In Proc.

of IEEE IAS Annual Meeting, Pittsburgh, Oct 11-19, 2002

Engine Torque Ripple Cancellation with an Integrated Starter

Alternator in a Hybrid Electric Vehicle: Implementation and Control
Roy I. Davis

Robert D. Lorenz

Associate Member, IEEE

Ballard Electric Drives and Power Conversion
15001 Commerce Drive North
Dearborn, MI 48121
rdavis9@ford.com

Fellow, IEEE
Departments of Mechanical Engineering
And Electrical and Computer Engineering
University of Wisconsin, Madison
Madison, WI 53711
lorenz@engr.wisc.edu

Abstract Integrated starter alternators (SA) allow greater

electrical generation capacity and the fuel economy and
emissions benefits of hybrid electric automotive propulsion.
The drive usually comprises a field-oriented induction motor or
vector-controlled permanent magnet motor coupled to the
crankshaft of the engine directly or by a belt. These drives have
high bandwidth torque control and may be used for active
cancellation of the significant ripple torque inherent to internal
combustion engines. This paper presents an analytical model
useful for the analysis of the combustion engine torque and
describes a control system that uses observer-based high fidelity
torque feedforward and engine speed feedback to generate a
disturbance input decoupling torque command for the SA to
cancel the engine torque ripple.

II. INTERNAL COMBUSTION ENGINE MODEL

An indirect injection four-cylinder diesel engine was used
for this research. A novel engine model was developed
consisting of a method for generating the combustion
pressure pulses for each of the four cylinder pressures, a
simplified representation of the slider-crank kinematics to
convert the pressure pulses to indicated torque, and a timevarying (or crankangle-varying) inertia expression to
represent the translating pistons as effective rotational
inertias. Fig. 1 shows an overall diagram of the engine,
clutch, and vehicle model, assuming the four cylinder
pressures P1() P4() as inputs, with the engine and vehicle
motion states as outputs. The indicated torque, MICE, as a
function of crankangle () can be expressed as

Index Terms Observer, integrated starter alternator,

engine torque, engine model, torque cancellation, active
flywheel, hybrid vehicle.

I. INTRODUCTION
Integrated starter alternators (SA) provide greater electrical
generation capacity and improve the fuel economy and
emissions of hybrid electric vehicles [1]. The integrated
starter alternator (SA) is coupled to the combustion engine
either directly or by a belt. In addition to enabling the fuel
economy and emissions improvements, the availability of the
SA to perform other functions provides further benefit to the
vehicle. One such function is active crankshaft torque ripple
cancellation, in which the SA machine is controlled as a
torque actuator to reduce or eliminate the ac content of the
torque produced by the impulsive cylinder pressures in the
engine [2, 3]. This is sometimes referred to as active
flywheel, although it does much more than a flywheel could
do.
This paper presents a "disturbance input decoupling"
control method using a highly accurate observer that
combines feedforward and feedback signals to synthesize the
torque control signal for the SA effectively to cancel the
ripple torque of the engine. The paper also describes a novel
internal combustion engine (ICE) model, the ac torque
observer used to create the SA control signal, and test results
that illustrate the system performance.

r
r
MICE () = k1 k2 ( 1.0+ cos()) - k3 ( 1.0- cos())
L
L

where
k1 =

(1)

B2rsin()
; k2 = P1()+P3(); k3 = P2()+P4();
4

and B is the cylinder bore, r is the crank radius, and L is the

connecting rod length.
MSA

P1 + P4
P2 + P3

Kinematics

Friction
Model

M ICE

_
MSHAFT

_ _

Jf + J( )

1
S

1
S

Variable
Inertia
1 J()
2

+
+

MCLUTCH

bcl

+
+
kcl
1
JV

Clutch
Model

_
1
S

1
S

_
M ROAD

Vehicle
Load

Fig. 1. Combined Engine, Clutch Damper, and Vehicle Model

It is important to include the crankangle-varying engine

inertia J() to properly model the dynamics of the crankshaft
torque.
The nonlinear geometry of the slider-crank
mechanism shown in Fig. 2 gives rise to two effects: an
inertia which is a function of the angle and a LaGrangian
term proportional to the partial derivative of the inertia and
the square of crankshaft speed [4]. These expressions are
given in (2) and (3), and are shown schematically in the
block diagram of Fig. 1. Depending on the relative
contribution, a friction torque term can also be included if
desired.
mrecr4
mrecr4
- 2mrecr2cos(2) cos(4)
2L2
2L2

J()
2mrecr4
= 4mrecr2sin(2) +
sin(4)
L2

(2)
(3)

where mrec is the total reciprocating mass.

With an eye toward real-time implementation of this
model as an observer, as described in the next section, a
simple but accurate method is required to compute the
impulsive cylinder pressures. The torque ripple that is
generated by the pressure pulses is considered here as a
disturbance input. Thus it is important to represent the
pressures as accurately as possible without relying on
complicated derivations of gas dynamics and chemical
thermodynamics [5-8].
The combustion pressure waveform can be accurately
modeled by considering the underlying mechanisms at work
in the cylinder. That is, by adding a motoring pressure
waveform due simply to the change in cylinder volume as the
crank rotates (without combustion), to an incremental
pressure waveform produced by combustion of the fuel. The
cylinder volume as a function of crankangle and the
resulting motoring pressure can be computed from
B
4

Motoring Prototype
Waveforms

Pzero14()
Pzero23()

RPM
FLVR

p23()

Motoring
Factor

c1

Firing
Factor

c2

P2()+P3()

+
+

Offset Pressure

P1()+P4()

p14()
Firing Prototype
Waveforms

p0(N)

Fig. 3. Simulation Block Diagram for Final Cylinder Pressure Model

( r + L - rcos()-

L - r sin ())

(4)

10

(5)

n
V()
(Vd+Vc))

12

P(0)

P() =

(6)

A block diagram showing this computation is given in Fig. 3.

Figure 4 shows very good matching of measured and
simulated pressure waveforms using this model.

where Vc is the cylinder clearance volume and Vd is the cylinder

displaced volume. P(0) is the intake manifold pressure.

12
1000 RPM - 10% FLVR

0
0

90

180
..

270

360

12
10

12

2500 RPM - 10% FLVR

0
0

Fig. 2. Geometry of Slider Crank Mechanism

90

180
.

270

90

180
..

270

360

2500 RPM - 50% FLVR

10

+ xp

1000 RPM - 30% FLVR

10

V() = Vc +

p(, N, FLVR) = pzero()c1(N) + p() c2(N, FLVR) + po(N)

J() = 2mrecr2 +

The motoring pressure waveform will differ only in

amplitude as a function of engine speed. By normalizing
these waveforms, a motoring prototype waveform, pzero(),
and a scaling factor c1 that is a function of speed, N, can be
found from experimental motoring data. The component of
cylinder pressure due to combustion is very difficult to
determine analytically. By capturing cylinder pressure
waveforms while the engine is firing at different fueling
levels (FLVR), and subtracting the appropriately phased
motoring waveform for each cylinder, a series of incremental
pressure waveforms due to the combustion of a known
amount of fuel is collected and analyzed. This analysis leads
to a similar firing prototype waveform, p(), and a scaling
factor c2 that is a function of speed, N, and fueling level,
FLVR. These prototype waveforms and scaling factors are
combined to compute individual cylinder pressure
waveforms at any speed and any fueling level as

360

90

180
.

270

360

Figure 4. Simulated [ _ ] & Measured [] Pressures (MPa) vs. Crankangle

III. ENGINE TORQUE OBSERVER DESIGN

300 VDC

To implement the "disturbance input decoupling" control,

a measurement or estimate of the torque ripple is required. It
is impractical due to expense and other issues to install a
physical sensor for crankshaft torque, so an observer is
designed to compute a non-lagging estimate of the engine
torque ripple (or ac torque). The control strategy is to
command the SA to produce the inverse of this ripple torque
which, when added to the crankshaft torque, will cause the
ripple to be cancelled or decoupled from the net output. It is
also important not to have the SA producing any average (or
dc) torque, so the observer is designed to capture an
estimate of only the ac content of the crankshaft torque.
The engine model described above must be extended to
include the contribution of the starter alternator torque.
Using this extended model as an observer, it is then possible
to determine the appropriate ac torque to use as the

"disturbance input decoupling" control signal, MAC . The

observer topology is shown in Fig. 5 in general terms, and in
detail in Fig. 6. The primary torque production process is the
internal combustion engine, while the secondary torque
production process is the SA. It is vitally important with this
technique to include the effects of both physical processes to
achieve a correct estimate, and to accurately represent the
complete physical system in the observer. As seen in both
Figs. 5 and 6, the command provided to the physical system
(the SA) must also be provided to the representation of the

SA in the observer. The determination of this signal, MAC , is

the primary result of this work.
Finally, because the feedforward portion of the observer is
open loop, motion state feedback from the physical system is
used in the observer to make it closed loop. This corrects for
any inaccuracies in the feedforward estimate [9, 10]
according to the bandwidth of the observer controller.
Torque
Control
Inputs

Disturbance Torque
Primary
+
Torque
Load
+_

Primary Torque
Production
Process
Secondary Torque
Production
Process

OBSERVER SYSTEM
Secondary
Torque
Estimate
Secondary Torque
Production
Process Model

Secondary
Load Model

M1
+

Observer
Controller

+
Load Model

FLVR

PWM
IGBT
Inverter

IFO
Current Regulator
(10 kHz)

MSA

1.8L IDI Engine

SA
Model

_ Jf + J()
1 J()

kio*Ts*z
z-1

kpo +

+
P1 + P4
P2 + P3

kdo*(z-1)
0.0015*z - 0.0005*z

+
1

bac

MICE

Jf + J()
1 J()
2

Ts*z
z-1

12

Friction
Model

Observer System with

Water Brake Load
(Executes at fs = 1.0 kHz)

Fig. 6.

Mac

+
Engine
Kinematics
Model

2 _

Ts*z
z-1

+
_

Combustion
Process
Model

SA
Induction
Machine

bWATER

BRAKE

Extended Crankshaft Torque Observer and AFW Controller

Implementation Block Diagram

Transfer function analysis is used here to tune the closed

loop observer, as has been done in previous work [11]. The
transfer function to analyze is that between the indicated

torque, MICE, and the observer output, MAC . This transfer

function is given in (7) and the corresponding Bode plot is
shown in Fig. 7 (with parameter variation included). Fig. 7

clearly shows that the output MAC is equal to MICE in the

desired passband, approximately 1 to 100 Hz, and attenuated
elsewhere, particularly at dc. This produces the desired
signal for decoupling the ac disturbance torque using active
flywheel control.

MAC(s)
=
MICE(s)

(7)
2

kdos + kpos + kio

2
2
Js + b
Js + b
(J s+b +1) +
b ac (kdos + kpos + kio) + kdos + kpos + kio


J s+b
J s (J s + b ac)

IV. ENGINE TEST STAND AND RESULTS

Disturbance Torque Estimate

for DID Control

Primary Torque
Production
Process Model

...

Measured
State(s)

Secondary
Torque

Primary
Torque
Estimate

IFO Controller
(Executes at
fs = 10 kHz)

Estimated
State(s)

Fig. 5. Extended Observer Topology for Active Vibration Control

The control system described above has been implemented

on a test stand consisting of a Ford 1.8 liter turbocharged
inter-cooled four-cylinder indirect injection diesel engine
with an indirect field-oriented induction motor starter
alternator directly coupled to the crankshaft where the
flywheel would normally reside. A water brake is provided
as a load. The test apparatus is shown schematically in Fig.
8. The field-oriented induction motor control block diagram
is shown in Fig. 9 [16, 17].
A significant amount of testing was carried out over the
speed range of idle (850 rpm) to 2000 rpm, under loaded and
unloaded conditions. Typical results for the observer and the
active flywheel performance are shown in Figs. 10 13.

Figs. 10 and 11 show a sample of the very good tracking

performance of the observed cylinder pressures, indicated
and net torque, and engine speed signals. Figs. 12 and 13
show a sample of the results achieved for the active flywheel
control using the starter alternator and the developed
observer.
V. CONCLUSIONS
This paper presents a simple method to actively control a
starter alternator to decouple unwanted ac disturbance torque
from internal combustion engines. The method is based on a
flexible and accurate dynamic model of an internal
combustion engine that is valid over all speeds and loading

conditions. The model is extended to develop an observer

capable of providing a very high fidelity, wide bandwidth
estimate of the crankshaft ripple torque content that can be
used as a disturbance input decoupling control signal for a
crankshaft mounted integrated starter alternator.
The
observer contains a feedback controller operating on
crankshaft speed feedback derived from the position sensor
used for the starter alternator control, such that no additional
sensors are required to implement this torque control system.
Test results are provided, showing very good performance
both for the observer and the active flywheel control system.

Linear Magnitude

1.5

0.5

4Lr N
iqse* +
N
3PLm2
D
_
D
idse*
*
1
+
Lm

vqse*

M*
0
0.01

0.1

10

3
1 10

100

90

PI

Rotate
PI
_

SLIP = f(r, M*,*)

Phase [degrees]

60

vqss*

vdse*

je

SLIP
e

vdss*
+

idse

30

idss

60
90
0.01

0.1

10

Frequency [Hertz]

Figure 7. Bode Plot of

MICE(s)

for bac =0.2 (___), 0.4(- - -),and 0.1 (. . .)

Cylinder Pressures

Indicated Torque

5 Pressures
Kistler Amplifiers(5)

Trigger

National
Instruments'
AT-MIO-16

ACE

IGBT
Inverter

300 V
Battery

200

2
-200
0

Dell 486/66MHz
PC
EEC-V

400

N-m

Abs Pressure, MPa

Fuel Control
Laptop PC
with
Xmath

A/D
_
_

Nm-rad
sec

Operator
Control Panel

Shaft Position

Fig. 9. Indirect Field Oriented Induction Machine Control Block Diagram

MAC(s)

SA

A/D

iass
ibss
icss

10 kHz Inner Loop Control Software

3
1 10

100

VDC
PWMA Inverter
PWM
PWMB
Duty
Cycle
PWMC

Quad.
Enc.

P/2

3
to
2

Rotate s
e -je iqs

iqse

2
*
to vbs
3
vcs*
r
+

30
0

vas*

200

400

600

Net Torque

180

360

540

720

Engine Speed

1.8L IDI
Diesel Engine

EEC-V: Electronic Engine Controller - 5th Generation

ACE: Advanced Controller for Electric Machines
IDI: Indirect Injection
IGBT: Insulated Gate Bipolar Transistor
SA: Starter Alternator
PC: Personal Computer

Daikin
Clutch

Process
Water
Clutch
Pedal
Signal
12 Volt Line
300 Volt Line
Torque Path

Fig. 8. Schematic Representation of The Engine Test Stand with SA and

Instrumentation

200

900

100

850
RPM

Fuel

Water
Brake

N-m

12V
Battery

Encoder

Kistler Pressure
Transducers

SA Machine

Load Control Valve

800

-100

750

-200

700

-300

180 360 540 720

Crankangle, Degrees

650

0.1
Time, Seconds

Fig. 10. Observer Test Results, 850 RPM, No Load:

Measured (....); Observer (___); Measured Speed (- - -)

0.2

Cylinder Pressures

Indicated Torque

Abs Pressure, MPa

N-m

Abs Pressure, MPa

Measured Cyl. Pressures

400
200
0

2
-200
0

200

400

600

180

360

540

4
2

720

Engine Speed

Net Torque

180 360 540 720

Crankangle, Degrees
Engine Speed

Net Torque

400

2060

1520
200

2040

1500
1480

180 360 540 720

Crankangle, Degrees

2000

0.02

0.04 0.06 0.08

Time, Seconds

0.1

Fig. 11. Observer Test Results, 1500 RPM, 20% FLVR:

Measured (....); Observer (___); Measured Speed (- - -)

Measured Cyl. Pressures

Engine Speed - Passive

950
900

4
RPM

Abs Pressure, MPa

850

2
800
0
0

750
0

180 360 540 720

Crankangle, Degrees

0.1
Time, Seconds
Engine Speed

950

200

900
RPM

N-m

Net Torque
400

850
800

-200
0

180 360 540 720

Crankangle, Degrees

750

2020

-200

1440
0

0
-100

1460
-200

RPM

100
N-m

RPM

N-m

200

0.05

0.1
0.15
Time, Seconds

Fig. 12. Observer & AFW Test Results, 850 RPM, No Load:
Measured Pressures (___); Measured (. . ..) and Observed (___)
Torque; Measured (- - -) and Observed (___) Speeds

-300
0

180 360 540 720

Crankangle, Degrees

1980
0

0.02
0.04
Time, Seconds

0.06

Fig. 13. Observer & AFW Test Results, 1500 RPM, 20% FLVR: Measured
Pressures (___); Measured (. . ..) and Observed (___) Torque; Measured (- -) and Observed (___) Speeds

VI. ACKNOWLEDGEMENT

[9]

The authors wish to acknowledge the motivation provided

by the Wisconsin Electric Machines and Power Electronics
Consortium (WEMPEC) of the University of WisconsinMadison, and the financial support of the Ford Motor
Company.

Luenberger, David G., "An Introduction to

Observers," IEEE Transactions on Automatic Control,
Vol. AC-16, No. 6, December 1971, pp. 596-602.

[10]

Gopinath, B., "On the Control of Linear Multiple

Input-Output Systems," The Bell System Technical
Journal, Vol. 50, No. 3, March 1971, pp. 1063-1081.

[11]

Rizzoni, Giorgio, "Estimate of Indicated Torque from

Crankshaft Speed Fluctuations: A Model for the
Dynamics of the IC Engine," IEEE Transactions on
Vehicular Technology, Vol. 38, No. 3, August, 1989,
pp. 168-179.

[12]

Rizzoni, Giorgio, Yue-Yun Wang, and Sergey

Drakunov, "Estimation of Engine Torque Using
Nonlinear Observers in the Crank Angle Domain,"
ASME DSC-56, Advanced Automotive Technologies,
1995, pp. 189-193.

[13]

Kao, Minghui, and John J. Moskwa, "Nonlinear

Diesel Engine Control and Cylinder Pressure
Observation," ASME Journal of Dynamic Systems,
Measurement, and Control, Vol. 117, June, 1995, pp.
183-192.

[14]

Kao, M., and Moskwa, J.J., "Engine Load and

Equivalence Ratio Estimation for Control and
Diagnostics via Nonlinear Sliding Observers," Int.
Journal of Vehicle Design, Vol. 15, Nos. 3/4/5, 1994,
pp. 358-368.

[15]

Lorenz, R.D., "New Drive Control Algorithms (State

Control, Observers, Self-Sensing, Fuzzy Logic, and
Neural Nets)," Proc. of PCIM Conf., Sept. 3-6, 1996,
Las Vegas, NV, pp. 275-289.

[16]

Lorenz, R.D., Lipo, T.A. and Novotny, D.W., "Motion

Control with Induction Motors," IEEE Proceedings
Special Issue on Power Electronics and Motion
Control, Vol 82, No. 8, pp. 1215-1240, August, 1994.

[17]

Novotny, D.W., and Lipo, T.A., Dynamics and

Control of Induction Motor Drives, Clarendon Press,
Oxford, 1995.

VII. REFERENCES
[1]

Kassakian, John G., H-C. Wolf, J.M. Miller, and C.J.

Hurton, "The Future of Automotive Electrical
Systems," Proceedings of the IEEE Conference on
Automotive Electronics, Dearborn, MI, USA, October
1996, pp. 3-12.

[2]

Tashiro, Hiroshi, et al, "Vibration Damping Control

Apparatus for Vehicle," US Patent 5,537,967, July 23,
1996.

[3]

Kouadio, L.K., P. Bidan, M. Valentin, J.P. Berry, "S.I.

Engine Idle Control Improvement by Using
Automobile Reversible 'Alternator'," Proceedings of
the IFAC 13th Triennial World Conference, San
Francisco, CA, USA, Paper No. 8b-03 5, 1996, pp. 9398.

[4]

Heywood, John B., Internal Combustion Engine

Fundamentals, McGraw-Hill, Inc., 1988.

[5]

Winterbone, D.E., Thiraurooran, C., and Wellstead,

P.E., "A Wholly Dynamic Model of a Turbocharged
Diesel Engine for Transfer Function Evaluation," SAE
770124, 1977.

[6]

[7].

[8].

Dobner, D.J., "A Mathematical Engine Model for

Development of Dynamic Engine Control," SAE
800054, 1984.
Cook, J.A., and Powell, B.K., "Modeling of an
Internal Combustion Engine for Control Analysis,"
IEEE Control Systems Magazine, August, 1988, pp.
20-26.
Jensen, J.P., Kristensen, A.F., Sorensen, S.C.,
Houbak, N., and Hendricks, E., "Mean Value
Modeling of a Small Turbocharged Diesel Engine,"
SAE 910070, 1991.