You are on page 1of 64

INSTITUTE OF TRANSPORTATION STUDIES

UNIVERSITY OF CALIFORNIA, BERKELEY

Engine and Transmission Modeling


for Heavy-Duty Vehicles

Diana Yanakiev,
Ioannis Kanellakopoulos

August 1995

PATH TECHNICAL NOTE 95-6

CALIFORNIA PARTNERS FOR ADVANCED TRANSIT AND HIGHWAYS


This work was performed as part of the California PATH Program of
the University of California, in cooperation with the State of Califor-
nia Business, Transportation, and Housing Agency, Department of
Transportation; and the United States Department of Transportation,
Federal Highway Administration.

The contents of this report reflect the views of the authors who are
responsible for the facts and the accuracy of the data presented herein.
The contents do no necessarily reflect the official views or policies of
the State of California. This report does not constitute a standard,
specification, or regulation.
Engine and Transmission Modeling for
Heavy-Duty Vehicles
Diana Yanakiev
Ioannis Kanellakopoulos
UCLA Electrical Engineering
Los Angeles, CA 90095-1594

MOU 124

PATH Technical Note

This work was performed as part of the California PATH Program of the University of
California, in cooperation with the State of California Business, Transportation, and
Housing Agency, Department of Transportation; and the United States Department of
Transportation, Federal Highway Administration.
The contents of this report re ect the views of the authors who are responsible for the
facts and the accuracy of the data presented herein. The contents do not necessarily
re ect the ocial views or policies of the State of California. This report does not
constitute a standard, speci cation, or regulation.

May 1995
ii
Engine and Transmission Modeling for Heavy-Duty Vehicles
Diana Yanakiev and Ioannis Kanellakopoulos
May 1995

Abstract
In this report we present an overview of our engine and transmission modeling for heavy-duty
vehicles. The models we construct are suitable for control design and simulation, and they are
used in our Advanced Vehicle Control Systems (AVCS) design for trucks and buses. We rst
review basic concepts from spark-ignition engine modeling and then proceed to a detailed
description of our three-state model of a turbocharged intercooled diesel engine, which is
the engine type used in heavy-duty commercial vehicles. Then we review basic concepts of
transmission modeling, including torque converter and transmission mechanicals, and argue
that automatic transmissions for heavy vehicles and passenger cars can be modeled the same
way, albeit with di erent parameter values.

Keywords
Advanced Vehicle Control Systems Control Systems
Buses Public Transit
Commercial Vehicle Operations Trucking

iii
iv
Executive Summary
In our research we are addressing the design of Advanced Vehicle Control Systems (AVCS)
which will allow the automated operation of heavy-duty vehicles on the intelligent highways
of the future. Our e ort focuses on longitudinal control algorithms, and hence requires
models which capture all the important characteristics of the longitudinal vehicle dynamics.
Hence, our rst task was the modeling of turbocharged diesel engines and hydrodynamic
automatic transmissions, both of which are used in the powertrains of commercial buses and
trucks.
Since our models are to be used for AVCS design, the level of detail of each subsystem
model must be dictated by its signi cance for the dynamical behavior of the whole vehi-
cle. Viewing the engine and the transmission as components of the system, rather than
as separate systems, determines a speci c objective for the modeling task, which results in
models that are highly nonlinear and complex enough for our purposes, but not nearly as
complicated as those used for engine or transmission design and control.
Instead of assuming full familiarity of the reader with engine and transmission modeling,
we provide an extensive review of the basic concepts. We start with spark-ignition engines
and then move to turbocharged intercooled diesel engines, the type used in most heavy-
duty commercial vehicles. We show that we only need three states to capture the dynamic
behavior of such an engine: the engine speed, the turbocharger rotor speed, and the pressure
of the intake manifold. The other two states usually associated with spark-ignition engine
models, the fuel mass ow rate and the exhaust gas recirculation rate, are not needed in
diesel engines because of the di erent fueling method, which injects fuel directly into the
cylinders before combustion.
Then we review basic concepts of transmission modeling, including torque converters
and transmission mechanicals, and argue that automatic transmissions for heavy vehicles
and passenger cars can be modeled the same way, albeit with di erent parameter values.

v
vi
Contents
1 Modeling a Turbocharged Diesel Engine for AVCS Design 1
1.1 Introduction : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 1
1.2 Nomenclature : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 1
1.3 Background : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 3
1.3.1 Spark ignition engine models : : : : : : : : : : : : : : : : : : : : : : : 4
1.3.2 Turbocharged diesel engine models : : : : : : : : : : : : : : : : : : : 9
1.4 Model Description : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 12
1.4.1 Selection of states : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 12
1.4.2 Modeling equations : : : : : : : : : : : : : : : : : : : : : : : : : : : : 14
1.4.3 Simulation results : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 26
2 Transmission Modeling 31
2.1 Introduction : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 31
2.2 Hydrodynamic Automatic Transmission (HDAT) : : : : : : : : : : : : : : : 32
2.2.1 Torque converter : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 32
2.2.2 Transmission mechanicals : : : : : : : : : : : : : : : : : : : : : : : : 33
2.3 Automatic Transmission Modeling : : : : : : : : : : : : : : : : : : : : : : : : 36
2.3.1 Torque converter modeling : : : : : : : : : : : : : : : : : : : : : : : : 36
2.3.2 Transmission mechanicals modeling : : : : : : : : : : : : : : : : : : : 43
2.4 HDAT for Heavy Duty Vehicles : : : : : : : : : : : : : : : : : : : : : : : : : 49
References 51

vii
viii
List of Figures
1.1 Schematic diagram of a carbureted SI engine. : : : : : : : : : : : : : : : : : 5
1.2 Diagram of engine dynamics submodel. : : : : : : : : : : : : : : : : : : : : : 6
1.3 Car model for longitudinal control. : : : : : : : : : : : : : : : : : : : : : : : 8
1.4 Schematic diagram of a turbocharged diesel engine. : : : : : : : : : : : : : : 10
1.5 TC diesel engine model representation in SIMULINK. : : : : : : : : : : : : : 13
1.6 Corrected compressor air mass ow rate (f1). : : : : : : : : : : : : : : : : : : 15
1.7 Compressor and intercooler model. : : : : : : : : : : : : : : : : : : : : : : : 16
1.8 Compressor eciency (f2). : : : : : : : : : : : : : : : : : : : : : : : : : : : : 17
1.9 Intake manifold, torque production, and crankshaft dynamics model. : : : : 19
1.10 Temperature rise in the engine (f6). : : : : : : : : : : : : : : : : : : : : : : : 22
1.11 Exhaust manifold, turbine, and TC rotor dynamics model. : : : : : : : : : : 23
1.12 Corrected turbine air mass ow rate (f7). : : : : : : : : : : : : : : : : : : : : 24
1.13 Turbine eciency (f9). : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 25
1.14 Apparent turbine eciency (f10). : : : : : : : : : : : : : : : : : : : : : : : : 26
1.15 Complete model of TC diesel engine in SIMULINK. : : : : : : : : : : : : : : 27
1.16 Engine speed, indicated torque, and friction torque response to given pro les
of fuel command (index) and load torque. : : : : : : : : : : : : : : : : : : : 29
2.1 Fluid coupling in action. : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 32
2.2 Split-guide ring designed to reduce turbulence. : : : : : : : : : : : : : : : : : 33
2.3 Planetary gearset. : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 34
2.4 Fluid coupling model. : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 37
2.5 Blade angle e ect on tangential velocity at the pump exit. : : : : : : : : : : 40
2.6 Torque converter diagram. : : : : : : : : : : : : : : : : : : : : : : : : : : : : 41
2.7 Schematic diagram of Hydramatic 440 . : : : : : : : : : : : : : : : : : : : : : 44
2.8 Actual coecient of friction. : : : : : : : : : : : : : : : : : : : : : : : : : : : 47
2.9 Complete bond graph diagram of the transmission mechanicals of Hydramatic
440 . : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 48

ix
x
Chapter 1
Modeling a Turbocharged Diesel
Engine for AVCS Design
1.1 Introduction
The problem addressed in this chapter is the development of a turbocharged diesel engine
model suitable for longitudinal control design for heavy duty trucks. The vast majority
of existing internal combustion engine models serves purposes such as engine performance
improvement or diagnostics. Engine models meant for vehicle control have been developed
only for normally aspirated spark ignition (SI) engines. Unfortunately, they cannot be
adapted to describe the compression ignition (CI) of the diesel engine and to capture the
e ect of the turbocharger.
After studying both gasoline and diesel engine models at di erent levels of complexity
we tried to build a model that best suits our needs. A serious e ort was made to achieve
simplicity without sacri cing accuracy. Simulations were performed using SIMULINK (The
MathWorks 1993) and the results obtained support the validity of the model.

1.2 Nomenclature
ai; bi; ci polynomial coecients, i = 0; 1; 2
Cpa speci c heat for air (kJ/kg K)
Cpx exhaust speci c heat (kJ/kg K)
(F=A)stoichiometric stoichiometric fuel to air ratio
pf friction mean e ective pressure (kPa)
Je e ective inertia of the engine (kg m2)
Jtc inertia of the turbocharger (kg m2)
k empirically determined constant
kf fueling constant (kg/m)
K empirically determined constant
m_ c compressor mass ow rate (kg/s)
m_ ai air mass ow rate into the manifold (kg/s)
m_ ao air mass ow rate into the cylinders (kg/s)

1
m_ egri EGR mass ow rate into the manifold (kg/s)
m_ egro EGR mass ow rate into the cylinders (kg/s)
m_ ex air mass ow rate into the exhaust manifold (kg/s)
m fuel mass (kg)
m_ f fuel rate (kg/s)
m_ fc commanded fuel rate in SI engine (kg/s)
m_ im air mass ow rate into the cylinder (kg/s)
m_ t turbine mass ow rate (kg/s)
Mc compressor torque (N m)
Mf engine friction torque (N m)
Mind engine indicated torque (N m)
Mload engine load torque (N m)
MP torque converter pump torque (N m)
Mt turbine torque (N m)
nr number of revolutions for each power stroke per cycle
Ne engine speed (rpm)
Ntc turbocharger rotor speed (rpm)
pamb ambient pressure (kPa)
pcx compressor outlet pressure (kPa)
pex exhaust manifold pressure (kPa)
pim intake manifold pressure (kPa)
QLHV fuel lower heating value (kJ/kg)
R gas constant (kJ/kg K)
rt turbine radius (m)
Sp mean piston speed (m/s)
Tamb ambient temperature (K)
Tcool coolant temperature (K)
Tex exhaust manifold temperature (K)
Tim intake manifold temperature (K)
Vd engine displacement volume (m3)
Vim intake manifold volume (m3)
vU=C turbine blade tip speed (m/s)
Y fueling index (or rack position) (m)
 fuel/air equivalence ratio
app apparent turbine eciency
c compressor eciency
ic intercooler eciency
ind indicated eciency
v volumetric eciency
speci c heat ratio of air
ex speci c heat ratio of exhaust gas
 density (kg/m3)
!e engine speed (rad/s)
!tc turbocharger rotor speed (rad/s)
t time (s)
2
1.3 Background
Despite their big variety, all internal combustion engines feature similar operation cycles.
The four-stroke cycle includes:
 intake | air/fuel mixture (for spark ignition) or air only (for diesel engines) is ingested
into the cylinder through the inlet valve(s); the piston moves from its top dead center
(TDC) toward its bottom dead center (BDC) position;
 compression | the piston goes from BDC to TDC with all valves closed; the resulting
pressure is much higher in the diesel engine and towards the end of this stroke the fuel
injection takes place;
 power | caused by spark ignition (SI) in the gasoline or by compression ignition (CI) in
the diesel engine, the combustion pushes the piston from TDC to BDC thus producing
torque;
 exhaust | now the exhaust valve(s) is open and the products of combustion are pushed
out by the piston moving from BDC up toward TDC.
The gas exchange or scavenging process is common to all internal combustion engines.
Its function is to e ect expulsion of the products of combustion from the engine cylinder and
their replacement by a fresh air charge in preparation for the next working cycle. However,
the scavenging methods used in the two-stroke and four-stroke engines are di erent. The
two-stroke cycle combines the air exchange strokes (intake and exhaust) into one. This
results in the following structure of the two-stroke working cycle:
 power | identical to the four-stroke case, but lasting only 105{130 degrees of crank
angle until about 50{75 degrees before BDC;
 scavenging | intake and exhaust valves are open simultaneously which inevitably
causes some mixing of the fresh air with the products of combustion; this stroke takes
place symmetrically about BDC occupying 100{150 degrees of crank angle;
 compression | same as in the four-stroke engine, but starting at about 50{75 degrees
after BDC and lasting 105{130 degrees until TDC.
Therefore, each cylinder in a two-stroke engine provides torque generation at each revo-
lution of the crankshaft, while in the four-stroke engine each cylinder produces torque once
every two revolutions. However, the two-stroke engine does not generate twice as much
torque and power as the four-stroke one. This is due to the fact that scavenging is not as
ecient in the exchange of exhaust gases with \fresh" air as in the exhaust/intake cycle
of the four-stroke engine. As a result, each combustion in the two-stroke engine yields less
energy than in the four-stroke one.
3
The purpose of turbocharging is to increase the mass of air trapped in the cylinders of
the engine by raising air density. This allows more fuel to burn, thus increasing the power
output of the engine for a given swept volume of the cylinders. A compressor is used to
achieve the increase in air density. The compressor is driven by a turbine, which itself is
driven by the exhaust gas from the cylinders. Obviously, the presence of a turbocharger
greatly alters the processes in the engine and respectively their description.
Other reasons for the diversity of engine models, besides the various types of engines,
are the di erent applications of the models and the computational considerations that might
exist. For example, when trying to improve engine performance, one needs a very detailed
description of the thermodynamic processes in the engine on an instantaneous basis. For
engine control, a less complex model is necessary to facilitate real time calculations. Even
further simpli ed models are appropriate for vehicle control where the engine is just a sub-
system and has much faster dynamics compared to the whole vehicle.
There are three groups of methods for engine simulation:
 quasi-steady;
 lling-and-emptying;
 wave action.
These have been listed in the order of increasing mathematical complexity and accuracy,
and decreasing reliance on empirical data. Simplicity and high speed of calculation are
usually achieved at the cost of extensive usage of empirical data, which restricts the model
validity only to a particular type of engine. On the other hand, if a model is based on
physical principles and can be easily modi ed to describe di erent types of engines, it is very
complicated and requires serious computational e ort.
Unfortunately, the diesel engine models that can be found in the literature are not readily
applicable to our problem. Even though some of them are created to be used for control,
they are meant for engine control, not for vehicle control.
Acknowledging the necessity for building our own model, we decided to:
 study the existing diesel and gasoline engine models;
 use the available gasoline engine models for longitudinal vehicle control as a guideline
for the level of complexity of our diesel engine model.

1.3.1 Spark ignition engine models


Of all the SI engine models studied, the following two proved to be the most useful:
 the model presented by Dobner (1980), which introduced a new approach to engine
modeling for control applications that has been widely adopted, and

4
exhaust gas
air/fuel recirculation spark advance load torque
? (EGR)
- EGR- ? ? engine-
air - combustion indicated- dynamics
throttle - carburetor fuel
air - intake
torque
- manifold fuel - speed
? ?
manifold pressure net torque

Figure 1.1 : Schematic diagram of a carbureted SI engine.

 the vehicle model by Cho and Hedrick (1989) and McMahon et al. (1990), which treats
the engine as a subsystem in a similar fashion as we need to.
Dobner (1980) has developed a mathematical model of the carbureted four-stroke SI
engine for application to dynamic engine control. Use of normalized parameters and a
modular structure make it possible to simulate di erent engines via changing component
performance characteristics and adding new model features. The diagram on gure 1.1
describes the principle elements of a four-stroke SI engine. The inputs available for control
are shown in boldface.
Nonlinear characteristics are incorporated in the detailed description of each component
thus making it possible to predict engine performance over a broad operating range.
The carburetor controls the air and fuel ow into the engine. Its response to inputs
is instantaneous | its model contains no dynamic elements. The air ow through the
carburetor is a function of the throttle angle ( ) and the intake manifold pressure (pim).
The fuel ow is determined by the carburetor characteristic, which shows that rich mixtures
are supplied near idle and at maximum power, and lean mixtures at normal road load
conditions.
The carburetor and the EGR valve control the ow into the manifold while the demands
of the cylinders and the valve systems dictate the ow out of the manifold. The mass of
each constituent is calculated by integrating the di erence between the ow-in and ow-out.
Some of the fuel evaporates while in the intake manifold (IM) and some is \atomized" in
ne droplets. The rest goes through the IM as liquid fuel lm on its walls. The division
of the IM constituents into fast- ow (air, EGR, fuel-gas, and fuel-droplets) and slow- ow
(fuel-walls) is re ected in the model by using two di erent transport delays.
A discrete form is chosen for the engine model | one computation per engine ring:
t = 120
speedeng numbercyl :
Consequently, the mass ow rates are transformed into incremental masses. Also the time
delays are obtained from the incremental delays using:
time delay = increm delay t :
5
load torque
{
+M brake -
indicated - +M? net - and
engine
load
engine - speed - engine
torque torque torque inertia acceleration integration speed
{ 6

friction torque calculation

Figure 1.2 : Diagram of engine dynamics submodel.

The fast ow incremental delay and the engine ow demand depend on the volumetric
eciency 1 , which is a nonlinear function of the engine speed and also increases as manifold
pressure increases.
The wall fuel mass ow rate is integrated back in time until the total wall fuel mass in the
IM is reached, in order to determine the slow- ow incremental delay and the corresponding
time delay, s. The latter is the time constant of a lag lter used in the model to account
for the inertia of the wall- ow fuel.
As a result of the calculations in the IM submodel, the ow-out incremental masses of air
and fuel are available. They are ingested into the cylinder to participate in the combustion
process, which generates torque. The torque production depends not only on the air ow
and air/fuel ratio, but also on the combustion eciency 2 of the engine and on its spark
control. Both air/fuel ratio and spark advance characteristics suggest a value which would
provide optimal torque. Deviations in either direction result in less torque. The combustion
eciency is a function of the engine speed and increases as charge density increases. All these
factors, along with the linear e ect of the air ow, determine the indicated torque. The fact
that the torque is produced a nite time after the charge is ingested into the cylinder is
accounted for by the intake to power stroke delay .
The nal step is obtaining the engine speed from the indicated torque considering the
engine dynamics. The transient performance depends on the inertia of the engine and its
load. The diagram on gure 1.2 describes the dynamics submodel.
Apart from yielding meaningful simulation results and being very exible, the model
presented by Dobner has the advantage of being comprehensive and yet compact and suitable
for control purposes.
A vehicle model of the typical front wheel drive car equipped with a V-6 engine has been
developed by Cho and Hedrick (1989) and McMahon et al. (1990). The block-diagram in
gure 1.3 incorporates all the factors which a ect the longitudinal performance of an auto-
mobile. The inputs available for control are shown in boldface again. The states belonging
1 Volumetric eciency is a measure of the e ectiveness of an engine's induction process and is de ned as
the volume ow rate of air into an engine divided by the rate at which the volume is displaced by the piston.
2 Combustion eciency is also known as thermal or indicated eciency.

6
to each major module are listed in the corresponding box.
Engine states
The state equation for the pressure in the intake manifold pim is derived using the ideal
gas law and the mass conservation principle:
p_im = , 4vVVd !e pim + RVTim (m_ ai + m_ egri) : (1:3:1)
im im
This equation shows how pim depends on the engine speed !e and the volumetric eciency
v, which, as mentioned earlier, is also a function of !e and pim itself. Vd is the displacement
volume, Vim is the IM volume, Tim is the IM temperature, m_ ai and m_ egri are respectively
the mass ow rates of the air and the EGR entering the IM. Similarly, the equation for the
exhaust gas mass ow rate from the IM m_ egro describes this state in terms of the engine
speed and volumetric eciency:
m egro = 4vVVd !e (m_ egri , m_ egro) : (1:3:2)
im
The equation associated with the fueling rate of the engine m_ f is valid for the sequential-
re port fuel injection system:
f m f + m_ f = m_ fc ; (1:3:3)
where m_ fc is the commanded fuel rate and f is the delay associated with the speci c fueling
method. Finally, the engine speed equation is obtained from the engine dynamics:
Je !_ e = Mind , Mf , MP ; (1:3:4)
where Je is the e ective inertia of the engine and torque converter, Mind is the indicated
engine torque, Mf is the friction engine torque, and MP is the torque converter pump torque.
The indicated torque production is a discrete process. However it is modeled as a contin-
uous time phenomenon. The production delays (intake-to-torque and spark-to-torque) are
employed to re ect the discrete nature of the four-stroke engine in the torque production
calculation.
Transmission states
During operation of the transmission in gear there is only one independent state variable,
the angular velocity of the turbine. It is the input to the planetary gear train, which
consists of two interconnected planetary gears (input and reaction ) and a single planetary
gear ( nal drive unit). During shift, however, there are two states. The turbine angular
velocity is related to the angular velocity of the reaction planetary carrier via the
speed reduction/multiplication ratio. In this case the reaction carrier velocity is the input
to the nal drive unit.

7
ENGINE
throttle - pim manifold pressure
spark - !e engine speed
fuel -
- m_ f fuel mass ow rate
EGR
- m_ egro EGR rate

production delays:
engine
| intake to torque brake
torque torque
converter | spark to torque
pump
speed

TRANSMISSION
!t torque converter turbine speed 
 clutches
and bands
!cr reaction carrier speed 
torque converter
- planetary
gear train
nal pump turbine
drive drive
output shaft
speed torque
DRIVETRAIN
- M axle shaft torque
s

disturbances - !fw front wheel speed


- V vehicle speed - vehicle speed
brakes

Figure 1.3 : Car model for longitudinal control.

8
Drivetrain states
The factors which a ect the longitudinal motion of the vehicle are considered in order to
derive the equations describing the drivetrain dynamics. More speci cally, the states of axle
shaft torque, front wheel speed, and vehicle speed are being determined using axle
shaft sti ness, constant rolling resistances of the tires, aerodynamic drag coecient, tire slip,
braking system, and suspension of the vehicle.

1.3.2 Turbocharged diesel engine models


A normally aspirated gasoline engine model cannot be modi ed to describe the turbocharged
diesel engine. However, the models reviewed in the previous section provide us with use-
ful techniques to approach our particular modeling task. Moreover, drawing a comparison
between the SI and CI engines, one can nd similarities as well as di erences, because as
mentioned earlier, there are common principles valid for all internal combustion engines. Sev-
eral models have been studied, covering the range from quasi-steady to lling-and-emptying
methods, and including intermediate levels of complexity.
The model developed by Winterbone (1977) is a wholly dynamic one, which represents the
engine and the turbocharger (TC) gas ows by a set of 30 interconnected rst order nonlinear
di erential equations. It employs the lling-and-emptying technique and, being based mostly
on physical principles, it gives an accurate description of the engine processes. Hendricks
uses the mean value method to create a series of models for di erent types of diesel engines
in (Hendricks 1986, 1989, Jensen et al. 1991). The mean value models are similar to the
quasi-steady ones with respect to simplicity. However, an important aspect of the mean value
method is the use of physically based models. In Hendricks' opinion, the fact that a given
system is complex does not necessarily imply that there are no underlying physical principles
which can give a simple overall picture of engine operation. The resulting models predict
correctly the steady state operating points as well as the most important aspects of dynamic
engine response. The model built by Jennings and Blumberg (1986) stands at a level between
the quasi-steady and the lling-and-emptying. It presents a method to calculate the engine
brake torque on an average rather than on an instantaneous basis, proving that there is no
loss of accuracy. A linearized model by Krutov is discussed in (Kullkarni et al. 1992). Flower
and Gupta (1974) give a state-space form of a simple discrete-time representation of the TC
diesel engine. Recently, Kao and Moskwa (1993) have tried to summerize the modeling
e orts at di erent levels of complexity presenting two models: mean torque production and
cylinder-by-cylinder . The rst one uses basically the quasi-steady approach and the latter
employs the lling-and-emptying technique.
Despite their di erences, all models describe the turbocharged diesel engine with charge
cooling, which is schematically presented on gure 1.4. The various modeling approaches for
the subsystems depicted in the diagram will be presented in parallel.

9
Air Exhaust Gases

  PPPP 6
? PPPP
 TC Rotor P
Compressor Turbine
PPPP  
PPPP 
P  6
?
Intercooler
Exhaust
Intake
Manifold
Manifold Fuel
6
? ?
- Speed
Engine
Diesel Engine and Crankshaft Assembly
 Load Torque

Figure 1.4 : Schematic diagram of a turbocharged diesel engine.

Compressor and turbine


Standard steady-state performance maps, which relate the mass ow rate and eciency of
the compressor to its pressure ratio, inlet temperature and the TC rotor speed are used to
model this subsystem. The disadvantage of this representation is its complete reliance on
empirical data provided by the compressor manufacturer. Unfortunately, it seems to be the
only possible approach, since it was used by all studied models. The turbine submodel is
built in a similar manner.
Intercooler
The process of compression raises temperature as well as pressure. Since the objective is to
increase inlet air density, intercoolers are often used to cool the air between the compressor
delivery and the cylinders, so that the pressure increase is achieved with the maximum rise in
density. The more involved intercooler models (Jennings et al. 1986, Winterbone et al. 1977)
employ energy balance on the gas and coolant ows through the intercooler to determine
their respective outlet temperatures. We adopted the simpler approach suggested in (Kao

10
and Moskwa 1993) to determine the outlet temperature using the cooling eciency.
Intake manifold
Di erent approaches are used to determine the intake manifold (IM) pressure. The lling-
and-emptying method used in (Jennings et al. 1986, Winterbone et al. 1977) provides detailed
consideration of all factors associated with the IM. Assuming that the heat transfer is negli-
gible, one can achieve certain simpli cation as in (Hendricks 1986, 1989, Jensen et al. 1991,
Kao and Moskwa 1993). In addition to that the IM can be viewed as a no volume component
(i.e., no temperature change occurs), which leads to an even simpler model. The air mass
ow into the cylinder depends on the engine speed and on the volumetric eciency. Jensen
et al. (1991) have stated the fact, that in the presence of a turbocharger, the volumetric e-
ciency is a function only of the engine speed as opposed to the case of a normally aspirated
engine, where it also depends on the intake manifold pressure.
Combustion and torque production
The major di erence in computational complexity of the models is due to their di erent
representation of the torque production process. The combustion can be described based
on physical principles as in (Winterbone et al. 1977). The advantage of this approach is
its capability to simulate the instantaneous uctuations of the engine speed. However, they
have no e ect on the vehicle dynamics, therefore applying this method will only increase the
computational burden, without contributing to the accuracy of our model.
The other commonly used technique for computation of the indicated torque is based
on using steady-state data. The engine speed and the air/fuel ratio are the factors that
a ect the combustion process. Empirical characteristics, speci c to the engine, are used to
determine the thermal eciency and respectively the produced torque.
Exhaust manifold
The importance of this subsystem comes from the presence of the turbocharger. The pres-
sure and temperature in the exhaust manifold are input conditions for the turbine. As in
the intake manifold model, several levels of detail are possible. The temperature can be de-
termined either from complete thermodynamic analysis of the combustion process, or using
empirical data for the temperature rise through the engine.
The pressure calculation depends on the turbocharging method. There are two di erent
ways in which the energy of the exhaust gases can be utilized to drive the turbine.
 With constant pressure turbocharging , the exhaust ports from all cylinders are con-
nected to a single exhaust manifold whose volume is suciently large to ensure that
its pressure is virtually constant. The unsteady exhaust ow processes at the cylinders
are damped into a steady ow at the turbine.

11
 The objective of pulse turbocharging is to make the maximum use of the high pressure
and temperature which exist in the cylinder when the exhaust valve opens, even at
the expense of creating highly unsteady ow through the turbine. In most cases the
bene t from increasing the available energy will more than o set the loss in turbine
eciency due to unsteady ow.
Pulse turbocharging is usually applied to automotive and truck engines. Employing the
pressure pulse to drive the turbine implies that an average exhaust manifold pressure model
would be inaccurate. Ledger et al. (1971) suggest how to avoid the involved thermodynamic
calculations and still account for the pulse e ect. In addition to the turbine eciency,
they introduce a corrective parameter called apparent eciency, which re ects the pulsating
pressure e ect on the turbine.
Crankshaft and TC rotor dynamics
All studied models use Newton's second law to derive the rotational models of the engine
and the TC rotor. The same technique is used in the spark ignition engine models.

1.4 Model Description


After careful analysis of the SI engine model for longitudinal control and extensive study
of the TC diesel engine modeling techniques, we have developed a model consisting of 3
di erential and several algebraic equations. The engine modeled is turbocharged, intercooled,
six-cylinder, with 0.014 m3 (14 liters) displacement volume. The model has been implemented
in SIMULINK for performing computer simulations. The diagram on gure 1.5 gives an
overall picture of the model structure, showing the subsystems as superblocks. As a result,
the interaction among the components can be clearly observed.

1.4.1 Selection of states


As in the SI engine case, two of the states are the intake manifold (IM) pressure and the
engine speed. However, due to the di erent fueling method, the fueling lag is not considered
here. In the diesel engine, the fuel is injected directly into the cylinder immediately before
the combustion takes place. This eliminates the need to account for the fueling dynamics.
The TC rotor speed is another state which needs to be introduced due to the presence of
the turbocharger.
The equation for the IM pressure pim has been derived by di erentiating the ideal gas
law:
p_im + 120
v Vd Ne
V p im = m
_ c
RTim ;
V (1:4:1)
im im

12
DIESEL ENGINE MODEL

TC rotor speed

Compressor
with IC TC rotor
Turbine
air compressor torque
mass turbine torque
flow T ex
rate IM Temperature
p ex

Intake Exhaust
Manifold Manifold

IM pressure
engine
speed air mass flow rate
F/A
fuel rate
2
engine speed
load 1
torque Torque Production 1
FUEL
COMMAND and Engine Dynamics
Engine
Speed
.

Figure 1.5: TC diesel engine model representation in SIMULINK.

13
where Tim and Vim are respectively the IM temperature and volume, v is the volumetric
eciency, Vd is the displacement volume of the engine, Ne is the engine speed in rpm (Ne =
!e 260 ), and m_ c is the compressor air mass ow rate.
The engine speed !e is obtained by integrating the angular acceleration of the crankshaft,
which is determined from Newton's second law:

Mind(t , i) , Mf (t) , Mload(t) = Je !_ e(t) : (1:4:2)

Mind is the indicated torque and Mf is the friction torque of the engine. Mload is the load
torque, determined by the transmission and the drivetrain subsystems of the vehicle model.
Je is the e ective inertia of the engine. The production delay i represents the average
di erence between the time of issuing a command to change the indicated engine torque
and the time when the injection valve can be operated. These events are determined with
respect to crankshaft angle. Therefore, the production delay and all other delays associated
with this model have constant values measured in crankshaft angle. Converted in seconds,
they become functions of the engine speed thus taking variable values.
The TC dynamics model is also derived using Newton's second law:

Mt , Mc = Jtc !_ tc ; (1:4:3)

where Mt is the torque provided by the turbine, Mc is the torque absorbed by the compressor,
and Jtc is the e ective inertia of the turbocharger. Integration of the TC angular acceleration
!_ tc yields the TC rotor speed !tc.

1.4.2 Modeling equations


Some intermediate computations are necessary to determine the other variables participating
in the state equations. Steady-state empirical characteristics and algebraic relations are
used in addition to the state equations to obtain a complete mathematical description of the
system. The subsystems will be discussed again in the sequence dictated by the mass ow
through the TC diesel engine, as indicated on gure 1.4. At each step we are also going to
\click" on the blocks of gure 1.5 to reveal the details.
Compressor model
The steady-state compressor performance map is used to determine the air mass ow rate
m_ c and the compressor eciency c, provided that the TC rotor speed Ntc (Ntc = !tc 260 )
and the pressure ratio across the compressor pcx=pamb are given. In order to eliminate any
variations due to the ambient pressure pamb and the ambient temperature Tamb, the maps
usually provide corrected instead of actual values for the ow rate and the TC speed. A

14
typical situation is when the corrected air mass ow rate is obtained from the compres-
sor characteristic and then the actual ow rate is calculated with respect to the ambient
conditions: !
m_ c = pp amb
f p ;N tc p cx
: (1:4:4)
Tamb 1 Tamb pamb
The function f1 (shown in gure 1.6) is represented in gure 1.7 as the two-dimensional look-
up table block C ow rate and the expression of equation (1.4.4) is computed in the mc dot
block. \link \center {\large \rm compressor\n corrected air mass flow rate}

f1
corrected air mass flow rate

6
compressor

0
3
9
2.5 8
p 2 7
cx /p 6
4
am 1.5 5 x 10
b 4 )
3 (rpm
1 speed
otor
TC r

Figure 1.6 : Corrected compressor air mass ow rate (f1 ).

The compressor eciency c is also speci ed from the performance map:


!
N
c = f2 pT ; ptc p cx
; (1:4:5)
amb amb

which is implemented using a two-dimensional look-up table in the CompE block of gure 1.7
and f2 is shown in gure 1.8. The temperature at the compressor outlet Tcx and the torque
absorbed by the compressor Mc then can be calculated as follows:
8 9
< 1 pcx ! , 1 3=
2
Tcx = Tamb :1 +  4 p , 15; (1:4:6)
c amb
2 ! , 1 3
m
_ C T p
Mc = c  pa! amb 4 p cx , 15 : (1:4:7)
c tc amb

15
COMPRESSOR AND INTERCOOLER

101.325 p_amb
1/p_amb
1/u 1
p cx/p amb
* f(u) (needed for
p cx/p amb {pr.term1} turbine
13.2 apparent
efficiency)

2590k C flow CompEff (f2)


f(u) Ntc corr f(u) 2
rate (f1)
Trq c compressor
Mux absorbed
Mux torque

mc dot f(u)
298.15
Mux f(u) 323.15
T_amb
Tcx T_cool

Mux f(u) f(u)


mass flow 14 Eff IC
parameter (f3) Mux Mux3
f(u) Mux f(u)
Memory p cx T im
compressor
1 2 3 mass flow 4
intake rate intake
TC rotor manifold mc dot
speed manifold
pressure temperature
N tc p im
.

Figure 1.7 : Compressor and intercooler model.

16
f2

efficiency}
0.8

efficiency
\rm compressor
0.6

compressor
0.4
\link \center {\large

0.2

0
3
9
2.5 8
p 2 7
cx /p 6
4
am 1.5 5 x 10
b 4 m)
1 3
speed (rp
otor
TC r

Figure 1.8 : Compressor eciency (f2).

The term containing the pressure ratio, which is identical for equations (1.4.6) and (1.4.7)
is computed as a Matlab function in the fpr.term-1g block of gure 1.7. Cpa is the speci c
heat and is the speci c heat ratio for air. The change of these parameters with ambient
conditions is negligible and they are considered to be constant in this model.
Intercooler model
An empirical characteristic has been used to specify the intercooler eciency ic given the
mass ow rate:
ic = f3(m_ c) : (1:4:8)
The eciency decreases almost linearly as the mass ow rate increases. The block E IC in
gure 1.7 contains the expression obtained by polynomial tting applied to empirical data.
The outlet temperature of the intercooler (i.e., the IM temperature Tim), is then calculated
from the cooling agent temperature Tcool and the compressor outlet temperature:
Tim = Tcx(1 , ic) + ic Tcool : (1:4:9)
The equation of gas ow through an ori ce is used in our model to determine the pressure
drop across the intercooler and later across the engine cylinders. This representation assumes
that the pressure drop of gas ow through an ori ce is proportional to the square of the mass
ow rate and inversely proportional to the density of the gas, yielding the following expression
for the intercooler:
m_ 2
p = k = k c im :
c m_ 2RT
(1:4:10)
im pim
17
Provided that we have calculated pim from equation (1.4.1), we can add the pressure drop
to it to get the pressure at the compressor outlet pcx:
pcx = pim + p : (1:4:11)
Equations (3.11) and (3.12) are combined in the p cx block of gure 1.7.
The dynamics of the compressor and the intercooler are fast compared to the rest of the
system, and thus neglecting them does not a ect the accuracy of this model. However, the
Memory block3 of gure 1.7 has been introduced to avoid the formation of an algebraic loop,
which would considerably increase simulation times.
Intake manifold model
As mentioned earlier, the volumetric eciency v of the turbocharged engine depends only
on the engine speed. This is due to the presence of the turbocharger, which makes the
system closed-loop and eliminates the independent e ect of the IM pressure on the volumetric
eciency. As suggested by Jensen et al. (1991), empirical data points relating the volumetric
eciency to the engine speed have been curve tted with a second order polynomial:
v = f4(Ne) = a0 + a1 Ne + a2 Ne2 ; (1:4:12)

which is the function in the eta vol block of gure 1.9.


The intake manifold pressure can be found from equation (1.4.1):

p_im + 120
v Vd Ne
V p im = m
_ c
RTim ;
V
im im

and then the air mass ow rate into the cylinder mim is calculated based on the volumetric
eciency and the air density:
m_ im = v 120
Vd Ne = v Vd Ne pim :
120 R Tim (1:4:13)

To avoid duplication of computational e ort, the term 120 v Vd Ne is being calculated only once
Vim
in the Product block of gure 1.9. It is needed for the implementation of equation (1.4.1) and
also for evaluating expression (1.4.13) in the m im dot block of gure 1.9.
Torque production model
The control input to the CI diesel engine is the fuel command. This is in contrast with the
SI engine where the throttle is the main control. In the diesel engine, the quantity m of
3 A Memory block holds its input signal for one integration step thus breaking an algebraic loop.

18
INTAKE MANIFOLD AND TORQUE PRODUCTION

air mass
flow rate 2
from
compressor
1 287/.0 1
intake R/Vim intake
manifold * manifold
pressure temperature
1/s +
2
p im
air mass
flow rate
083333 Mux f(u) into the
*
Vd/(120*Vim) m im dot cylinder
Product
f(u) Mux f(u)
eta vol (f4) F/A
KN eng
* 4
3 1.e mf dot fuel
A/F stoichK mass
Fuel kf flow
Command K
Qlhv/k rate
Trq f f(u) Trq
4 Load 01.05

* 5
f(u) Mux
Trq ind F/A
1/s omega eng eta ind equiv
+ Sum Trq Prod f(Neng)*f5 ratio
Delay
.33 f(u) tau i 3
1/J eng engine
speed

Figure 1.9 : Intake manifold, torque production, and crankshaft dynamics model.

19
fuel injected is proportional to the index 4 Y , which is determined by the governor5 at idle
and maximum speed, and by the accelerator pedal at normal road operation:
m = kf Y: (1:4:14)
The index is represented as the Fuel Command block in gure 1.9, which is a variable gain.
This is the input available for control.
The fuel mass rate is obtained from the engine speed:
m_ f = m Ne ; (1:4:15)
and is being used to calculate the fuel/air ratio in the cylinders:
(F=A)actual = mm _f ;
_ im (1:4:16)
which is done in the F/A block of gure 1.9. The actual fuel/air ratio is then divided by its
stoichiometric value for diesel fuel to determine the equivalent fuel/air ratio :

 = (F=A(F=A)actual : (1:4:17)
)stoichiometric
Since (F=A)stoichiometric is just a constant, (3.18) is represented as the gain block A/F stoich
in gure 1.9.
The indicated eciency of the engine ind is speci c for the particular engine and is
obtained from empirical data. It shows how the torque production depends on the engine
speed and on the fuel/air ratio. The eciency is higher when the mixture is leaner. It can
be related to the engine speed by a second order polynomial:
ind = (b0 + b1 Ne + b2 Ne2) f5() : (1:4:18)
The e ect of both factors has been decoupled in the empirical data and then curve tted with
polynomials. The resulting function is included in the eta ind block of gure 1.9.
Finally, the achieved torque is proportional to the fuel mass and the indicated eciency:
Mind = m QLHV ind = kf Y QLHV ind : (1:4:19)
The lower heating value for diesel fuel QLHV is a constant.
4 The index Y is de ned as the position of the fuel pump rack, which determines the amount of fuel
provided for combustion.
5 The governor is a device which controls the speed of a diesel engine by regulating the intake of fuel. In
road vehicles this device is active only when the engine is (a) idling, or (b) operating close to its rated speed.
In all other operating conditions, the governor is inactive and the fuel intake is controlled by the accelerator
pedal.

20
Crankshaft dynamics model
The engine rotational dynamics model has been presented already in equation (1.4.2):
Mind(t , i) , Mf (t) , Mload(t) = Je !_ e :
The variable delay block Trq Prod Delay in gure 1.9 has been used to represent i, which
depends on the engine speed.
The load torque Mload becomes an input from the automatic transmission torque con-
verter, once the engine model is incorporated in a vehicle model.
Calculation of the friction torque Mf is necessary for the torque balance in equation
(1.4.2). We have adopted a model which assumes quadratic dependency of the friction upon
the mean piston speed and linear dependency on the engine speed. The coecient c0 in the
expression for friction mean e ective pressure (pf ),
pf = c0 + 0:048 Ne + 0:4 Sp 2 ; (1:4:20)
is determined by tting empirical data. The mean piston speed Sp can be expressed in terms
of engine speed, given the number of cylinders and the stroke of the engine. Hence equation
(1.4.20) is written in the more general form:
pf = c0 + c1 Ne + c2 Ne2 : (1:4:21)
Finally, the friction torque is found from
Mf = pf V2d 1000
nr ; (1:4:22)
where nr stands for number of revolutions for each power stroke per cycle.6 Both equations
(1.4.20) and (1.4.22) are included in the Trq f block of gure 1.9.
Subtracting the friction and load torques from the indicated torque is performed in the
Sum block of gure 1.9. The result is the net torque, which is then divided by the e ective
engine inertia Jeng (actually multiplied by its inverse in the gain block 1/J eng) to yield the
crankshaft acceleration. The latter is then integrated in the omega eng block of gure 1.9 to
get the engine speed. Finally, the N eng block converts it from rad/s to rpm as needed for
most computations.
Exhaust manifold model
Another e ect of the combustion process is the temperature rise in the engine Te, which we
will need in order to determine the exhaust manifold conditions. Similarly to the indicated
eciency, it is a function of the fuel/air ratio and the engine speed (shown in gure 1.10):
Te = f6(; Ne) : (1:4:23)

6For example, a 4-stroke 6-cylinder engine has nr = 1=3, since it produces 6 power strokes per cycle and
each cycle comprises two revolutions.

21
\link \center {\large \rm engine temperature rise}
f6

800

engine temperature rise


600

400

200

0
1.5

1 2000
equ
iva 1500 Ne
len 0.5
t fu
el/a 1000
ir r (rpm)
atio 0 500 e speed
engin

Figure 1.10 : Temperature rise in the engine (f6 ).

It is represented by the two-dimensional map block T e rise in gure 1.11. The engine
temperature rise is found from empirical data. In general, it increases linearly with increasing
engine speed and mixture richness.
In order to determine how the exhaust manifold temperature is being a ected by the
intake manifold temperature, the latter has been delayed by 3, which is the time from intake
valve opening (IVO) to exhaust valve opening (EVO). Therefore, this delay also depends
on the engine speed and has been implemented by the variable delay block IVO to EVO in
gure 1.11. The exhaust manifold temperature is obtained in the t ex block by adding the
temperature rise in the engine to the delayed IM temperature:
Tex(t) = Tim(t , 3) + Te(t) : (1:4:24)
The mass ow rate out of the engine is determined by adding the fuel and air mass ow
rate into the engine with the appropriate delays:
m_ ex(t) = m_ f (t , 1) + m_ im(t , 2) : (1:4:25)
The time from fuel injection (FI) to EVO is denoted as 1, and 2 is the time from intake
valve closing (IVC) to EVO.
We calculate the pressure drop through the engine as in (Horlock et al. 1986), using
the ow through an ori ce representation. This results in the following expression for the
exhaust manifold pressure pex:
1  q 
pex = 2 pim + pim , 4 m_ ex K (2 Tim + Te) ;
2 2 (1:4:26)

22
EXHAUST MANIFOLD AND TURBINE

2 1 compressor
intake intake absorbed 7
manifold manifold p cx/p amb 4 torque
temperature pressure
p ex
Mux f(u)
3
air mass p ex/p amb f(u)
flow rate apparent
42.5/u IVC to EVO
turbine
tau 2 1050 1.053.2 efficiency
5 + 3450 (f10)
+
fuel m ex dot
mass flow
FI to EVO rate {1pr.term} f(u)
flow 22.5/u
rate (f7)
tau 1
Ntc
f(u) corr KCpx

88.3/u
Mux
tau 3 IVO to EVO Mux f(u)
+ mt dot
+ *
8 T ex
+
F/A Mux Trq t
equiv eta t
ratio T e rise (f9)
(f6) f(u) 1/u
U/C 01.12 1/omega tc
(f8)
6 K 1/s K
engine 5002100 N tc omega tc 1/J tc
speed TC rotor
N eng 1 speed

Figure 1.11 : Exhaust manifold, turbine, and TC rotor dynamics model.

23
\link \center {\large \rm turbine\n corrected air mass flow rate}
f7
6

corrected air mass flow rate


5

4
turbine
3

0
3
3500
2.5
p 3000
ex /p 2 2500
am 1.5 2000
b 1500 speed
1 1000 Cr otor
cted T
corre

Figure 1.12 : Corrected turbine air mass ow rate (f7 ).

where K is an empirical constant which accounts for the port area and the dynamic e ects
in the manifolds. Equation (3.26) is being computed in the p ex block of gure 1.11.
Turbine model
Similarly to the compressor model, the mass ow rate and the eciency of the turbine
are determined from performance maps. The ow rate and the TC speed are corrected to
eliminate the variation due to di erent exhaust manifold conditions. After specifying the
corrected mass ow rate from the map of gure 1.12, the actual value can be obtained as
follows: !
p
m_ t = p f7 p ; p
ex N tc p ex
: (1:4:27)
Tex Tex amb
The two-dimensional look-up table in the ow rate block of gure 1.11 represents f7. The
actual mass ow rate is then computed in the mt dot block as in equation (1.4.27).
The turbine eciency t is a function of the blade tip speed vU=C , which is computed in
the U/C block of gure 1.11 as:
vU=C = q !tcrt ; (1:4:28)
2TexCpx
where rt is the turbine radius. The turbine eciency is then obtained via the one-dimensional
look-up table shown in gure 1.13:
t = f9(vU=C) ; (1:4:29)
24
1

0.9

0.8

0.7

0.6

turbine
0.5
efficiency
f9 0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1
turbine blade tip speed (m/s)

Figure 1.13 : Turbine eciency (f9).

in the turb e block of gure 1.11. As suggested in (Ledger et al. 1971), the apparent eciency
of the turbine app, which depends on the ratio pcx=pamb, is determined from experimental
data. The one-dimensional look-up table of gure 1.14 is used in the apparent turbine eciency
block to obtain: !
app = f10 p pcx
: (1:4:30)
amb
As we recall, the apparent eciency is used to account for the e ect of the pulse turbocharg-
ing method.
The torque produced by the turbine is then calculated:
2 ! ex ex,1 3
m_ C T  
Mt = t px ! ex t app 41 , pamb p 5: (1:4:31)
tc ex

The variance of the exhaust gas speci c heat Cpx and speci c heat ratio ex with changing
pressure and temperature is negligible. They are assumed to be constant like the correspond-
ing parameters for air in the compressor model. Equation (3.31) is represented by the Trq t
block in gure 1.11.
TC rotor dynamics model
In order to obtain the TC rotor speed, we use equation (1.4.3):
Mt , Mc = Jtc !_ tc :

25
2

1.5

apparent
turbine
efficiency
f 10 1

0.5
1 1.5 2 2.5 3
p cx /p amb

Figure 1.14 : Apparent turbine eciency (f10 ).

The turbine torque has been determined in equation (1.4.31) and the compressor torque
in equation (1.4.7). The di erence between these torques is divided by the TC e ective
inertia (actually multiplied by its inverse in the 1/J tc gain block). The result is the angular
acceleration, which is integrated in the omega tc block of gure 1.11 to obtain the TC rotor
speed. The gain block N tc performs the conversion from rad/s to rpm.
This concludes the description of the mathematical model of the TC diesel engine. The
complete model representation in SIMULINK is shown in gure 1.15.

1.4.3 Simulation results


Some simulation results are presented in gure 1.16. In this particular simulation, the
response of the system to given input signal pro les is being observed. The inputs to the
road vehicle diesel engine are the fuel command, represented by the index Y which is directly
related to the accelerator pedal command, and the torque absorbed by the transmission and
the drivetrain Mload. The pro les of the index and the load torque have been selected to
demonstrate the reaction of the system states to di erent input changes. First, at t = 12 s,
both input signals are decreased simultaneously, leaving the engine speed unchanged. The
decrease of the indicated torque (in the bottom diagram) due to the decreased fuel mass
is compensated for by the higher load torque. As a result, the net torque is zero and thus
the engine speed remains constant. The e ect of changing the load torque and the index
Y separately is shown next. At t = 23 s, the load torque alone is increased. This renders
the net torque negative, and hence the engine speed drops. At t = 40 s, the index alone is

26
101.325 p_amb DIESEL ENGINE MODEL
1/p_amb
1/u
* f(u)
p cx/p amb {pr.term1}
13.2
COMPRESSOR

2590k C flow CompEff (f2)


f(u)
f(u) Ntc corr rate (f1)
Trq c
Mux
Mux

Figure 1.15
mc dot f(u)
298.15
T_amb Mux f(u) 323.15
Tcx T_cool

Mux f(u) f(u)


mass flow 14 Eff IC INTERCOOLER
parameter (f3) Mux
f(u) Mux f(u)
Memory p cx T im
EXHAUST
MANIFOLD

27
287/.0
* R/Vim p ex
Mux f(u)
1/s +
INTAKE MANIFOLD
p im p ex/p amb f(u)

8333 Mux f(u) 42.5/u IVC to EVO app


* turb
Vd/(120*Vim) m im dot tau 2 1050 1.053.2 eff
Product + 3450 (f10)
f(u) Mux f(u) +
eta vol (f4) F/A m ex dot flow
KN eng FI to EVO rate {1pr.term} f(u)
* 22.5/u (f7)
mf dot tau 1 N tc
1 2 1.e
A/F stoichK f(u) corr
Eng Speed Fuel kf KCpx
Command K
Qlhv/k
Load 88.3/u
Trq f f(u) 1 01.05 Mux

: Complete model of TC diesel engine in SIMULINK.


Trq tau 3 IVO to EVO Mux f(u)
+ mt dot
* f(u) Mux + *
1/s omega eng Trq ind eta ind T ex
+ Sum Trq Prod f(Neng)*f5 +
Mux Trq t
Delay eta t
TORQUE T e rise
PRODUCTION f(u) (f9) 1/u
.33 f(u) tau i (f6)
U/C 01.12 1/omega tc
1/J eng (f8)
K 1/s K
5002100 N tc omega tc 1/J tc

TURBINE
increased, which causes an increase in the indicated torque and the engine speed.
Since we have not obtained experimental data yet, we built our simulation model based
on empirical data available in the literature. The results obtained using this model support
its validity. The response of the engine speed and the TC speed to changes in the fueling
rate and the load torque are as expected. Most of the other variables in the model also
behave accordingly.

28
-3
x 10 Index Y
3.4

3.2
m

2.8
0 10 20 30 40 50 60
time (s)
Load Torque
600

550
N.m

500

450
0 10 20 30 40 50 60
time (s)
Engine Speed
1750

1700
RPM

1650

1600

1550
0 10 20 30 40 50 60
time (s)
Indicated and Friction Torques
800

600
N.m

400

200

0
0 10 20 30 40 50 60
time (s)

Figure 1.16: Engine speed, indicated torque, and friction torque response to given pro les of fuel
command (index) and load torque.
29
30
Chapter 2
Transmission Modeling

2.1 Introduction
The development of fully automated highway systems, in which the driver is not part of the
control loop, would require only vehicles equipped with automatic transmissions (ATs) to
be considered. Hence, AT modeling is an integral part of any e ort to construct complete
vehicle models suitable for AVCS design.
Three types of ATs have been developed for commercial vehicles including buses and
trucks. The rst is the hydrodynamic automatic transmission (HDAT), which has the de-
sirable features of self-controllability between speed and load, durability and ease of mass
production, but the disadvantage of low fuel economy. The second is the hydrostatic au-
tomatic transmission (HSAT), which has the advantages of power utility as well as fuel
economy due to its continuously variable speed range, but presents some diculties in the
implementation of the control mechanism, durability, and manufacturing of oil-hydraulic el-
ements. The third is the mechanical automatic transmission (MAT), which is characterized
by high eciency and low cost due to its ecient use of a conventional gearbox and clutch,
but presents a problem in controlling the fully automatic gear shift with smooth clutch en-
gagement. Of these three, HDAT has been increasingly employed for commercial vehicles
following its great success in passenger cars.
While many detailed models of passenger car ATs have been developed (Cho and
Hedrick 1989, Runde 1986, Tugcu et al. 1986), models for heavy duty truck ATs do not
seem to be available in the literature. The fact that HDATs for passenger cars and heavy
duty vehicles obey the same principles of operation makes it feasible to modify existing
models for passenger car transmissions for use as truck AT models.
31
2.2 Hydrodynamic Automatic Transmission (HDAT)
2.2.1 Torque converter
Manual transmissions require a clutch to uncouple the transmission from the engine so that
gears can be shifted. In an AT, hydraulic pressure is used to control shifts. Instead of
owing through the clutch to the gears, engine power ows from the crankshaft to the torque
converter . The latter is lled with oil and acts as a uid coupling . It allows gear shifting
while power is owing through the transmission.
Fluid coupling
A uid coupling consists of two members as shown in gure 2.1. The driving member, called
the pump is attached to a drive plate on the engine crankshaft. The driven member, called
the turbine , is attached to the transmission input shaft. When the engine starts running,
the pump begins to rotate and the oil is set into motion: It leaves the pump and strikes the
vanes in the turbine, thus applying torque to the turbine. In a uid coupling, the torque
applied to the turbine can never exceed the torque applied by the pump.1

Pump
' $
Turbine

 ? ? 

? ?
   
rotation of rotation of
crankshaft transmission shaft
& %
Figure 2.1 : Fluid coupling in action.
When sucient torque is applied to the turbine, it begins to turn. As the turbine
approaches the speed of the pump, the e ective force of the oil on the turbine vanes is
reduced. If the two members turn at the same speed, no oil passes from one member to the
other, and respectively no power is transmitted through the coupling.
In order to reduce the undesired turbulent ow of oil in a uid coupling, a split guide
ring is centered in the members as shown in gure 2.2.
Torque converter
Fluid couplings transmit torque at maximum eciency when the pump and the turbine are
turning at close to the same speed. When the driving member turns much faster than the
driven members, then the eciency of any uid coupling is reduced because, as the pump
1 This is a consequence of the so called bounce-back e ect, described in detail in the next section.

32
Pump Turbine

Split-guide
ring

Figure 2.2 : Split-guide ring designed to reduce turbulence.

rotates, the oil is thrown onto the turbine vanes with considerable force. As the oil strikes
the turbine vanes, it splashes and bounces back toward the pump vanes, striking the oil
being thrown out of the pump vanes. When there is a big di erence in pump and turbine
speeds, much of the driving torque is used up in overcoming this \bounce-back" e ect. As
a result, torque is lost and there is a torque reduction through the uid coupling.
In order to overcome this e ect and at the same time increase the torque on the turbine,
a third member, called stator or reactor is added between the pump and the turbine. With
this addition the uid coupling becomes a torque converter.
The reactor is mounted on a one-way clutch, which permits the reactor to run free in
the same direction as the pump when both the pump and the turbine are turning at about
the same speed. However, when torque increase takes place, the oil still has great forward
motion as it bounces back and leaves the turbine. This causes the reactor to try to turn
backward and it locks to the reactor shaft via the one-way clutch. Now the reactor vanes
turn the oil from the trailing edges2 of the turbine blades into a \helping" direction before
the oil reenters the pump. Thus torque multiplication takes place.
As the turbine speed approaches the pump speed, the torque increase gradually drops
o until it becomes 1:1. At this point, the oil begins to strike the back faces of the reactor
vanes so that the reactor begins to turn. The torque converter acts simply as a uid coupling
under these conditions.

2.2.2 Transmission mechanicals


Planetary gears are an essential component of automatic transmissions. A simple planetary
gearset is shown in gure 2.3. In the center is the sun gear . Meshing with the sun gear
are two or more planet-pinion gears . They are mounted on shafts that are part of a plate
called the planet-pinion carrier . The carrier is also mounted on a shaft, so that the shaft
2 Trailing edges are the edges of the vanes that the oil passes last as it leaves a member.

33
HH
Y
H
HH
ring H planet
gear H
j
H 6 sun , pinions
gear ,
,
,
,
,

?

,
,

Figure 2.3 : Planetary gearset.

can rotate, thereby carrying the planet pinions around in a circle. The internal gear also
meshed with the pinions is called ring gear . Each part of the planetary gearset is called a
member . Thus, the three members are:
 the sun gear;
 the planet-pinion carrier; and
 the ring gear.
A planetary gearset can provide any of the following ve operating conditions:
1. overdrive (speed increase and torque decrease);
2. reduction (speed decrease and torque increase);
3. lockup (direct drive);
4. neutral (no torque transmitted); and
5. reverse (again speed decrease and torque increase).
All these di erent conditions (except neutral) are made possible by applying the input torque
to a di erent planetary member while holding one of the other two members stationary.
Allowing all of the members to spin freely provides neutral. Locking any two of the members
together causes the entire planetary gear system to be locked and thus produces direct drive
through the gearset.
Gear equations. The angular velocities of the members of a planetary gearset can be
related to each other as follows. Let NS, NR, and NPL be the circumferences of the sun, the
ring, and the planet pinions respectively. They are proportional to the number of teeth on
the gears and can be expressed as
NS = 2rS; NR = 2rR ; (2:2:1)
34
where rS and rR are the radii of the sun and the ring respectively. Then the radius of a
planet pinion is
rPL = rR ,2 rS = NR 4, NS ) NPL = 2rPL = NR ,2 NS : (2:2:2)
The angular velocity of the planet pinion carrier is:
!C = r +vCr = N4v C
; (2:2:3)
S PL S + NR
where vC is the tangential velocity of the pinion center.
 Let us rst consider the case where the carrier is still, i.e., its angular velocity is !C = 0.
vR '$
q
q carrier is still
q - vS
&%

Then we have
NS ! :
vR = ,vS ) !RrR = ,!SrS ) !R = , N (2:2:4)
S
R
Hence, if the sun is the input and the ring is the output, we have reverse with speed reduction,
since NS < NR.
 Next, consider the case where the ring is still, i.e., !R = 0.
'$ring is still
q -v C
q -v
&% PL = vS

Then we have
vS = vPL = 2vC ) !C = N 2v S
= 2rS!S = NS ! ) ! < ! : (2:2:5)
NS + NR NS + NR S C S
S + NR
Hence, if the sun is the input and the planet pinion carrier is the output, we have reduction.
If the planet pinion carrier is the input and the sun is the output, we have overdrive.
 Now consider the case where the sun is still, i.e., !S = 0.
'$ q - vPL = vS
q - vC
&%
sun is still

35
Then we have
vR = vPL = 2vC ) !C = N2v R
= 2rR!R = NR ! ) ! < ! : (2:2:6)
NS + NR NS + NR R C R
S + NR
Hence, if the ring is the input and the planet pinion carrier is the output, we have reduction
(but with a di erent ratio than in the previous case). If the planet pinion carrier is the input
and the ring is the output, we have overdrive (again with a di erent ratio than before).
Finally, if none of the members is held still, the combination of the above three cases
results in vC = 12 (vS + vR):
vR '$
q
 qv-
C
-

&%
q - vS
Substituting into equation (2.2.3) yields:
!C = 2N(vS++NvR) = NSN!S ++ NNR!R = N N+SN !S + N N+RN !R : (2:2:7)
S R S R S R S R

From (2.2.7) we obtain the previous equations as special cases:


1. !C = 0 ) !R = , NNRS !S ;
2. !S = 0; !C = NSN+RNR !R ) !C < !R ;
3. !R = 0; !C = NSN+SNR !S ) !C < !S ;
4. !C > !R ; NS
NS +NR!S = !C , NSN+RNR !R > !C (1 , NSN+RNR ) = NSN+SNR !C ) !S > !C > !R ;
5. !C > !S ; NSN+RNR !R = !C , NSN+SNR !S > !C (1 , NSN+SNR ) = NSN+RNR !C ) !R > !C > !S .
Many planetary gearsets used in ATs are simple planetary gearsets. Other ATs are
equipped with compound planetary gears. They consist, for example, of a sun gear, two sets
of planetary gears with carriers, and two ring gears.
In ATs, drums , bonds , and clutches are used to control planetary gears and cause a
change of gear ratio through the transmission.

2.3 Automatic Transmission Modeling


2.3.1 Torque converter modeling
A static nonlinear terminal model of the torque converter has been derived and justi ed by
Kotwicki (1982). His approach is to start from simple physical models which capture the
36
-
6
?
Pump Q Turbine
rPx 6

6
z z z MT rTx
? ?

!P MP !
:
T
-
?

6


Figure 2.4: Fluid coupling model.

structure of the torque converter and then recast them into the form:
MT = f1(!P; !T )
MP = f2(!P; !T ) ;
which are the desired explicit terminal relations between torques and angular velocities of
the torque converter turbine and pump respectively.
Torque converters and uid couplings are based on the simple physical principle that the
rate of change of angular momentum of the uid passing through a stationary or rotating
element of a turbomachine must equal the torque applied to that element. Rotational angular
momentum transferred to circulating uid by the blades of a pump is carried by the uid to
a turbine, where it is absorbed by the turbine blades to produce output torque.
Torque and power in a uid coupling
Let us consider the uid coupling of gure 2.4, operating at constant speed. The applied
torque MP rotates the pump and induces an annular volume ow rate Q through the pump
and turbine elements. The tangential velocity at the pump exit is equal to the product of the
angular velocity !P and the pump exit radius rPx. This tangential velocity times the pump
exit radius is the tangential angular momentum per unit mass of the uid leaving the pump.
Multiplication by uid density and volume ow rate gives the exit angular momentum ow
rate of the pump L_ Px = QrPx 2 !P . Similarly, at the turbine exit L_ Tx = Qr2 !T . The rate
Tx
at which angular momentum enters the turbine is the rate at which it leaves the pump. The
di erence gives the rate of change of angular momentum across the turbine and the turbine
torque:
MT = L_ Tx , L_ Px = Q(!T rTx2 , ! r2 ) :
P Px (2:3:1)
37
A similar argument gives the pump torque:
MP = L_ Px , L_ Tx = Q(!P rPx
2 , ! r2 ) :
T Tx (2:3:2)
Therefore, MT = ,MP. The pump input power is the product of pump torque and angular
velocity:
PP = MP!P = Q!P (!PrPx2 , ! r2 );
T Tx
which is positive, indicating power ow into the uid coupling. Respectively, the turbine
input power is negative, indicating power ow from the uid:
2 , ! r2 ):
PT = MT!T = Q!T (!TrTx P Px

Although angular momentum has been conserved, mechanical power has not. Summing
pump and turbine input powers gives the rate at which the coupling dissipates energy under
steady state conditions:
2 , ! ! r2 + ! 2 r2 , ! ! r2 ):
PP + PT = Q(!P2 rPx P T Tx T Tx P T Px

This dissipation is zero only when pump and turbine rotate at the same speed:
!P = !T ) !P2 = !T2 = !P !T ) PP + PT = 0 :
In order to obtain the relation in the desired explicit form, we need to solve for Q in
terms of !P and !T . We are using the fact that for steady ow conditions, the energy added
to a uid particle in a complete circle around the coupling must be zero. Flow energy and
losses are commonly described in terms of total pressure:
PP + PT = p + p = (!2 r2 , ! ! r2 + !2 r2 , ! ! r2 ); (2:3:3)
P T P Px P T Tx T Tx P T Px
Q Q
where pP + pT is the net total pressure added from external inputs.
Total pressure losses around the ow circuit are due to two sources:
 uid friction, and
 shock loss.
Fluid friction is proportional to the square of the volume ow rate:
pf = 12 Cf Q2: (2:3:4)
Shock loss is caused by the sudden change in velocity as uid crosses the boundary between
one blade system and the next. This energy loss is similar to the total pressure loss in pipe
ow due to sudden expansions or contractions of the pipe cross section. The shock loss

38
is proportional to the square of the di erence in rotational velocities. Summing the shock
losses due to the two transitions the uid makes around the coupling, we obtain:
ps = psP,T + psT,P = 12 (rPx2 + r2 )(! , ! )2:
Tx P T (2:3:5)
Adding the pressure inputs, subtracting the pressure losses, and requiring that the total
pressure added be zero for steady state operation, results in the following expression:
pP + pT = pf + ps : (2:3:6)
Substituting (2.3.3){(2.3.5) into (2.3.6) yields:
2 , ! ! r2 + ! 2 r2 , ! ! r2 ) = 1 [C Q2 + (r2 + r2 )(! , ! )2 ]:
(!P2 rPx P T Tx P T Px
T Tx 2 f Px Tx P T

Hence,
2 , ! ! r2 + ! 2 r2 , ! ! r2 = 1 C Q2 + 1 ! 2 r2 , ! ! r2 + 1 ! 2 r2
!P2 rPx P T Tx P T Px
T Tx 2 f 2 P Px P T Px 2 T Px
+ 12 !P2 rTx
2 , ! ! r2 + 1 ! 2 r2 ;
P T Tx
2 T Tx
which after some algebraic manipulation is reduced to
2 , r2 )(! 2 , ! 2 ) = C Q2 :
(rPx Tx P T f

Then for Q we obtain: s 2 q


Q = rPx C, rTx !P2 , !T2 ;
2
(2:3:7)
f
and the explicit relations (2.3.1){(2.3.2) for the torques become:
s2 2 q 2 
MP = ,MT =  rPx C, rTx !P2 , !T2 rPx !P , rTx
2 ! :
T (2:3:8)
f

E ect of curved blades. If the blade entrance or exit angles di er from zero, the expres-
sions for tangential velocity directly involve the ow rate, because the blade can convert uid
ow rate into additional tangential velocity in the plane of rotation of the pump or turbine.
As we can see from gure 2.5, the expression for the tangential velocity at the pump exit
vPx
t becomes
t = Q tan + ! r ;
vPx (2:3:9)
Px P Px
APx
where APx is the pump exit area and Px is
the blade angle at the pump exit.
The new angular momentum ow equation is:
 
t r = Q Q tan + ! r
_LPx = QvPx Px Px P Px rPx :
APx
39
Px = 0 Px 6= 0 Q tan
Px
Q Q APx
APx APx Px
Q
!P rPx !P rPx APx !P rPx
Flat Blade Curved Blade

Figure 2.5 : Blade angle e ect on tangential velocity at the pump exit.

The change of tangential velocity from the pump exit to the turbine entrance becomes
 
vt = vPxt , v t = Q tan Px , tan Te + ! r , ! r ;
Te P Px T Te
APx ATe
where Te and ATe are respectively the blade angle and the area at the turbine entrance.
The shock loss in total pressure becomes:
   2
psP,T = 12  Q tanA Px , tanA Te + !PrPx , !T rTe : (2:3:10)
Px Te

Torque and power in a torque converter


In the torque converter case, uid leaving the turbine element is de ected by the reactor so
that at the reactor exit, uid tangential velocity is in the direction of the pump rotation.
Since the pump angular momentum ow L_ Px is the sum of input from the reactor and that
added by the pump, the turbine sees a larger angular momentum ow than that due to the
pump alone. This e ectively multiplies the pump input torque.
Since the reactor does not rotate, it produces no work. The reactor is typically mounted
on a one-way clutch so that it can only add, not subtract torque. When the entrance and exit
momentum ows of the reactor are equal, it freewheels and the torque converter operates as
a uid coupling.
A schematic diagram of a torque converter is shown in gure 2.6. The expressions for
the pump and turbine torque are:
 
MP = !P rPx 2 Q +  rPx tan Px , rRx tan Rx Q2 (2.3.11)
APx  ARx 
MT = (!T rTx , !P rPx)Q + 
2 2 r Tx tan Tx rPx tan Px
, Q 2: (2.3.12)
ATx APx
40
APx -  ATe
6 Q
P  T
APe HH ATx
rPx = rTe 
HH
HH
R 
j 


 OC
z MT
6
z z  C rTx = rRe =

 C
C
= rPe = rRx
?  C ?

!P MP 
 C
C !T :
ARx ARe
Figure 2.6 : Torque converter diagram.

The net pressure equations then become:


 
pP = PQP = MQ P !P 2 + ! rPx tan Px , rRx tan Rx Q
= !P2 rPx P
APx  ARx
P M T !T r Tx tan Tx rPx tan Px

pT = Q = Q = !T (!T rTx , !PrPx) + !T
T 2 2
A , A Q:
Tx Px
The uid friction loss is again pf = 12 Cf Q2, and the shock losses are:
   2
psP,T = 12  Q tanA Px , tanA Te + !P rPx , !T rTe
1   tan Px

Te
Tx tan Re
 2
psT,R = 2  Q A , A + !T rTx
Tx Re
1   tan Rx tan Pe  2
psR,P = 2  Q A , A , !PrPe :
Rx Pe
Adding the pressure inputs, subtracting the pressure losses, requiring that the total
pressure added be zero for steady state operation, and using the identities rPx = rTe, rTx =
rRe = rPe = rRx, yields the following equation for Q:
IQ2 + (H!T , G!P )Q , (E!P2 + F!T2 ) = 0 ;
where
"   tan Tx tan Re 2  tan Rx tan Pe 2 #
1 tan
I = 2 A , A Px tan Te 2
+ A , A + A , A + Cf
Px Te Tx Re Rx Pe
 Te , rRe tan Re 
H =  rTe tan
ATe ARe 
 rTe tan Te , rPe tan Pe
G =  ATe APe
41
E = 21 (rPx
2 , r2 )
Pe

F = 12 (rTx
2 , r2 ) :
Te

Solving for the volume ow rate Q, we obtain:


q
(H!T , G!P )2 + 4I (E!P2 + F!T2 )
Q = ,(H!T2I, G!P ) + 2I : (2:3:13)
Substituting (2.3.13) into (2.3.11){(2.3.12) the explicit equations for MP and MT can be
derived. Although the resulting expressions are too complex to be applied directly in control
analysis, they can be used as the basis for extracting a simpli ed model.
To derive an approximate static terminal model of the torque converter, one can approx-
imate the expression for the volume ow rate Q in (2.3.13) by performing a Taylor series
expansion in the variable !T about !T = n!P , where n can be any constant. If !T = !T ,
the ow rate varies linearly with pump speed:
q
(Hn , G)2 + 4I (E + Fn2) , (Hn , G)
Q(!P; !T ) =

2I !P : (2:3:14)
In order to simplify the evaluation of derivatives in the Taylor series expansion, we choose
n = HG , which reduces (2.3.14) to:
s 2
Q(!P; !T ) = EH IH
 + FG2 ! :
2 P

Performing the expansion about !T = !T yields:


@Q (! , ! ) + 1 @ 2Q (! , ! )2 +   
Q(!P; !T ) = Q(!P; !T ) + @! T T 2 T T
s 2 T 2 @! T

= EH IH + FG2 !
2 P
0 1 
H
+ @, 2I + q FG A !T , G !P + O(!T , !T )2 :
I (EH 2 + FG2) H
Actual torque converter data justi es a rst order approximation:
Q~ 1(!P ; !T ) = !P + !T : (2:3:15)
Using equation (2.3.15) in the torque expressions (2.3.11){(2.3.12) makes the terminal equiv-
alents take on quadratic forms in the vector of speeds (!P ; !T)T :
MP = a0!P2 + a1!P!T + a2!T2
MT = b0!P2 + b1!P !T + b2!T2 : (2.3.16)
42
These expressions with suitable coecients provide a simpli ed analytical model for the
torque converter in converter mode, where torque multiplication takes place. Because the
approximated equations do not change in coupling mode, suitable adjustments of parameters
allow use of these forms for the description of the converter in coupling mode too.
The approximate model (2.3.16) has been used as the basis for regression ts of actual
torque converter data. Kotwicki (1982) obtained excellent results by tting data from a
variety of converters. Therefore, the approximate model (2.3.16) proves to be suitable for
control design. The described approach has been adopted by Cho and Hedrick (1989) for
the AT subsystem of their vehicle model.
The HDAT used in buses and heavy duty trucks has the same structure and principles of
operation as the typical light duty vehicle HDAT (Tuck et al. 1978). Therefore, the approach
suggested by Kotwicki and employed by Cho and Hedrick for their car model, can be readily
applied in the heavy duty vehicle model once actual torque converter data are available.

2.3.2 Transmission mechanicals modeling


Several di erent approaches have been explored for modeling the AT mechanicals. In the
lever analogy method presented by Tugcu et al. (1986), each gear set is represented by a
single lever. The torques acting on the gears are represented by forces on the levers. The
gears and shafts are assumed to be rigid, and no backlash in the gearsets is accounted for.
The force and moment balance equations of motion for the planetary gearset are derived
from the FBD of the lever system.
The bond graph method (Thoma 1990, Blundell 1982) was used by Runde (1986) to
model the kinematics of transmission mechanicals. Cho and Hedrick (1989) have adopted
this approach in their vehicle model. The reasons that make it more desirable for control
are:
 the states are directly associated with physical variables, and
 the possible gear shaft compliances and backlash e ects can be easily included.
The schematic diagram for the Hydramatic 440 AT modeled by Runde (1986) is shown
in gure 2.7.
Let us now see how bond graphs can be used to model the transmission mechanicals.
First we consider the ideal planetary gearset of section 2. Its schematic diagram is
R

C
S
and its equations are repeated here for convenience:
!C = N N+SN !S + N N+RN !R = RS!S + RR!R
S R S R

43
'$ M e
P T

,,
,,,
, &%
,,R@@
,,
,,,
,,
,
,,
,,,
,,
, B12
C2 BR ,,
,,,
,,
,,
,

C1 C3
,,
,,,
,,,,
,

C4 @@,, @ , M
s

Figure 2.7 : Schematic diagram of Hydramatic 440 .

MS = N N+SN MC = RSMC
S R
N
MS = N +SN MC = RSMC :
S R
This gearset can be described by the following bond graph:
!S !R
1 TF
, ..
0 @ TF
.. 1
MS , ,, @ @ MR
@
RS !C MC R R

,
,
1
A compound planetary gearset , with schematic diagram

Input Reaction
Planetary Planetary

44
is described by the following equations:
!CI = RSI!SI + RRI!RI
!CR = RSR!SR + RRR!RR :
These equations, combined with the kinematic constraints:
!CI = !RR
!RI = !CR ;
result in a system of only 2 independent equations with 4 unknowns:
!RR = RSI!SI + RRI!CR
(2:3:17)
!CR = RSR!SR + RRR !RR :
Therefore, the compound planetary gearset can be represented by:
!SR !RR
1 , ..
,
TF ,
,
0 @@
TF
.. @
@
1
RSR !CR RRR ,
,
,
,
,
, ,
,
1 ,
,

ll ,
,
,
, l
l
1 l
l
@ l
@ l
!CI l
l
ll

1 TF
.. 0 TF
.. 1
!SI ,
, ,
, @
@ @
@ !RI
RSI RRI

which is equivalent to:


!RR = !CI !SR
1 ,
,
TF
.. ,
,
0 @
@
TF
.. @
@
1
@
@ RRR RSR
,
,
1 TF
.. 0 @ TF
.. @ 1
!SI ,
, ,, @ @ !RI = !CR
RSI RRI
This bond graph shows that the ring structure will create an algebraic loop when the system
45
equations are solved. To avoid it, one can use the following representation, which uses the
multiport junction PL:
!RR !SR
@ MRR @
, MSR
@ , ,
,
@
@ ,
PL
, .. @
!SI ,, N @@ !CR
@ MSI
@ MCR ,
,

Depending on the causality in the junction PL, di erent values for the matrix N can be
found from (2.3.17). For example, in the expression
" # " #
!SR = A !SI ;
!RR !CR
the matrix A corresponds to the following causality:
!RR !SR
@
@ @ , ,
,
@ ,
@, ,
, PL
.. @
@
@
,
! SI ,
, A @
@
!CR
@
@ ,
,

Then, A can be determined from (2.3.17) as:


2 3
, R RR RSI 1 , R RR R RI
A = R1 64 75 :
SR RRIRSR RRIRSR
By power conservation:
h i " MSI # h i " MSR #
!SI !CR M + !SR !RR M = 0;
CR RR
or h i " MSI # h i T " MSR #
!SI !CR M = , !SI !CR A M ;
CR RR
which implies: " # " #
MSI = ,AT MSR :
MCR MRR
Adding the inertias and torques to the bond graph results in:

46

0.1 ? 


0.05 ?

?` ` ` `
0 40 80 120 slip speed, ! (rad/s)
Figure 2.8 : Actual coecient of friction.

IRR MBR ISR


A  A
A  A
A  A
1 1  MB12
@ ,
@ ,
@, ,
PL
. .
@
@,
, @
,
,
A @
@,
1@
@ ,1

 A 
 A 
 A 
ISI MC4 ICR

The stationary clutch torque applied to the respective inertia opposes its motion. The
lumped inertias of the reaction ring, input carrier, and reverse reaction band are included
in IRR. The input sun, the C 4 clutch, and both sprag clutches inertias are combined in ISI.
The reaction carrier, input ring, and nal drive gear inertias are the components of ICR. The
clutch torques for plate clutches are calculated from:
MCi = P(!Ci )ACi RCi sign (!Ci ); (2:3:18)
and for bonds from:
MB = PABRB(e(!B)B , 1) sign (!B); (2:3:19)
where P is the clutch pressure, (!) is the coecient of friction, ACi;B are respectively the
clutch plate or the band piston area, RCi;B are the e ective radii of clutch plate or band, and
B is the band wrap angle. As pointed out in (Tugcu et al. 1986), the coecient of friction
(!) follows the solid curve in gure 2.8, rather than the dotted line which indicates the
static friction coecient.

47
Ir MC2 IRR MBR ISR
A A A  A
A A A  A
MT A A A  A

!T  1 0 1 1
   MB12
@ ,
@ ,
@ ,
PL
, .. @
,
,
A @@ Ms
 0  1 1  TF
.. !o



 A
A 
 RD
  A 
MC1 + MC3 ISI MC4 ICR

Figure 2.9 : Complete bond graph diagram of the transmission mechanicals of Hydramatic 440 .

gure 2.9 contains the complete bond graph diagram of the transmission mechanicals.
The inertias of the torque converter turbine, chain assembly, and C 1 and C 3 clutch housings
are combined in IT. The inverse of the nal drive gear ratio is RD. The inputs to the
mechanical section of the transmission are the torques MT , Ms and the clutch pressures.
The output is the transmission output speed !o. The clutch slip speeds are determined from
gure 2.9 as follows:
!C4 = !SI
!B12 = !SR (2.3.20)
!BR = !RR
based on the fact that at \1" junctions ows are equal, and
!C1 = !C3 = !T , !SI
!C2 = !T , !RR ; (2.3.21)
since at \0" junctions the sum of ows equals zero.
Before the system equations can be found from the bond graph, the causal assignment
must be made. The causality of each clutch torque will depend on its mode of operation.
If the clutch is slipping, then the clutch is a torque source and the torque is found from
(2.3.18){(2.3.21). If, however, the clutch is locked up, then the clutch is a speed source
and the clutch slip speed is zero. The reaction torque of the clutch in locked-up mode is
determined via another bond graph.

48
For a locked up clutch, the magnitude of the clutch torque as de ned in equation (2.3.18)
or (2.3.19) must be greater than the clutch reaction torque as de ned by the system equations.
If this is not true, then the clutch will begin slipping. The causality of the clutch will therefore
change from a speed source to a torque source whose torque is found from (2.3.18){(2.3.21).
A slipping clutch can only go into lock-up mode when the clutch slip speed numerically
passes through zero. When this happens, the system equations corresponding to the clutch
in its locked-up mode are used.

2.4 HDAT for Heavy Duty Vehicles


As mentioned earlier, AT models for heavy duty vehicles are not readily available. However,
the analysis of such transmissions (Tuck and Smith 1978) shows that they obey the same
operation principles as passenger car transmissions. Even though the parameters of the
torque converter and the mechanicals are di erent, but one can use the same modeling
approach if provided with torque converter data and gear speci cations.
A detailed description of 3 representatives of the Detroit Allison's AT-MT-HT family of
automatic transmissions is presented in (Tuck and Smith 1978). One 4-speed and two 5-
speed ratio arrangements are discussed. As in the modeled Hydramatic 440 , the mechanical
part of the transmission consists of planetary gearsets, clutches, and bands. Hence, the
bond graph method presented in the previous section can be used for deriving the system
equations for the transmission mechanicals.

49
50
References
Blundell, A. J. 1982. Bond Graphs for Modelling Engineering Systems . New York: Halsted
Press.
Cho, D., and J. K. Hedrick. 1989. Automotive power train modeling for control. ASME
Transactions, vol. 111, pp. 568{576.
Dobner, D. J. 1980. A mathematical engine model for development of dynamic engine
control. SAE Transactions, paper number 800054, pp. 373{381.
Flower, J. O., and R. K. Gupta. 1974. Optimal control consideration of diesel engine
discrete models. International Journal of Control , vol. 19, no. 6, pp. 1057{1068.
Hendricks, E. 1986. A compact, comprehensive model of large turbocharged, two-stroke
diesel engine. SAE Transactions, paper number 861190, pp. 4.820{4.834.
Hendricks, E. 1989. Mean value modelling of large turbocharged two-stroke diesel engine.
SAE Transactions, paper number 890564, pp. 1{10.
Horlock, J. H., and D. E. Winterbone. 1986. The Thermodynamics and Gas Dynamics of
Internal Combustion Engines . Oxford: Claredon Press.
Jennings, M. J., P. N. Blumberg, and R. W. Amann. 1986. A dynamic simulation of Detroit
diesel electronic control system in heavy duty truck powertrains. SAE Transactions,
paper number 861959, pp. 5.943{5.966.
Jensen, J. P., A. F. Kristensen, S. C. Sorenson, and E. Hendricks. 1991. Mean value mod-
elling of a small turbocharged diesel engine. SAE Transactions, paper number 910070,
pp. 1{13.
Kao, M., and J. J. Moskwa. 1993. Turbocharged diesel engine modeling for nonlinear
engine control and state estimation. Proceedings, ASME Winter Annual Meeting,
DSC-vol. 52. Symposium on Advanced Automotive Technologies: Advanced Engine
Control Systems. New Orleans LA.
Kotwicki, A. J. 1982. Dynamic models for torque converter equipped vehicles. SAE Trans-
actions, paper number 820393, pp. 1595{1609.
Kullkarni, M. M., R. P. Sinha, and H. C. Dhariwal. 1992. A reduced order model of
turbocharged diesel engine. Proceedings, American Control Conference, Chicago IL,
pp. 927{931.
Ledger, J. D., R. S. Benson, and N. D. Whitehouse. 1971. Dynamic modelling of a tur-
bocharged diesel engine. SAE Transactions, paper number 710177, pp. 1{12.
Lilly, L. C. R. 1984. Diesel Engine Reference Book . London: Butterworths and Co.
The MathWorks, Inc., 1993. SIMULINK User's Guide . Unpublished.
McMahon, D. H., J. K Hedrick, and S. E. Shladover. 1990. Vehicle modelling and control
for automated highway system. Proceedings, American Control Conference, San Diego
CA, pp. 297{303.

51
Runde, J. 1986. Modeling and Control of an Automatic Transmission. S.M.M.E. thesis,
Massachusetts Institute of Technology, Cambridge MA.
Thoma, J. U. 1990. Simulation by Bondgraphs: Introduction to a Graphical Method . New
York: Springer-Verlag.
Tuck, R. M., and R. Smith. 1978. Automatic transmissions for commercial and public
service vehicles. Inst. Mech. Eng. Conference Publications. Automatic and Semi-
automatic Gearboxes for Heavy Commercial Vehicles Conference, Solihull GB.
Tugcu, A. K., K. V. Hebbale, A. A. Alexandridis, and A. M. Karmel. 1986. Modeling
and simulation of the powertrain dynamics of vehicles equipped with automatic trans-
mission. Proceedings, ASME Winter Annual Meeting. Symposium on Simulation of
Ground Vehicles and Transportation Systems. Anaheim CA.
Winterbone, D. E., C. Thiruarooran, and P. E. Wellstead. 1977. A wholly dynamic model of
turbocharged diesel engine for transfer function evaluation. SAE Transactions, paper
number 770124, pp. 1{11.

52

You might also like