P., Differences
• Horiba Instruments Ltd, UK
in Pre- and Post Converter Lambda Sensor Characteris-
• Johnson Matthey, Catalytic Systems Division, UK tics, SAE paper 960335
• Ricardo Consulting Engineers Ltd, Vehicle Control 9. Peyton Jones J.C., Atherton D.P., Root Locus Diagrams,
and Calibration and the Effect of Zeros on System Response , Int. J. Electri-
• Rover Group, Powertrain, Gaydon, UK cal Engineering Education, Vol.34, No.1, pp.46-67, 1997
• SAGEM, Research and Development