You are on page 1of 54

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245328932

Modeling and Simulation of an M1 Abrams Tank with Advanced Track


Dynamics and Integrated Virtual Diesel Engine

Article  in  Mechanics of Structures and Machines · January 1999


DOI: 10.1080/08905459908915707

CITATIONS READS

15 736

12 authors, including:

Dennis N. Assanis Matthew P Castanier


University of Delaware U.S. Army Ground Vehicle Systems Center
342 PUBLICATIONS   9,928 CITATIONS    129 PUBLICATIONS   3,566 CITATIONS   

SEE PROFILE SEE PROFILE

Gregory Hulbert Dohoy Jung


University of Michigan University of Michigan-Dearborn
112 PUBLICATIONS   7,910 CITATIONS    43 PUBLICATIONS   966 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Lightweight vehicle structures that absorb and redirect destructive energy away from the occupants View project

DAAD Step-by-step integration mehtods View project

All content following this page was uploaded by Dennis N. Assanis on 18 July 2016.

The user has requested enhancement of the downloaded file.


MECH. STRUCT. & MACH., 27(4), 453-505 (1999)

Modeling and Simulation of an M1


Abrams Tank with Advanced Track
Dynamics and Integrated Virtual
Diesel Engine*

Dennis N. Assanis, Walter Bryzik,l Matthew P. Castanier,


Ian M. Darnell, Zoran S. Filipi, Gregory M. Hulbert,
Dohoy Jung, Zheng-Dong Ma, Noel C. Perkins,
Christophe Pierre, Craig M. Scholar, Yongsheng Wang,
and Guoqing Zhang 2
AUTOMOTIVE RESEARCH CENTER
DEPARTMENT OF MECHANICAL ENGINEERING AND ApPLIED MECHANICS
W. E. LAY AUTOMOTIVE LABORATORY
THE UNIVERSITY OF MICHIGAN
ANN ARBOR, MI 48109-2121

ABSTRACT

New capabilities for simulating a tracked vehicle are presented, including an


advanced dynamic track model, a high-fidelity diesel engine system model, and
an integration scheme to perform a coupled simulation of vehicle/powertrain dy-
namics. These capahilities are essential for understanding the interplay of vehicle
dynamics and powertrain dynamics. including track vibration (and durability). sus-
pension response. and engine performance. The dynamic track model considers the
track as an equivalent continuum and captures longitudinal and transverse track
vibrations, static sag, and superposed translation. A low-order discrete model is
developed by employing modal track coordinates. The continuum approximation

'Communicated by E. J. Haug
'Chief Scientist, U.S. Army Tank-Automotive and Armaments Command.
'Current Address: with Navistar International Transportation Corporation. Melrose Park,
Illinois.

453

Copyright © 1999 by Marcel Dekker, Inc. www.dekker.com


454 ASSANIS ET AL.

for the track is validated through experiments on a representative track span. This
track model is extended and implemented into a commercial multibody dynamics
code-DADS-through development of a new user-support-force element that in-
tcgrates the track element with the vehicle hull and suspension system. A range
of dynamic track models results that allows one to tailor the degrees of freedom
to a selected frequency range of interest in order to balance computational cost
and accuracy. A virtual diesel engine model is developed as a tool to investigate
the possible replacement of the current gas turbine engine used in the M I Abrams
tank. This study demonstrates the power of this simulation tool for evaluating new
vehicle concepts prior to prototyping and manufacturing. The engine model is
developed within the MATLAB/Simulink environment. Therefore, the integrated
vchicle/powertrain model requires the coordination of two coupled models that
reside in distinct simulation environments. To achieve this integration, a new nu-
merical method-referred to as the leading-following approach-is developed,
based on an explicit predictor-corrector scheme. This approach allows independent
simulation environments to be coupled, offers easy extension to multiple applica-
tions, promotes efficient simulations, and requires only simple implementations
of the software interfaces compared to the conventional master-slave integration
approach. Numerical examples are reviewed in the paper, to highlight capabilities
of the fully integrated simulation of a diesel-powered M I tank.

I. INTRODUCTION

Analysis, design, and optimization of complex ground vehicle systems are


time-intensive processes that involve expensive testing of physical prototypes.
As a cost-effective alternative to physical testing, the Automotive Research Cen-
ter" (ARC) has been pursuing the development of a flexible vehicle simulation
system composed of a hierarchy of models of varying resolution. In this frame-
work, subsystem models, algorithms, and simulation modules are selected ac-
cording to the objectives of the overall vehicle simulation. By using an agile,
tailored simulation environment, significant improvements can be made to the
product development process for military and commercial ground vehicles.
This paper investigates the development of such a virtual proving ground,
while focusing on modeling and simulation issues associated with tracked vehi-
cles. As a particular example, a case study is presented for a tracked vehicle
used in the US military, the M I Abrams tank. It is important to note that the
simulation of heavy, tracked vehicles presents unique challenges. For instance,

"The ARC (hl1p://arc.engin.umich.edu) is a US Army Center of Excellence for Automo-


tive Research at the University of Michigan, currently in partnership with the University
of Alaska-Fairbanks, Clemson University, University of Iowa, Oakland University, Uni-
versity of Tennessee, Wayne State University, and University of Wisconsin-Madison.
MODELING AND SIMULAnON OF M I ABRAMS TANK 455

enhanced capabilities are required for modeling the track dynamics as well as
the high-output powertrain system. Consequently, the objectives of this case
study are as follows:
• Develop advanced tracked vehicle models for efficient prediction of track
vibration, engine performance, and vehicle dynamics.
• Provide a virtual prototyping tool to support conceptual design of a pro-
posed engine system.
• Create a flexible simulation environment for integration of different com-
ponent/subsystem models to produce a complete vehicle system simula-
tion.
This case study addresses the need to efficiently model and evaluate track
vibration and its transmission into the vehicle structure. Track vibration is a
major source of vibration energy transmitted into a tracked vehicle. It may
adversely affect track life, as well as seriously degrade the performance of
on-board instrumentation and personnel. The dominant deformation modes of
vehicle tracks include both longitudinal and lateral (vertical-plane) displace-
ments. These track displacements couple to rigid body rotations of the driver
and follower sprockets, support wheels, road wheels, and torsion bars (suspen-
sion elements). These components are shown in Fig. I, a schematic representa-
tion of a tracked vehicle.
Previous models of tracked vehicle systems have utilized a variety of repre-
sentations for the track. Early investigations of tracked vehicle dynamics fo-
cused on kinematic track effects, assuming negligible track inertia. Wheeler [I]
incorporated a quasi-static model for track stretching, to approximate vertical
forces of wheel-ground interaction. Garnich and Grimm [2] used the same track

Support
Track
Roller
Hull

Drive
Sprocket

Fig. I. Example tracked vehicle: M I Abrams tank.


456 ASSANIS ET AL.

model, but accounted for the effects of track bridging, drive sprocket interaction,
and a track-compensating linkage. Bennett and Penny [3] extended the quasi-
static model by accounting for initial sag of the track. McCullough and Haug
141 developed a tracked vehicle model assuming quasi-static track response,
referred to as the track superelernent, in which rigid body equations of motion
were derived for systems containing recurring subsystems. Dhir and Sankar, in
developing a model of a high-speed tracked vehicle [5], also approximated track
loads by considering the quasi-static stretching of a massless belt. However,
since all of these models neglect the inertia of the track, they cannot capture
track vibration.
On the other hand, tracked vehicle models have been developed in which
each pitch of the track is assigned the requisite degrees of freedom (OaF) in
forming a large multibody dynamic system. Choi [6] presented a multibody
dynamics model in which the track consisted of individual links connected by
single-Dfif revolute joints. Highly detailed models of this nature have also been
developed for the US Army Tank-Automotive and Armaments Command
(TACOM) by Wilcox [7,8]. These models allow analysis of track motion and
track/sprocket interaction to high fidelity. Similarly, models by Galaitsis [9]
employ a multibody representation for the tracked vehicle that was subsequently
used to predict dynamic track loads during flat terrain traversal. However, these
models require the solution of a very large number of degrees of freedom, thus
leading to simulations that are computationally expensive.
In this paper, low-order, efficient models of vehicle tracks that capture track
vibration are presented. High efficiency is achieved by introducing a hybrid
model that treats the track as a continuum that interacts with rigid-body elements
(sprockets, road wheels, and support rollers). Each track span is represented as
a continuous elastic member, with the longitudinal (stretching) and transverse
track response described using a selected set of vibration modes. Thus, a modal
representation of track dynamics may be generated within user-specified fre-
quency ranges. This modeling approach captures dynamic effects with few de-
grees of freedom, relative to established multibody dynamic formulations.
A full tracked vehicle model is then developed, with the objective of provid-
ing new capabilities for simulating track and suspension system dynamic re-
sponse. These capabilities are essential for predicting the durability of the track,
as well as vibration transmission into the interior of the vehicle. By using the
continuous track model, a full vehicle model involving relatively few degrees
of freedom is assembled for the example military tank. A mixed Eulerian/La-
grangian description is employed, wherein the rigid-body elements of the hull
and suspension are coupled to component modes of the track spans. Implemen-
tation of this model within a commercial multibody dynamics software package
(DADS) is accomplished through the development of a new force element for
the track. This force element allows one to tailor model fidelity to achieve par-
MODELING AND SIMULATION OF MI ABRAMS TANK 457

ticular simulation goals. As an example, the response of an M I tank traversing


a bump course is evaluated. Calculated results include the dynamic track ten-
sion, normal contact forces, and vehicle acceleration levels. These results em-
phasize the importance of track vibration in tracked vehicle modeling and illus-
trate advantages in using a modal representation for the track.
To predict dynamic response of the track and suspension system with high
fidelity, it is critical that precise loading conditions be prescribed for dynamic
track models. For this to be accomplished, the time-varying torque at the drive
sprocket needs to be accurately specified. It should be noted that very different
sprocket torque histories may result for a given driver demand and road profile,
depending on selection of the engine, the driveline, and their control systems
(e.g., fuel injection and transmission shift logic). Consequently, a high-fidelity
simulation of dynamic response for tracked vehicles demands a fully coupled
powertrain model.
Since the inception of the Automotive Research Center in 1994, one of its
active research areas has been propulsion system simulation. As a result, a
framework for flexible powertrain modeling has been developed within the
MATLAB/Simulink environment [10,11]. Engine and driveline modules of
varying resolution have been built into this simulation system so that it may be
tailored to required applications.
In a previous study, Anthony et al. [12] applied this flexible simulation ap-
proach to the study of the M I tank, with emphasis on driveline components
and transients initiated by the tank steering system. In their work, the current-
production, 1500-HP (1l18-kW) gas turbine engine could be represented by
either a torque-speed map or a thermodynamic model of the gas turbine cycle.
However, the diesel engine has emerged as a potential powerplant for the next
generation of battle tanks, due to its attractive power density, fuel economy,
reliability, and maintenance characteristics. Therefore, one objective of the cur-
rent study is to explore the integration and performance of a virtual diesel pow-
ertrain in a tracked vehicle.
To establish the performance characteristics of such a powertrain, a high-
fidelity engine simulation module is embedded into the powertrain simulation,
which in tum is integrated with advanced structural dynamics models of the
tracked vehicle. This approach not only allows the prediction of performance
for the virtual powertrain, but it also enables accurate coupling of the power-
train, the vehicle structure, and the driver (or cruise controller). The integrated
engine-in-vehicle simulations that are reported in this work demonstrate the im-
portance of providing an accurate profile of the torque that drives the track.
Furthermore, it is demonstrated that the engine can experience significant varia-
tions in speed and load torque as a result of the road profile and the attendant
dynamic response of the track and suspension system.
Integration of the engine/driveline model with the vehicle dynamics model
458 ASSANIS ET AL.

presents one of the major technical challenges in this case study, due to the
different software environments in which the original simulations were devel-
oped. The transient engine and the drive line models were implemented in MAT-
LAB/Simulink to maximize flexibility and to allow easy coupling of component
modules. The vehicle dynamics model, however, was developed using a com-
mercial multibody dynamics code, DADS. One possible approach to coupling
these models would be to use an existing interface software called DADSlPlant,
which allows MATLAB to call DADS as a special function. In this integration
approach, MATLAB is employed as a master code that integrates and solves
the system equations, controlling integration time step and solution error. Conse-
quently, DADS is employed as a slave code, whose integrators are inactive
during the simulation process. To perform a more seamless integration that may
be extended to general applications, independent of the software packages to be
integrated, an alternative approach is presented.
In the approach employed in this study, the relationship between MATLAB/
Simulink (powertrain subsystem) and DADS (vehicle structure subsystem) is
not master and slave, but rather leading code and following code. MATLAB/
Simulink leads the simulation by predicting the angular velocity of the drive
sprocket, solving the powertrain subsystem problem, and outputting torque to
the vehicle model. DADS follows MATLAB/Simulink during the simulation by
inputting the torque from the powertrain model, solving the vehicle subsystem
problem, and correcting the angular velocity predicted by the powertrain model.
Therefore, this approach may be considered an explicit predictor-corrector
method. Communication between MATLAB/Simulink and DADS is achieved
through an interface developed at the ARC.
This paper is organized as follows. Section II focuses on the vehicle structure
model and introduces a new track model. This track model is then extended and
implemented in a multibody dynamics model for the MI Abrams tank, using
DADS. Results for the vehicle dynamics model are reviewed to illustrate the
important role played by track vibration. In Section III, a virtual, turbocharged,
intercooled V-12 diesel engine for the M I Abrams tank is introduced. This is
followed by a description of the models used to generate the diesel engine sys-
tem and driveline modules. To validate the diesel engine simulation, predicted
steady-state engine performance is compared to that of an existing engine of
similar size. Numerical results are presented for powertrain simulations using
both conventional and Low Heat Rejection engine designs, integrated with a
simple point-mass vehicle model. In Section IV, this point-mass is replaced by
the complete vehicle and track DADS model of Section II, in order to create a
coupled vehicle/powertrain model. The interface developed to couple the DADS
vehicle model and MATLAB/Simulink powertrain model is described. The
high-fidelity tank simulation is then used to study vehicle response during accel-
eration on flat terrain and during low-speed operation over rough terrain. Exam-
MODELING AND SIMULATION OF MI ABRAMS TANK 459

pIe results show the significance of coupling between the vehicle structure and
the powertrain, in terms of obtaining realistic subsystem responses and assessing
the relationship between track vibrations and engine torque/speed fluctuations.
In Section Y, conclusions from this study are summarized, and. topics of future
research are discussed.

II. TRACKED VEHICLE MODEL

In this section, a low-order model for an M I Abrams tank is presented that


supports efficient simulation of track vibration and overall vehicle response
(Fig. I). The vehicle model employs a continuum element for a translating track
and captures both transverse and longitudinal track vibration. Solution efficiency
derives from the use of low-order vibration modes to describe track response.
Using a mixed Eulerian/Lagrangian description of motion, the rigid-body ele-
ments of the hull and suspension are coupled to component modes of the track
spans, to form a full vehicle model. The resulting model is implemented within
a commercial dynamics software package (DADS), through the development of
a new force element for the track.

A. Track Element Model

There are two dominant modes of track vibration; transverse and longitudinal.
Transverse track vibrations generate deformation normal to the track span and
in the vertical plane. This deformation requires relative rotation of adjacent
pitches about the connecting pinlbushing assembly. Transverse track vibrations
arc readily visible and may also reach amplitudes that are sufficiently large to
produce impacts of the track with the underside of the hull. Longitudinal track
vibrations generate deformation tangent to the track span and require the relative
separation of adjacent pitches through radial compliance of the pinlbushing as-
sembly. Longitudinal vibrations produce dynamic track tension and may pro-
mote bushing failure. These two modes of vibration are coupled, in general, due
to the small static sag of the tensioned track.
A continuum model is employed that describes the in-plane dynamics of a
massive, translating, flexible track sagging under its own weight. With reference
to Fig. 2, this continuum model represents one span of the track circuit shown
in Fig. I and reduces to that of a shallow-sag cable [13]. The coordinates U(s,/)
and V(s,t) denote, respectively, longitudinal and transverse span displacements
about an equilibrium configuration. Here, t denotes time, and s is an independent
span coordinate defined on a domain s E [0, L], where L is the span length.
The equilibrium configuration can be assumed to be a parabola, since the
track satisfies the small sag condition KL < I, where K is the curvature of the
460 ASSANIS ET AL.

.... ~ Dynamic configuration


-;<, «:
# ....
# ....
# ....
## ~,
# U(S,t) -,

Equilibrium configuration

Fig. 2. Schematic of (rack clement.

equilibrium configuration [13,14]. The equations for free in-plane motion about
this equilibrium are derived in Ref. 13. After linearization, these equations of
motion become

( V,, - v)L
2
C .,
k] = -U
[ U - -V
L
I
g.n +
1¥V;L
-g- [U.' --V
k]
L (I)
.' J

(2)

with the nondimensional parameters

EA
,
2 2 Po 2 C
VI =--, v, =--, V, = -, k = KL (3)
pgL pgL gL

Here, EA is track section modulus, pg denotes track weight/length, Po is static


track tension, and c denotes track translation speed relative to the hull. The
quantities V" v, V" and k are (nondimensional) speed of propagation of longitudi-
nal waves, speed of propagation of transverse waves, track transport speed, and
equilibrium curvature, respectively.
Equations (I) and (2), coupled in U(s,t) and V(S,I), provide the mathematical
description of track span dynamics used here, When employed in a full vehicle
model, these equations of motion couple to the motion of the vehicle through
boundary conditions that describe moving supports (sprocket, idler wheel, sup-
port rollers, etc.).
The modeling of a track (composed of a finite number of pitches) by an
equivalent continuum requires justification, To this end, a series of experiments
461
462 ASSANIS ET AL.

Discernible peaks mark the natural frequencies of the first four span modes.
Table I compares these measured frequencies to those predicted using the track
model. The conclusion is that a small sag cable theory provides excellent agree-
ment for the first two resonance frequencies of a track span and lower-bound
estimates for the next two. In Section 11.0, it is shown that the low-frequency
track response is well described using just the first two (component) modes of
each track span.
Using this track model, the natural frequencies and mode shapes of the upper
track circuit can be calculated. Figure 5 shows the first four natural frequencies
and corresponding mode shapes for the case in which the front and rear road
wheels are fixed to the hull.

B. Generalized Model for Coupled TracklVehicle Motion

There are multiple sources of excitation to the track. At the track-terrain inter-
face. substantial excitation is delivered to the lower track circuit (that contacting
the terrain). Moreover, the motions of the front and rear road wheels provide
excitation at the boundary of the upper track circuit. Finally, rigid-body vehicle
motions generate inertial loading for the entire circuit. All of these sources can
produce substantial track vibration.
The equations of motion. Eqs. (I) and (2). represent the track model for the
simplest case when the supporting points of the track are fixed in space. Thus.
in Eqs. (I) and (2). the track deformations U and V are measured with respect
to a spatially fixed coordinate system. Considering that the track is attached to
a moving vehicle, and this vehicle is often driven over rough terrain (possibly
at high speeds). the track model presented in the previous subsection is now
extended.
Consider a track segment that is attached to two general support rollers. as
shown in Fig. 6. The domain of the track segment extends from point 0 to point
B. which both lie along the common tangent to the support rollers. Let oX)' be
a local coordinate system. with the x axis originating at point 0 and passing

TABLE 1
Comparison of predicted frequencies and
experiment results (Hz).

Predicted Experiment

fi 7.2 7.2
f, 14.6 14.7
h 23.2 25.3
j; 29.7 37.0
MODELING AND SIMULATION OF MI ABRAMS TANK 463

(a) It = 3.7 Hz

(b) h = 5.5 Hz

(c) fa = 7.3 Hz

(d) 14 = 7.6 Hz
Fig. 5. Vibration mode shapes of upper track circuit.
464 ASSANIS ET AL.

Roller 1
o x
Fig. 6. Track segment in a moving coordinate system.

through point B, and y measured positive along the outward normal to roller I.
Let (x",Yo) represent the coordinates of point 0 at the global coordinate system
OXY, and let 8 represent the angle between the x axis of the local coordinate
system and the X axis of the global coordinate system. Also, let I:.) and .r;
(t, f,) represent reaction forces acting on the left and right ends of th~ track
segment, respectively, and let (g" gy) represent components of gravity acting
along the axes of the local coordinate system. Subscripts x and yare used to
denote the components directed along the local x and y axes, respectively.
Consistent with this application, assume that the static sag and dynamic defor-
mation of the track are considerably smaller than the length of the track seg-
ment. Furthermore, for this study, the vehicle achieves speeds considerable
slower than the propagation speeds of transverse (and longitudinal) waves in the
track (less than 20 MPH). Therefore, terms containing the translational speed c
in Eqs. (I) and (2) are small and can be ignored [16]. Based on these assump-
tions, the equations of motion of Eqs. (I) and (2) can be readily generalized to
include the motion of the support rollers. This leads to the following generalized
Newton-Euler equations for the track segment/roller system:

(EAU. x - k, \I).x = p(U. 1I - zev, - xe' + .1'0) (4)

(PoV:x + k,U).x - k,V= P(V:II + 2eU., + xe + Yo) (5)

L L
m x; - n8., + f pU. lldx - 28.f pV:,dx = Ix"+ Ix + mg, (6)
o 0
MODELING AND SIMULATION OF MI ABRAMS TANK 465

L L
my" + lie + Jo p Y"dx + 28 J pU.,dx == r; + f; + mg;
0 . -
(7)

L L
Ie + IIYo + J0
pxY"dx + 28 J pxU.,dx == Lf; + IIg,.
0 '
(8)

where

I 2 I
m == pL, II == zPL , 1== _pLJ k == EA pgy k, == EA(pgyP (9)
3 ' 1 Po' - po}

Note that Eqs. (4) and (5) are extensions of Eqs. (I) and (2) (for Vc == 0) that now
capture the inertial loading on the track due to support motion. In particular, the
last three terms on the right sides of Eqs. (4) and (5) represent coupling effects
of vehicle rigid body motion with track deformation. Equations (6)-(8) represent
the resulting Newton-Euler equations for the three degrees of freedom for the
local coordinate system (i.e., the equations of rigid body motion of the overall
track segment). Thus, Eqs. (4)-(8) constitute a flexible multi-body dynamics
formulation of the continuous track model, under the assumptions of small track
deflection and large track rigid body motion. These equations are employed as
the basis for implementing the track model in an overall vehicle model.

C. Vehicle Model
l. Background. To employ Eqs. (4)-(8) in a full vehicle model, the equa-
tions are discretized using a Component Mode Synthesis (CMS) method [17,18].
A mixed EulerianlLagrangian description is employed, wherein the rigid-body
elements of the hull and suspension are coupled to component modes of the
track spans. As illustrated in Fig. 7, an Eulerian description is used for the 78-
pitch translating track (Eqs. (4)-(8)), while a Lagrangian description is used for
the rigid-body elements of the hull and suspension. The rigid-body elements
include the hull/turret, drive sprocket, idler wheel, track-adjusting link, two sup-
port rollers, and seven pairs of road arms and road wheels (see Fig. I).
The track equations of motion, Eqs. (4)-(8), are described using four types
of component track modes: (I) a rigid-body mode, (2) a constraint mode, (3)
longitudinal vibration modes, and (4) transverse vibration modes. The rigid-
body mode describes superimposed translation speed, and the constraint mode
describes static track elongation. The longitudinal and transverse vibration
modes describe dynamic deformation about the sagged and translating equilib-
rium state. The resulting track equations are solved within the commercial code
466 ASSANIS ET AL.

Eulerian Description of Track Motion

lagrangian Description of Hull + Suspension Motion

Fig. 7. Formulation of full vehicle model.·

DADS by developing a new general-purpose force element. Traditional force


elements allow for only stiffness and damping. This new element also accounts
for track inertia. which allows the track spans to be conveniently coupled to the
suspension components (support rollers. idler wheel, road wheels, and sprocket)
as standard elements.
The track-terrain interface is modeled as a nonlinear elastic foundation, with
a no-slip condition imposed at the track segment below the fourth (middle) road
wheel. Each road wheel has a rotational degree of freedom. It is assumed that
the track maintains normal contact with the suspension elements at all times. In
designing a simulation, arbitrary terrain profiles, as well as sprocket drive torque
curves, can be specified via data files.
2. Simulation capabilities. Through user menu choices, modal track models
of variable fidelity can be constructed and used to support different simulation
goals. Currently, six models are available, listed here in order of increasing
complexity/accuracy, as follows:
• Static (spring-damper) model: this model reduces to that of a massless
track as employed in Refs. 4 and 5.
• Lumped mass-spring-damper model: the track inertia in the upper spans
is lumped into the support rollers while the track is modeled as massless.
• Quasi-static (Guyan reduction) model: the track inertia is lumped into the
support rollers using static condensation.
• Dynamic model with longitudinal vibration: this model captures dynamic
track stretching using longitudinal span vibration modes [19).
• Dynamic model with transverse vibration: this model includes transverse
span vibration modes and assumes quasi-static track stretching [15).
• Dynamic model with coupled transversellongitudinal vibration: this model
captures track vibration by including both transverse and longitudinal span
modes.
MODELING AND SIMULATION OF MI ABRAMS TANK 467

The user is free to select the number of modes (longitudinal and/or transverse)
to be used in any of the dynamic track models. Possible simulation outputs
include descriptions of tank rigid body dynamics, longitudinal track vibration,
transverse track vibration, dynamic track tension, track contact forces, dynamic
response of discrete components, and animation sequences. A selection of these
outputs is used to discuss simulation results in the following subsection.

D. Results for M1 Abrams Tank: Case of


Prescribed Torque

A major source of track vibration is response of the tracked vehicle to uneven


terrain. To examine this, consider the response of the M I Abrams tank driven
over a bump course. The section of the bump course under consideration is 500
feet (15 m) in length and includes numerous bumps, ranging in profile from I
ft (30.5 em) high and 25 ft (762 em) long-a large obstacle-to 0.25 ft (7.6
em) high and 2 ft (61 em) long-a small obstacle. See Fig. 8 for a depiction of
the bump course profile, which is that of Aberdeen Proving Ground Course 4.

1.2 r--.----.-----.--...,----.-----r----r----,--r---,
Shownin
Fig. 9

o Ll.-..J..J...........L.......1L.....U_.LI..
500 450 400 350 300 250 200 150 100 50 0
Distance (It)

Fig. 8. Discretized terrain profile for the bump course. Note that the height is greatly exaggerated
with respect to distance. The segments of the bump course shown in Fig. 9 and Fig. 10 are labeled.
468 ASSANIS ET AL.

Fi~. 9. Bump course: tank traversing a large obstacle.

Starting from rest, the vehicle accelerates over this course and reaches a maxi-
mum speed of approximately 25 MPH. Figures 9 and 10 are snapshots from this
simulation, showing the tank trasversing large and small obstacles, respectively.
In Section IV, this simulation will be revisited, after inclusion of a vehicle
powertrain model. To provide a point of reference, the results in this section are
from the uncoupled (pure vehicle) model, wherein the torque at the drive
sprocket is a prescribed function of time. Results are presented using three dif-
ferent track models: (I) static model (21 OaF), (2) dynamic model with longitu-
dinal vibration (26 OaF, one mode per track span), and (3) dynamic model with
coupled transverse/longitudinal vibration (31 OaF, two modes per track span).
As a first result, consider acceleration of the vehicle mass center as the vehi-
cle traverses the bump course. Figure II illustrates a portion of the time history
of the longitudinal acceleration component of the vehicle center of mass, where
positive acceleration corresponds to a deceleration of the vehicle as it travels in
the negative .r direction. Results from the static track model (Fig. II a), indicate
large vehicle deceleration levels as obstacles are encountered. These levels are
unrealistically high. since this track model is very stiff. Results of the two dy-
namic track models (Figs. lib and lie) exhibit significantly greater frequency
content due to track vibration. Moreover, these results illustrate how a softer

Fig. 10. Bump course: small obstacles.


MODELING AND SIMULATION OF Ml ABRAMS TANK 469

"",---,----,------.----,,----.,------,

.... 1----'---'----'---'----'------'
o 10 15 20 2S 30
TIme(s)
(a) Static Model

"".---.,------,-----,----,--.,---------,

.... L-_-L-_---'--_---'-_-----'_ _->-----J


o '0
"Time(5)20 " 30

(b) Longitudinal Vibration Model

""r----,---.,------,---.,------,-----,

"j
·········t···········l············~·
......... ~ l. i + L .
: : 1 1 ~
.... 1-_--'--_1-_--'--_1-_--'-------'
o '0 15

TIme!s)
20
" 30

(c) TransverseJLongitudinai Vibration Model

Fig. 11. Longitudinal acceleration of vehicle center of mass predicted by three track models.
470 ASSANIS ET AL.

track (dynamic representation for longitudinal and transverse deformation) may


absorb a larger portion of the obstacle impulse forces.
Figure 12 illustrates a portion of the time history of the dynamic track tension,
sampled at the location of the rear support roller (see Fig. I). Results from the
static model (Fig. 12a) cannot capture the dynamic amplifications due to track
vibrations that are apparent for both of the dynamic models (Figs. 12b and 12c).
Furthermore, in the frequency range considered here, the longitudinal vibration
model (Fig. 12b) underestimates the peak dynamic tension, which is affected by
low-frequency transverse vibration modes. Note that the peak values of tension
are approximately 30,000 Ib, which is nearly a 100% increase over the initial
(static) track tension. Tension extremes of this magnitude are expected on harsh
bump courses, and they can only be captured using dynamic track models.
The dominant vibration modes are revealed by the results of Fig. 13, which
show the fast Fourier transform (FFT) of the dynamic tension histories of Fig.
12. The peak dynamic tension near 3.7 Hz in Fig. 13c correlates with the natural
frequency of the first mode of vibration of the track circuit as predicted by the
coupled transverse/longitudinal model in Section II.A (Fig. 5). The peak dy-
namic tension near 14.0 Hz in Figs. 13b and 13c correlates with the natural
frequency of the first track circuit mode predicted by the longitudinal vibration
model of Scholar and Perkins [19]. The mode near 3.7 Hz is dominated by
transverse track response. Therefore, the transverse span modes are required in
order to fully resolve the dynamic tension, which can be seen by comparing
Fig. 13b with l3c and Fig. 12b with 12c. The mode near 14.0 Hz is dominated
by longitudinal response (track stretching), so it is reasonably well captured by
either dynamic model. Finally, the results of Figs. 12 and 13 clearly illustrate
the limitations of the static model in resolving dynamic track tension.
As another comparison, consider the computed dynamic normal contact force
that exists at the rear support roller, while traversing the same bump course.
Figure 14 illustrates the fast Fourier transform of the computed normal force
from two of the three models. The static model results are not shown since they
are null throughout the frequency range. For the coupled transverse/longitudinal
model of Fig. 14b, two clear peaks in the force signal occur near 3.7 and 5.5
Hz. These frequencies correlate with natural frequencies of the track circuit
vibration modes involving transverse vibration in the spans adjacent to this
roller, which are shown in Fig. 5. The longitudinal vibration model (Fig. 14a)
is simply unable to describe the major dynamic changes of this normal contact
force, since they are largely due to the transverse track dynamics near the roller.
Computation of the dynamic normal contact force is crucial to calculating en-
ergy transmission into the tank hull, as well as predicting when the track loses
contact with the roller and possibly impacts the underside of the hull. Thus, it
is vital to include transverse track dynamics in any model that seeks to predict
these phenomena.
MODELING AND SIMULATION OF M I ABRAMS TANK 471

o L..-...L.---'-_-'-----'----'_-'---'----'_-'---'
o 12151'21242730

Time(.)

(a) Static Model

Time I')
(b) Longitudinal Vibration Model

i
·····r······
I I !
····t······f······t······t·····_!······t· ···j······1······!"·····
o 1--'-------L-L----'_L....-'-------L-i.----'c.......J
o 12 15 II 21 ~ V »
Time I.)

(c) TransverselLongitudinal Vibration Model

Fig. 12. Dynamic tension predicted by three track models.


472 ASSANIS ET AL.

....[ i .+••••••••••••••L ~ l i ~ .
1 I ! ! ! ! iii
····..·····.I······.;.··_··.~····· •.I·····.4-···.··I····.-·1···.-·..·····

IJ~;!;~t~~
Frequency (Hz)

(a) Static Model

12 1$ 18 21 2~ 'l7 30
Fl8quency (Hz)

(b) Longitudinal Vibration Model

Frequency (Hz)
(c) TransverseILongitudinal Vibration Model

Fig. 13. Fourier transform of dynamic tension.


MODELING AND SIMULATION OF Ml ABRAMS TANK 473

1200r-~_...,--_;---,--,..--,_--,-_.,----,_..,

4.5 6 7.5 9 10.5 12 13.5 15

Frequency (Hz)
(a) Longitudinal Vibration Model

1200 ...---,_-,--_,...---,_-,--_,...---,_-,--_,...-,

····v······;···
. . . . . . .
··~······.·······;······~······t····_··;.······.·······
ill l l ! 1
: : : : : : :
g: .~ .••• ~ ~ ~ ..••_.: ~ l....•••

~ l ~ 1 j l j !
{1. ·+····1······+·····+······1······+·····1·······
iU 1 ! ! i ! ! !
E
o -f··· :·······~······+······i·······~······~·······
z : : : : : : :
'0 ! : ! ; 1 1 i
.. ·t:···· .L L .:. L i ,.. 1 .
tu. : l j ! j j
1.5 3 4.5 7.5 9 10.5 12 13.5 15

Frequency (Hz)
(b) TransverseILongitudinal Vibration Model

Fig. 14. Fourier transform of dynamic normal contact force at second support roller.

III. DIESEL-ENGINE POWERTRAIN FOR M1 TANK


A. Virtual 1500-HP Diesel Engine System

The M I Abrams main battle tank is a very complex ground vehicle system
requiring a propulsion system with high power per unit of mass or volume,
low fuel consumption, high reliability and ease of maintenance. Two alternative
474 ASSANIS ET AL.

powerplants are primary contenders for this vehicle: the gas turbine and the
diesel engine. In current production, the M I Abrams main battle tank uses a
1500-HP (I I 18-kW) gas turbine. However, with increased emphasis on fuel
economy of military vehicles, the advanced diesel engine, with its higher effi-
ciency and less demanding logistics, is emerging as an attractive option for
future generation vehicles. This section explores the possibility of introducing a
highly turbocharged and intercooled, V-12 diesel engine to meet the propulsion
needs of this 65-ton tracked vehicle. The proposed configuration of the engine
system is shown in Fig. 15. Use of a large number (12) of cylinders helps limit
the size of the individual cylinder, thus allowing engine operation at reasonably
high speeds, a prerequisite for achieving high specific power output. Use of two
turbochargers and intercoolers, one per bank of cylinders, is preferred because

Exhaust Exhaust
Air Gas Gas Air

V12 ENGINE

Fig. 15. Schematic of the virtual V 12 diesel engine system.


MODELING AND SIMULATION OF MI ABRAMS TANK 475

it allows the use of smaller turbochargers for better response and eliminates
risks of unbalanced operation, which would be possible in applications in which
two compressors are connected to a single intercooler and a common intake
manifold. The bore (0. [59 m) and stroke (0.159 m) of the virtual engine cylinder
correspond to the production CUMMINS KTTA38-C engine, yielding 37.7 L
of total displacement. However, the engine compression ratio of 15, as well as
other engine design parameters including the method of cooling, are different
from the specifications of the reference CUMMINS engine. Nevertheless, the
basic geometric similarity between the reference and the virtual engine allow
for validation of performance prediction trends before the virtual engine is in-
stalled in the vehicle.

B. Powertrain System Simulation

The complete diesel powertrain for the M I tank includes the diesel engine
system, the torque converter (TIC), automatic transmission, and the final drive
(D) that provides torque to the sprockets via two drive shafts of the single
drive axle (see Fig. [6). To generate the complete powertrain system simulation,
appropriate subsystem modules need to be developed and integrated. The con-
tent of the primary modules is summarized below. While these models are coded
in different languages, i.e. FORTRAN, C, and MATLAB, MATLAB/Simulink
allows to convert any of them into a format suitable for generating S-functions
that can be used as building blocks in the Simulink graphical programming
environment. The flexible powertrain simulation structure in MATLAB/Simul-
ink that was originally developed by Rubin et al. [20] for 6 x 6 truck with a
six-cylinder engine and a drivetrain with three axles was the foundation for
reconfiguring the diesel powertrain simulation for this study. In particular, the
user-friendly features of the Simulink graphical interface are utilized to the full-

Powertrain

Dr;~~"" .........
Sprocket

Fig. 16. The coupling interface between the powertrain and structure is at the drive sprocket
(circled). From a modeling perspective, the output from the powenrain model is the drive sprocket
torque. which is the input for the structure model.
476 ASSANIS ET AL.

est in reconfiguring the powertrain system to represent the V-12 cylinder engine
with two intake and exhaust banks, two sets of turbochargers and intercoolers,
and a single drive axle.
I. Engine module. The engine module is comprised of multiplexed single-
cylinder modules, links with external component modules; i.e., compressors,
turbines, intercoolers, and intake and exhaust system elements. The engine cyl-
inder model tracks the thermodynamic processes within the cylinder throughout
a cycle, as a function of crank angle. An engine dynamics module provides a
link with the vehicle through the drivetrain.
Thermodynamic diesel engine cylinder module: The foundation of the die-
sel engine cylinder module used in this work is the physically based, thermody-
namic, zero-dimensional model developed by Assanis and Heywood [21]. In the
parent model, the cyclic processes in the cylinder are represented by a blend of
more fundamental and phenomenological models of turbulence, combustion,
and heat transfer. The parent simulation has been validated against numerous
test results from diesel engines of various sizes, ranging from highway truck
engines [21] to large locomotive engines [22]. The engine cylinder system is
open to the transfer of mass, enthalpy, and energy in the form of work and heat.
Cylinder contents are represented as one continuous medium, uniform in pres-
sure and temperature, characterized by an average equivalence ratio. Quasi-
steady, adiabatic, one-dimensional flow equations are used to predict mass flows
past the intake and exhaust valves. Ignition delay is predicted using an Arrhen-
ius correlation, based on in-cylinder conditions following fuel injection. Com-
bustion is modeled as a uniformly distributed heat release process, using Wat-
son's correlation [23], which accounts for premixed and diffusion-controlled
burning. Convective heat transfer is modeled using Nusselt number correlation
based on turbulent flow in pipes. An energy cascade zero-dimensional turbu-
lence model 121,24] provides the characteristic velocity and length scales re-
quired for the convective heat transfer model. Radiative heat transfer is added
during combustion [21]. Both steady-state and transient combustion chamber
surface temperatures of the piston, cylinder head, and liner can be predicted
from a specification of the wall structure and coolant boundary conditions [21].
The temperatures of the oil reservoir and the coolant in the block and head are
computed using a lumped heat capacitance thermal network, based on the work
of Bohac et al. [25].
The parent diesel engine model [21] and its transient extension [26] were
originally coded in FORTRAN. To use the single cylinder model as a building
block for assembling a multicylinder simulation in the MATLAB/Simulink en-
vironment, a FORTRAN-MEX file containing all necessary state derivatives
and gateway routines is created. More details on the conversion technique can
be found in Ref. 27.
MODELING AND SIMULATION OF MI ABRAMS TANK 477

Engine dynamics: An engine in a tracked vehicle experiences frequent, often


rapid, variations of driver demand and external load, due to track vibrations
induced by the road profile. Hence, the powertrain simulation requires an engine
model that is capable of dealing with dynamics resulting from these varying
operating conditions. The primary inputs to the engine dynamics system are the
driver demand, which is translated into mass of fuel per cycle, the ambient
conditions, and the resistance torque determined by the vehicle dynamics. Out-
put from the cylinder module is the indicated torque, as the result of only the
gas force acting on the piston. A friction submodel based on Millington and
Hartles correlation [28] is used to predict engine friction losses and convert
indicated to brake quantities. In this application, the model uses instantaneous
engine speed, rather than mean engine speed, as used in the traditional approach.
Next, the external load torque, imposed on the engine by the vehicle or the
dynamometer, is subtracted from the brake torque and the net value is passed
on to the engine dynamics equation. The variable crankshaft inertia is calculated
in a separate block and supplied to the nonlinear dynamics equation applied to
the complete engine [29]. The new value of instantaneous crankshaft speed is
returned after every integration step.
Turbocharger: The turbocharger model [10] uses compressor and turbine
maps to correlate instantaneous mass flow rate and turbomachinery efficiency
with shaft speed and pressure ratio. Creating a virtual engine implies that a real
turbocharger that would be a suitable match for the engine is not readily avail-
able. Hence, a hypothetical turbocharger is created, based on a much smaller
unit of known characteristics designed for the 12.7-L, DDC Series 60 truck
engine. The reference truck-size turbomachinery maps are scaled using tech-
niques described in Refs. 30 and 31, to increase mass flow rate by 60%, thus
accommodating the increased cylinder size, while accounting for the fact that
the virtual engine system uses one turbocharger for each bank of six cylinders.
The scaled turbocharger unit operates with lower shaft speeds and higher polar
moment of inertia than the reference unit. The instantaneous rate of change of
angular velocity is determined based on the balance between compressor and
turbine torque, divided by the polar moment of inertia of the rotor.
Intercooler: There are two intercooler units in the system, one for each bank
of six cylinders. The intercooler model [20] uses specified values for wall tem-
perature and overall heat transfer coefficient. It consists of a control volume
connected to an orifice element. Outputs are the mass flow and enthalpy flow
at the intercooler outlet.
2. Drivetrain module. The drivetrain module, containing the torque con-
verter, transmission and rear driveline (see Fig. 17) plays a critical role in inte-
gration of the engine with the vehicle. Driveline component models used in this
module are based on models developed by Munns and Rubin [10,20]. The en-
478 ASSANIS ET AL.

i----------------·-----~I
I
I
I i __ .!

~-------------------
,.... ,
T
Engine Trans D
C
~

--------------------
I
I~
~--.
J
I

Fig. 17. Schematic of the powertrain with automatic transmission.

gine's output torque is connected to the torque converter's pump shaft (input),
while the torque converter's turbine torque (output) is transmitted to the
sprocket through the gear train. Pump speed is obtained as the solution of the
engine dynamics equation, while the transmission speed depends directly on
the solution of vehicle dynamics. At any instant, using the pump to turbine
speed ratio, the pump's K factor is determined from a look-up table. Then, the
pump torque is calculated based on the instantaneous engine speed and the K
factor. Subsequently, the turbine torque is determined from a look-up table,
using the pump torque value and the pump to turbine speed ratio. The initial
torque converter data are based on a production unit for an M I tank, but the K-
factor look-up table is scaled properly to match the much lower rated speed of
the V-12 diesel engine.
The torque converter's output shaft is connected to the four-speed automatic
transmission with planetary gears. Operating modes are defined by combinations
of locked and unlocked clutches. Shifting is accomplished through simultaneous
disengagement and engagement of appropriate clutches during a user-specified
interval of time. For this study, the transmission model is based on data avail-
able for a prototype Allison X 1100-1C automatic transmission, developed for
military purposes, and the heavy-duty, truck type Allison HT-740 transmission
[10,32]. The original transmission model [10] is augmented with the gear upshift
logic [32], based on engine speed, to allow for vehicle acceleration runs over
the entire operating range. For the study presented here, the shifting speed is
1850 RPM. The transmission output (speed and torque) is further transformed
by the propshaft, differential, and driveshaft modules [10,11], to finally provide
torque at the sprocket. Original models developed by Munns [10] for a truck-
trailer driveline are modified for the tank driveline, consisting of a single rear
MODELING AND SIMULATION OF MI ABRAMS TANK 479

axle driving the track via a pair of sprockets. The final gear ratio in the differen-
tial is also modified to match the driveline with the engine rated speed of 2100
RPM, since the X1100-1C unit was originally designed for an engine that would
have a considerably higher rated speed.
3. Engine model validation. As stated in Section lILA, while design differ-
ences exist between the virtual V-12 diesel and the reference Cummins
KTTA38-C engine, simulated performance characteristics of the virtual engine
are compared in this subsection to those published for the production engine.
The goals of this exercise are to confirm that (I) simulated performance trends
of the virtual engine are realistic; and (2) the virtual engine can deliver an output
level comparable to the gas turbine, prior to integration of the engine in the
vehicle.
Figure 18a compares the brake power and torque characteristics for the
KTTA38 engine and three versions of the virtual V-12 engines, the baseline
engine with an all metal combustion chamber and two low heat rejection (LHR)
configurations. The latter utilize O.5-mm and I.O-mm thick zirconia coatings
sprayed on the combustion chamber surfaces. Clearly, predicted performance
trends for the virtual engines closely track the observed trend lines for the pro-
duction engine. Note that the production engine, in its two-stage turbocharged
version, delivers 1350 HP (1007 kW) at 2100 RPM. Both the baseline and LHR
virtual diesel engines produce higher power output than the reference engine,
with the increases being 9.9%, 15.7% (0.5 mm of zirconia), and 19.8% (I mm
of zirconia), respectively. This is achieved by allowing higher values of boost
pressure and maximum combustion pressure in the cylinder. It has been assumed
that the technology and the materials used in a military engine would allow
higher engine rating, perhaps even at the expense of somewhat inferior durabil-
ity compared to the reference, industrial engine. As even the baseline configura-
tion of the virtual engine produces a rated power within 2% of the target power
level, the task of developing a diesel engine of a rating close to the gas turbine
has been accomplished.
As can be observed from Fig. 18a, the power output of the baseline (all metal)
engine at lower speeds is slightly better than that of the LHR versions. However,
the slight disadvantage of the LHR engines at low speeds is reversed at medium
speeds, and they eventually achieve much higher power levels at the rated speed.
These trends can be explained with the volumetric efficiency tradeoffs caused
by increased component surface temperatures and increased exhaust tempera-
tures in the LHR engines. Owing to their higher surface temperatures [21], LHR
engines experience considerably reduced net heat transfer, compared with the
conventional all-metal engine. This increases the level of specific exhaust en-
ergy, which is the source of propelling the turbocharger turbine. At the same
time, substantial heat transfer takes place from the hotter walls to the incoming
gas, thus reducing its density (and thus volumetric efficiency) and increasing
480 ASSANIS ET AL.

6000 2000

5500
............
5000
-. 1500

r
~
4500
~
.,
::J ~

2"
0 1000 ~
0
l- c,

~:....-_--------...,.,
500
- - Conventional Eng.
- - . LHR Eng. (0.5 mm Coating)
-LHR Eng. (1.0 mm Coating)
- ... Cummins Eng.
2000 0
800 1000 1200 1400 1600 1800 2000 2200
Engine Speed (RPM)
(a)

4.0 . - - - - - - - - - - - - - - - - .

"E" 3.5
!
~ 3.0
0..

~ 2.5

j --Conventionel Eng.
2.0
- - ·LHR Eng. (0.5 mm Coating
--LHR Eng.(1.0 mm Coaling)

Fig. 18. (a) Comparison of the brake power and torque values for the Cummins KTTA38 engine
and three versions of the virtual V-] 2 engines. (b) Comparison of the intake manifold pressures for
the conventional and LHR engines.
MODELING AND SIMULAnON OF M I ABRAMS TANK 481

compression work. Dominance of one effect over the other changes with engine
and turbocharger operating conditions. Over the medium-to-high engine speed
range. the higher exhaust energy of the LHR engines result in higher intake
manifold pressures (see Fig. 18b). which more than make up for the effect of
reduced fresh charge density during the intake process. Admission of more air
allows more fuel to be burned. thus higher power output from the LHR than the
conventional engine. On the other hand. at low engine speeds. the difference
between the boost pressure levels produced in the conventional. versus LHR
engine systems. is small. Under those circumstances, the adverse effect of the
hot combustion chamber walls on volumetric efficiency becomes more domi-
nant, thus explaining the reduced levels of engine torque in the LHR configura-
tions.

C. Mobility Studies with the Point Mass Vehicle Model

Depending on simulation objectives, alternative vehicle dynamic approaches


can be used to perform mobility studies using an integrated powertrain model.
At one extreme, a point mass vehicle model [10) can be selected, assuming that
vehicle mass is lumped at the center of gravity. Such an approach can give
sufficiently high-fidelity predictions of vehicle acceleration and speed on flat,
smooth roads. As an illustrative example. a study comparing the full throttle
performance of a point mass tank, powered by both conventional and LHR
engine configurations, is presented in this section. All three main modules-i.e.,
the engine, the driveline, and the point mass vehicle modules-were available
in Simulink. Thus, integration was done by creating graphical links between
blocks. A common solver from the MATLAB/Simulink library was used to
simultaneously integrate differential equations for the complete engine-vehicle
system [32).
A detailed multibody vehicle dynamics model is necessary for investigation
of vehicle-powertrain-track interactions during transients associated with a
tracked vehicle riding over a rough terrain that excites significant track vibra-
tions. Integration of the powertrain with a multibody DADS model. featuring the
tank structure and the track superelement described in Section 11, is presented in
Section IV.
1. Tank acceleration with conventional engine. Figures 19a and 19b show
the angular velocity and torque histories for various shafts in the system during
the first 20 sec of full power acceleration from standstill. Engine speed increases
sharply after the start. thus causing a large speed slip between the TIC pump
wheel (connected to the engine shaft) and the TIC turbine wheel (connected
to the transmission). As a result, significant multiplication of engine torque is
experienced at the TIC turbine shaft and propagated through the transmission
during the first 3 sec of the transient. as seen in Fig. 19b. Note that the increas-
482 ASSANIS ET AL.

250

200

' .: . : f/
II
:ag I
I \ J
\ I
I \
I'
'" \
\

al
a
en
150
, i,m~m::f/
iij

~
c:
.2 100

a:
"J
\ i

Transrnlsslon Out

50
Drive Shaft

.~ -
_.-
0
0 5 10 15 20
Time (s)
. (a)

4‫סס‬oo

5 10 15 20
Time (s)
(b)

Fig. 19. Histories of (a) rotational speeds and (b) torques for various shafts in the conventional
powertrain system during full power acceleration on flat terrain.
MODELING AND SIMULATION OF Ml ABRAMS TANK 483

ing TIC pump torque increases the external load applied on the engine. Hence,
engine speed shows hesitation during the first second of the run, until the engine
is able to balance the external load and eventually produce enough brake torque
to start accelerating again. The rate of change of shaft speed is determined by
the instantaneous difference between the engine and pump torque levels and the
effective polar moment of inertia of the engine and TIC pump wheel. Since a
significant turbocharger lag occurs at the beginning (see also Section III.C.2 and
Fig. 21a), engine torque and speed build up gradually until the first gear shift.
At that time, the speed ratio in the torque converter diminishes, thus the engine
torque line approaches the TIC out line.
The gear shift creates a very sharp fluctuation of shaft angular velocities and
torques. The gear ratio change moves the transmission-in (TIC out) speed closer
to the transmission-out speed, almost instantaneously. This introduces again a
large slip in the torque converter, and thus multiplication of torque. Therefore,
a saw tooth can be identified on the TIC output torque history (see Fig. 19b)
after the shift. Increased slip in the TIC increases the pump torque loading the
engine, and thus a sharp decrease of engine speed is observed immediately after
the shift. After the shift is completed, the engine accelerates with minimal effect
of turbocharger lag, since turbocharger speed and boost never drop down to
levels seen at vehicle launch. Reducing the overall gear ratio during the next
two shifts produces a similar transient behavior, with reduced amplitude of
speed and torque spikes during the shift. The torque levels on the driveshaft
show a general decreasing trend, with increase of vehicle speed and decrease
of overall gear ratio, as expected. Ultimately, transmission-in (TIC out) and
transmission-out angular velocities become the same in fourth gear, where the
gear ratio in the transmission becomes unity.
2. Tank acceleration with low heat rejection engine. Acceleration runs are
performed for the same vehicle, but with ceramic insulation (O.5-mrn and 1.0-
mm thick sprayed zirconia coatings) applied on the LHR engine combustion
chamber. All input parameters remain the same as for the base, all-metal engine
combustion chamber, including the fuel control logic. As before, the run starts
with a step increase of driver demand to 100%. The overall response of the
system, as well as the nature of the interaction of the engine with the torque
converter, and the fluctuations during gear shifts are quite similar to those of
the conventional powertrain. However, LHR engine versions yield better vehicle
performance than the conventional one; i.e., higher speeds after 20 sec and
shorter acceleration and passing times. Table 2 summarizes vehicle performance
metrics for all three engine versions.
The speed differentials between each of the LHR versions and the conven-
tional one are plotted versus time in Fig. 20. While the LHR versions show
slower initial response, they eventually overtake the conventional one. The tran-
sition occurs when the higher exhaust energy level of the LHR engines enables
484 ASSANIS ET AL.

TABLE 2
Effect of combustion chamber surface coating on vehicle mobility.

Vehicle Speed Time to Accelerate


after 20 sec
Engine Design (kmlh) 0-60 kmlh (s) 20-60 kmlh (s)

Conventional 65.67 17.2 13.2


LHR with O.5-mm thick coating 66.29 16.95 12.9
LHR with I.O-mm thick coating 66.43 16.9 12.8

their turbocharger to deliver higher boost pressures, thus offsetting the adverse
effect of elevated wall temperatures on volumetric efficiency. As shown in Fig.
21 a, the initial turbocharger lag period is quite similar for both the conventional
and LHR engines. During the period of low boost pressure levels, engine volu-
metric efficiency is a strong function of the combustion chamber surface temper-
ature in each of the engines. Clearly, from Fig. 21 b, both the surface temperature
level and its cyclic fluctuations increase at a much faster rate in the LHR engine

1.0

0.8

~
0.6
::!!.
CD 0.4
g.
l!!
CD 0.2
!l:
0
'0 0.0
CD
CD
0.
00
-0.2

-0.4

5 10 15 20
Time (5)

Fig. 20. Speed differences during acceleration of vehicles powered by the conventional and the
two LHR engine versions.
MODELING AND SIMULATION OF Ml ABRAMS TANK 485

700
g
Gi
ci.
~ E
e
:l I!!-
~ 8
Q.
'"::>
'l:

J C/)
c
.~
a. Conventional Engine

0 5 10 15 20
5 10 15 20
TIme(s) Time (s)

(a) (b)

5 10 15 20
Time (s)
(c)

Fig. 21. Comparison of the conventional engine and LHR engine with I.O-mm coating during the
20-sec vehicle acceleration: (a) boost pressures: (b) piston surface temperature variations; (c) ex-
haust gas temperatures.
486 ASSANIS ET AL.

(with a I.O-mm thick coating) than the conventional engine. Elevated surface
temperatures decrease filling of the LHR cylinder with fresh mass during the
turbocharger lag phase. On the other hand, the LHR engine experiences a higher
rate of increase of its exhaust temperature levels (see Fig. 2Ic). Consequently,
after about 5 sec from start, the LHR engine's intake manifold pressure departs
significantly from the line calculated for the conventional engine, thus becoming
high enough to overcome the adverse effect of charge heating during intake.
Higher charge density yields higher engine air flow, and this allows higher fuel
flow through the engine. Consequently, the LHR engine brake torque output
eventually exceeds that of the conventional engine during the full load accelera-
tion, thus explaining the behavior shown in Fig. 20.

IV. INTEGRATED SIMULATION OF


ENGINE-VEHICLE SYSTEM
A. Integration Algorithm

The vehicle dynamics model of the M I tank, developed using the DADS
simulation package, is presented in Section II. Diesel engine and driveline mod-
els for the M I Abrams tank in MATLAB/Simulink are presented in Section III.
To create a complete system simulation, it is necessary to integrate these two
modules in a seamless way. A conventional approach for this integration is to
employ a master-slave approach, with the MATLAB/Simulink code as the main
calling the DADS package as a special function. This approach may result in a
difficult numerical integration problem for the MATLAB/Simulink solvers.
Hence, an alternative approach is presented, based on an explicit predictor-cor-
rector scheme. Here, MATLAB/Simulink is employed as a leading code (the
predictor) and DADS is employed as a following code (the corrector).
The major features of this approach are as follows:
• Independence: suitable for distributed simulation environments, in which
integrated codes are language independent and they can be executed in
different platforms at different locations
• Generality: general enough to integrate different commercial codes
• Simplicity: requires only a simple interface for the codes to be integrated
• Efficiency: each subsystem is solved in parallel, using its own solver for
the subsystem problem
Figure 22 shows that the coupling interface is at the drive sprocket, and the
major data exchange between the two subsystem models is the angular velocity
(RPM) of the dri ve sprocket and the output torque of the powertrain system.
The integration process works in the following way. First, the MATLAB/Simul-
MODELING AND SIMULATION OF Ml ABRAMS TANK 487

Powertraln Structure

Matlab™
',L-
, ---J
,
,,
Drive ',Sprocket
,

Fig. 22. Integration of powertrain and structure models.

ink module predicts the angular velocity of the drive sprocket, based on angular
velocity and angular acceleration obtained at the last integration time step. Then
the predicted angular velocity is used as an input to the powertrain model for
predicting engine response and output torque from the powertrain system. To
be more specific, the instantaneous sprocket speed is propagated back through
the driveline, and it eventually determines the slip in the torque converter; i.e.,
the ratio between the torque converter turbine and pump speeds. This ratio af-
fects both the TIC active and resistance torque values, the latter being used in
the engine dynamics equation. Therefore, the system of differential equations
for the engine/driveline can now be solved using the numerical integrator in
MATLAB/Simulink, and the solution includes the engine speed and torque. The
DADS-MA TLAB interface then reads the torque from the powertrain model,
and this torque is applied to the drive sprocket of the DADS vehicle model at
the current DADS integration time. The subsystem equations of the vehicle
model are then solved by the DADS solver to obtain dynamic response of the
vehicle, including the angular velocity and angular acceleration of the drive
sprocket. Finally, the angular velocity and angular acceleration obtained from
DADS calculations are sent to the powertrain model for the next integration
step.
MATLAB/Simulink leads the simulation, but only by a preset amount of
time. If MATLAB/Simulink is running too fast, it will idle to wait for DADS.
DADS is behind MATLAB/Simulink, but by no more than a given time interval.
If DADS is running too fast, it will be forced to idle and wait for MATLAB/
Simulink. Note that since MATLAB/Simulink and DADS use different integra-
tion time steps, which can vary within each code, an interpolation method is
employed to produce a torque value at the required integration time, using the
known values of torque.
488 ASSANIS ET AL.

Two interfaces (one for MATLAB/Simulink, the other for DADS) have been
developed for communication between the two codes. On the MATLAB/Simul-
ink side, a user-subroutine has been developed, which is called by MATLABI
Simulink as an S-function. This user-subroutine reads the angular velocity, an-
gular acceleration, and other necessary data from the DADS output file, predicts
the angular velocity for the current MATLAB integration time, and writes the
output torque and the corresponding time to the MATLAB/Simulink output file
that DADS will read. On the DADS side, an engine-force element has been
developed, which reads outputs from the powertrain model, interpolates the
torque value for the current DADS integration time, and applies the torque to
the drive sprocket of the vehicle model. The engine-force element also writes
the converged values of the vehicle's response to a file that the MATLABI
Simulink user-subroutine will read. The user-force element has been included
in the DADS menu. Therefore, parameters in the engine force element can be
changed in an interactive way.
Figure 23 shows a schematic of the complete M I Abrams tank model. As
shown in Fig. 23, there are three different integration processes at two different
levels. Two lower level integration processes include: (I) integrating the diesel
engine model and the drivetrain model and (2) integrating the dynamic track
model and the vehicle body/suspension model. Both lower level processes inte-
grate the subsystem modules within their own software environments, as de-
scribed in Sections II and III. The top level integration of the powertrain and
vehicle modules in the distributed computing environment has been described
previously. In a physical sense, the information flow from one module to an-
other is as follows: the engine provides torque at the flywheel; i.e., the torque
converter pump wheel. This torque is converted in the TIC and propagated
through the transmission and the rest of the driveline, with its value being modi-
fied at each step in accordance with the gear ratio of each element. The resulting
torque acting on the drive sprocket then accelerates the vehicle through
sprocket-track-terrain interactions. As the vehicle moves, obstacles on the un-
even terrain decelerate the vehicle and induce vibration in the track and suspen-
sion components, which result in variations of the sprocket/driveshaft speed.
This is propagated back through the driveline to ultimately determine the TIC
turbine speed. The speed ratio between the TIC pump and turbine wheels deter-
mines the TIC pump torque; i.e., the resistance torque loading the engine. Inte-
gration of the engine dynamics equation, together with all the engine process
equations, closes the loop and provides data for the next calculation step.

B. Intelligent Vehicle Speed Controller


1. Background. A tracked vehicle is intended to operate on a variety of
terrains that often include obstacles. A cruise controller provides an element of
MODELING AND SIMULATION OF M I ABRAMS TANK 489

-------------- ..
I
I

.. _---------
I -------
':,\ ..L A- .!-A .!- 1- F

Multi-Body
Vehicle Dynamics
,.----------m:I:b[l1
IL J Including
Longitudinal
and Transverse
Diesel Engine Drivetrain
Track Vibration
System

MATLAB I Simulink DADS


Fig. 23. Schematic of the integrated M I Ahrams tank model.

driver intelligence that is needed in order to maintain relatively constant vehicle


speed, especially while driving over bumpy courses. Such an intelligent vehicle
speed (IYS) controller should minimize vehicle jerk, thus enhancing the ability
of the vehicle to carry out its mission, while minimizing fatigue on both the
driver and sensitive vehicle electronic systems. The basic structure of the IYS
controller is composed of two parts, a controller and an actuator, as illustrated
in Fig. 24.
Speed control algorithms used in automotive cruise controllers typically are
of proportional-integral (PI) type [33,34]. Mathematically, the PI controller can
be described as follows [35]:

U (t) = «, [e(t) + '!'J' e (t) d't] (10)


~o
490 ASSANIS ET AL.

Required Cruise Controller


Vehicle
Speed \6(1)

Disturbed Torque

Fig. 24. Block diagram of vehicle speed control.

where U (I) is output provided to the actuator controlling the fuel rack position;
e(t) is control error between the desired vehicle speed, V,(t), and the actual
vehicle speed, Vall); K; is proportional gain; and T, is integral time constant.
The implementation of the PI controller in MATLAB/Simulink is shown in Fig.
25. Notice that a saturation block is located after the PI controller of Fig. 25.
This nonlinear link enforces a limit for maximum fuel delivery.
The challenge in modeling an IVS controller is how to determine proper
values of the parameters K, and T,. If a closed-form, mathematical model of the
controlled system (diesel engine, drive train, and vehicle) can be formulated, in
the form of a transfer function or a state-space equation, controller parameters
K, and T, can be readily determined from basic control theory. However, in the

Vehicle Speed
Setting Vr(s)

u(s)

Sum2

Fig. 25. Simulink block diagram for the PI controller.


MODELING AND SIMULATION OF Ml ABRAMS TANK 491

case at hand, such models are constructed in the form of complex differential
and algebraic equations, which makes it infeasible to derive a simple transfer
function.
An alternative, practical approach to set the controller's parameters is based
on the open loop, Ziegler-Nichols tuning method. The method is based on mea-
surements of the dynamic response of the system output, when a unitary input
is imposed on the system input. The normalized fuel rack position, U" of the
diesel engine is selected as the input to the vehicle system; while the normalized
vehicle speed, Va" is selected as the output of the system. These variables can
be expressed as

(lla)

(lIb)

where Va and R are vehicle instantaneous speed and fuel rack position of the
diesel engine at time t: subscript 0 designates steady-state values before a uni-
tary step change is supplied; and subscript h designates rated speed at which the
engine produces rated power at full rack.
Assuming that open loop response of the system to a unitary step input exhib-
its an S-shaped characteristic without overshoot (see Fig. 26), two performance
parameters can be used to characterize the transient response of vehicle speed.
One characteristic parameter is related to the transport lag and the other de-
scribes the rate of response. The point at which the slope of the response curve
reaches its maximum value, usually at the point of inflection, is first determined,
and a tangent line at this point is drawn (Fig. 26). This tangent line intersects
the horizontal axis (the time axis) at X and intersects the vertical axis (the vehi-
cle speed axis) at Y. The tuning parameters, K, and TI , for the PI controller can
be determined from X and Y as
0.9
tc, = TYT (l2a)

TI = 3!XI (l2b)
At this point, it is important to stress that the vehicle dynamic system is
highly nonlinear, because characteristics of the controlled system change with
operating conditions (e.g., vehicle mass, road slope angle) and with time (e.g.,
the performance of engine and drive train decay as time goes on). Although a
cruise controller can be well-tuned, or designed under one circumstance, it may
exhibit sluggish behavior when vehicle parameters and/or road terrain are
changed. Since several operating conditions exist in one simulation process,
providing a schedule of gains is a practical approach to deal with the nonlinear-
492 ASSANIS ET AL.

?
;;
'5
Co
:l
o
'0
s
?
;:)
S
0-
.5
8
~
Ql
Co
o
16

Fig. 26. Output response Vo,(t) to a unit step input U,(t).

ity. According to suitable scheduling variable(s), (e.g., vehicle mass or slope


angle), the parameters K, and T, are tabulated as a discrete set of values, or
fitted as a set of linear segments or a polynomial. Consequently, when operating
conditions change, the scheduling variables also change, hence the parameters
of a cruise controller are tuned accordingly (Fig. 27).
2. Fuel system. The output of the IVS controller is a signal that changes
from 0 to I and could be interpreted as the rack position, according to traditional
diesel engine terminology. The signal is then translated into the actual amount
of fuel injected per cycle in the smart actuator module. The actuator module
provides two additional control functions to the turbocharged diesel engine; i.e.,
the constant phi function and the two-speed governor function, consistent with
modern practice in which electronic controls replace mechanical governors and
pneumatic smoke limiters. The constant phi actuator function determines the
amount of fuel delivered to the cylinder, based on the IVS signal, and the actual
air-flow through the engine. The two-speed governor function of the actuator
prevents the engine from overspeeding and stalling at low speeds. The amount
of fuel is corrected proportionally to the speed difference term. The complete
actuator model has been implemented in MATLAB/Simulink.
The implementation of the constant phi actuator function is done in two steps.
The target fuel/air equivalence ratio at full load (q>,g,) is determined from
MODELING AND SIMULATION OF Ml ABRAMS TANK 493

Controller's SCheduling
Perameter: Kp, TI I Variable(s)
Table
I
I I

Desired
Vehicle
Speed

Cruise Controller Vehicle System -.


,.-. --+ System
Output

Actual VehicleSpeed

Fig. 27. Block diagram of gain scheduling.

8
$rgr = 0.45 + 8.8757 X 10- (ro, - 2100)' (13)

where ro, is engine speed in RPM. The correlation is extracted from experimen-
tal data acquired by the ARC team, using a six-cylinder heavy-duty truck diesel
engine [36]. Then, the cyclic mass of fuel injected is determined by

mFinjlcyc = .
mai,<ptgt (A)
- (Rack) (60n R
- - l J ) ,) (14)
F ncyl
s

where mFi,ycy, is amount of fuel injected per cycle per cylinder, mo' , is mass flow
rate of the intake air, (A/F)s is stoichiometric air/fuel ratio, (Rack) is the signal
from the driver/IVS (varying from 0 to I), nR is the number of crank revolutions
for each power stroke per cylinder, and n,yl is total number of cylinders. Applica-
tion of this two-part control logic assures sophisticated correction of the amount
of fuel during rapid engine transients. As an example, if air flow is reduced due
to turbocharger lag, the above function will automatically reduce the amount of
fuel delivered to the engine and prevent conditions leading to excess smoke
emission.

C. Integrated Simulation Results

The integrated system model has the capability to simulate a complete system
under real dynamic conditions. In contrast to results presented in Section II.D,
where a prescribed torque was used for vehicle structure system simulations,
494 ASSANIS ET AL.

torque on the drive sprocket now comes from a real transient engine and drive-
line model. The engine feels the resistance torque that may include a high fre-
quency content in addition to the main, slower varying signal. This sheds new
light on engine-vehicle interaction, especially if the system is excited with very
dynamic input signals. The fully integrated vehicle utilizes the speed controller
included in the powertrain to enable runs over anomalous terrains, with almost
constant speed. Fuel input is automatically adjusted to respond to terrain condi-
tions, thus causing engine torque to fluctuate at a high rate-a sharp contrast
with results given in Section II, where sprocket torque was supplied as a con-
stant input parameter. Hence, the integrated tool allows analysis of both the
vehicle structure system and the powertrain system, under realistic operating
conditions, which is important for evaluation of component and system designs.
The first run presented for this integrated simulation is acceleration of the M 1
tank driven over flat terrain. The rack position is fixed at ninety percent (i.e.,
driver demand is 90% of full power), and the transmission upshift logic is modi-
fied to allow shifts at lower engine speeds, in accordance with a less aggressive
acceleration schedule.
Figure 28 illustrates the time history of powertrain output torques. The engine
torque (bottom line) is multiplied in the torque converter as a result of transient
conditions and a large slip in the torque converter. The sprocket torque is further
multiplied in the transmission, with the multiplication effect being the greatest
in the first gear. After each shift, the gear ratio-and thus the sprocket torque-is
decreased in a discrete step. The effects of gear shifts on track tension are shown
in Fig. 29, where the three spikes on the track tension line correspond to three
gear shifts. Consequently, these sharp fluctuations of force excite track vibra-
tion, as illustrated in Fig. 30. These results clearly demonstrate the capability of
the integrated system model to capture the effects of powertrain dynamics on
response of the vehicle track/structure.
Next, the simulation is used to investigate response of the diesel powered M I
Abrams tank, driven over the bump course described in Section II.D. As in the
previous run, the rack position was fixed at 90%. Figure 31 compares time
histories of the torque at the sprocket during the run over the bump course and
the flat terrain. The rough terrain has significant effects on powertrain output
torque. As expected, the second and third gear shifts are delayed, and vehicle
acceleration suffers as more drive torque is required to overcome the obstacles
and propel the tank over the bumps.
A comparison of vehicle speed histories obtained with and without the power-
train model for the bump course is given in Fig. 32. In the case without the
powertrain model, the tank clearly reaches a much higher speed (27 MPH),
relative to cases that include the full powertrain model. This speed is unrealisti-
cally high, since a typical driver would operate the tank at much lower speeds,
given a terrain profile that includes I-foot-high obstacles. In other words, these
MODELING AND SIMULATION OF Ml ABRAMS TANK 495

x 10·
3.5 r-----,------,-----,----......,-r===:;;::;==;:::::::::::::::::::;:=:==il

-E
I

-
Z
Q)
::l
2

e-
O
I- 1.5

/",.
II

J
n--------J
0.5 .. I .:.;. .. _._ _ ,;:.~.~.;.;.;.;;;.;.;..;.;.;.;.;.;;.;.;.;.

I
/
__ -
_.- - --'--
~;.;;;.;:~.;.:.;;.;._

:
.~-

oL- L- -L- ....I- ~----=---__:

o 5 10 30
Time (s)
Fig. 28. Integrated simulation: shaft torques for flat terrain.

driving conditions lead to unrealistic predictions for dynamic response of the


vehicle and attendant loads on vehicle components. With an integrated power-
train model, it is possible to reduce vehicle speed with a specified rack position
and reduced transmission upshift speed. However, as seen in Fig. 32, with rack
position fixed at 90%, the tank can still reach a speed that is significantly higher
than the target speed of 5 MPH-a speed shown to be typically chosen by
human drivers on the bump course. Of course, it would be possible to adjust
the rack position setting further, but hunting for the exact value that would
produce conditions closest to desired would be lengthy trial and error process.
This process is also limited in terms of the minimum acceptable constant fueling
rate. P.elow a certain value for the rack position, the engine will not be able to
',enOUgh torque to overcome the first bump and the vehicle will stall.
,tically, the only way to adjust and maintain low speed, normal for the
.. \ such roughness, is to activate the intelligent driver interface described
496 ASSANIS ET AL.

c:
.9
III
c:
~

13050L-_.l.-_.L-_...L-_...L_....l..._....l.._--l.._--l._----1_---J
10 12 14 16 18 20 22 24 26 28 30
Time (s)

Fig. 29. Integration simulation: track tension for flat terrain.

in Section IV. This is illustrated by repeating the run over the bumpy course
with the IVS controller activated, at a setting of 5 MPH. Figure 33 presents the
rack position history produced by the speed controller during the complete run
over the bump course. The result-i.e., the vehicle speed history line-has been
superposed on the other two lines shown in Fig. 32. The controller quickly
increases fuel input to 100% at vehicle launch, in order to accelerate the vehicle
to the desired speed as soon as possible. Then, as a speed of 5 MPH is reached
before the 8th second of the run, rack position is decreased to maintain that
speed. Soon, the tank hits the first bump and a sharp spike in fuel input is
observed as the controller responds to sudden deceleration of the vehicle. The
remainder of the run over the bump course displays a number of sharp fluctua-
tions of the rack position, covering the range between 10% and 100%. These
fluctuations are often of high frequency, due to the impact of track vibrations
on vehicle and powertrain response. Although oscillations are also visible on
the vehicle speed line shown in Fig. 32, overall the controller is able to maintain
the speed close to the target value of 5 MPH.
The impact of such conditions on the engine system is visible in Fig. 34,
where the rotational speeds of various shafts in the powertrain system are plotted
MODELING AND SIMULATION OF Ml ABRAMS TANK 497

-0.09 .---_,.---_,.-------,,.-------,----,----,_----,_----,_----,_----,_---,

-0.095 .·; . .
· .··: ;
... ...
~

· . ..
···· .... ...
-0.1
·· .. .
.

-0.105 ..... ...... . .


~

··· ..: ;.
...
.
··· ... ..
.
··· ... ....
-0.11 -: . . .: : -:: .
.
. .
.
~ ~ ~

..
., .
··: ....... : : : : :.
·· . ...
-0.115 ..... '.' \ i .....• '".. . ; ~ , '" •......

-0.12 '--_.l...-_-'---_-"-_--'-_--'--_---!.-_-'-_----'-_---l._----'
10 12 14 16 18 20 22 24 26 28 30
Time (s)

Fig. 30. Integrated simulation: transverse track deflection for flat terrain.

as a function of time. Fluctuations of engine speed are especially dramatic, as a


result of the combined effect of rapidly varying fuel input, governed by the IVS
controller, and equally rapidly changing load torque while the tank rides over
bumps. These rapid engine transients are certainly a challenge, in terms of main-
taining combustion stability and optimum fuel economy, while operating condi-
tions are varying at very high rates. Emissions, especially soot, can also reach
unacceptable levels during periods of extremely high rates of increase of fuel
input. In this light, the design of the controller/injection system module is clearly
a challenge. It should be noted that, although the controller is modeled in detail,
the rest of the injection system is idealized as if it is reacting instantly to the
change of the control signal. Nevertheless, the current version of the simulation
would be a powerful tool for the analysis of fuel injection system design, if an
appropriate model was added to the engine module.
Engine/torque converter interaction is characterized by dramatic variations of
slip between the turbine and pump wheels. During certain intervals, such as
between 20 and 30 sec in the run, the slip is minimal, while during some other
periods the speed ratio becomes very large. If conditions like these persist, they
can become critical from the point of view of T/C fluid cooling. Sprocket speed
498 ASSANIS ET AL.

35000r-_ _--,-_ _----:- ,...-_ _....,...._ _----:-_ _---,

..............................................
· .
g ··· ...
D Bump Course
c ·
.....•.............
··· _.. . .
..
Ien
C-
·· ..

.~
~ 15000 : -: .

110000 .. . . Fla.'. ~errai~ : \ r:::'::::'::::'::


~ ~ V:
eF! 5000 ...........•............. .
·· ..
~

O'--_ _-'--_ _...l-_ _--'-_ _---l..._ _----'_ _----J


o 5 10 15 20 25 30
TIme (s)

Fig.31. Integrated simulation: shaft torques for flat terrain vs. bump course.

is characterized by high-frequency oscillations superimposed over the lower


main frequency. The lower frequency is to a large extent determined by fluctua-
tions of torque provided by the powertrain system. At the same time, the ride
over bumps excites higher frequency track vibrations (see Fig. 35) that translate
into high-frequency fluctuations of sprocket angular velocity. These fluctuations
create transient conditions in the driveline that is unique to a tracked vehicle.
Hence, high rates of change of rotational speeds, in combination with gear iner-
tia and lash, must be considered in durability and noise emission studies of
driveline components.
A comparison of the track tension histories during the high-speed run (pre-
scribed torque) and the 5-MPH run (with integrated powertrain and IVS control-
ler) is shown in Fig. 35. Clearly, the high-speed case produces significantly
higher peak amplitudes for track tension. This may lead to an erroneous predic-
tion on the yielding condition of components in the track system, thus signifi-
cantly affecting the accuracy of fatigue life predictions. Using results obtained
from the 5-MPH run would allow track and suspension element analysis under
much more realistic conditions, thus avoiding unrealistic high peaks and provid-
ing more reliable predictions of both track and overall vehicle system dynamics.
MODELING AND SIMULATION OF Ml ABRAMS TANK 499

30

25 ........

I
a. 20
5
"0
CD
CD
C- 15
OO
CD

s:
CD 10
>

o ""-_---L_ _"---_----'-_ _-L-_----'-_ _--'-_----''--_--'


o 10 20 30 40 50 60 70 80
Time (5)

Fig. 32. Vehicle speed predicted by simulations with and without the powertrain model.

c
o

e~ 0.6
~
II: 0.4

Fig. 33. Rack position history controlled by the Intelligent Vehicle Speed controller over the bump
course.
500 ASSANIS ET AL.

2500,---,-----,----r----,---,---;=:::<::;;===:::::c===j1
- EngineOut
. _. TorqueConvertor Out
- - Transmission Out
• DriveS rocket Out
2000

Time (5)

Fig. 34. Integrated simulation: shaft speeds for bump course.

V. CONCLUSIONS AND FUTURE WORK

This paper addresses three major issues for modeling and simulating a com-
plete tracked vehicle. First, an advanced model is presented for a critical compo-
nent in the vehicle structural system, the track. The new track model captures
both longitudinal and transverse vibrations of the track, using relatively few
modal coordinates. This track model has also been extended to a flexible multi-
body dynamics formulation and implemented into a general multibody dynamics
code, DADS, for simulating tracked vehicles with various configurations.
Through DADS graphical user interface menu choices, track models with vari-
ous levels of fidelity can be constructed and employed to support different simu-
lation goals. Numerical examples have shown that if track dynamics are ne-
glected, the traditional static track model may produce erroneous results for the
track tension and dynamic response of the vehicle. By employing the new track
model, dynamic effects of the track system can be efficiently captured.
Second, a virtual diesel engine model for the M I Abrams tank is developed.
The diesel engine is an intriguing alternative to the current-production gas tur-
bine powerplant, due to the diesel engine's attractive fuel economy, reliability,
MODELING AND SIMULATION OF Ml ABRAMS TANK 501

......:.. :. .:. . .
................ . ., .
.
; . . ~

. .

0
0 3 6 9 12 15 18 21 24 27 30

Time (s)
(a) Without Powertrain

35000

....... .. .... ........ ....


:0-
~

-=-c:
0
'iii .".
c:
Ql
I-
.><
o
e
l-

..... ·1·· ..... ~........;...... "1


. .
0
0 8 18 24 32 40 48 58 64 72 80
Time (s)
(b) With Powertrain

Fig. 35. Track tension predicted by simulations with and without the powertrain model.
502 ASSANIS ET AL.

and maintenance characteristics. The objective was to develop and use a new
tool to explore integration and performance of the proposed powertrain system
in a tracked vehicle. A high-fidelity, flexible engine system simulation has been
developed, based on the phenomenological diesel engine cycle simulation. Tran-
sient capabilities have been integrated with external components, within the hi-
erarchical system simulation structure in MATLAB/Simulink. This software en-
vironment is used to couple the engine system with the driveline. Flexibility of
the new tool has been utilized to create a virtual V-12 diesel system with twin
turbochargers, which is suitable to power the 65-ton main battle tank. The vir-
tual engine system has been analyzed in both conventional and Low Heat Rejec-
tion (LH R) setups. For this purpose, the powertrain has been linked with the
point mass vehicle dynamics model in MATLAB/Simulink, and the tank has
been run over the flat terrain. Transient results illustrate the trade-off between
the higher available exhaust energy in the LHR engine configuration and the
loss of volumetric efficiency due to charge heating in contact with the hot,
insulated combustion chamber walls. Overall effects indicate that the LHR en-
gine provides certain advantages during extended, hard-acceleration runs.
Third, the powertrain model in MATLAB/Simulink has been integrated with
a multibody dynamics model (including the dynamic track model) in DADS.
This has been achieved by developing a novel integration algorithm for running
concurrent, coupled subsystem simulations-under different software applica-
tions-in a distributed computing environment. This algorithm is based on a
predictor-corrector scheme, rather than using a conventional master-slave ap-
proach. This new integration approach has several advantages; it maintains soft-
ware independence, provides a framework for performing efficient simulations,
and requires only a simple implementation for software interfaces.
Using this integration scheme, vehicle/powertrain interactions were investi-
gated during transients associated with the tracked vehicle riding over rough
terrain. In addition, an intelligent vehicle speed controller has been added to the
engine, in order to be able to simulate driver action and maintain desired vehicle
speed over difficult terrain. Results obtained with the fully integrated tank simu-
lation with the intelligent speed controller illustrate the following critical aspects
of engine/driveline/vehicle/track interactions:

• Running the fully integrated tank model allows track and suspension ele-
ment analysis under much more realistic conditions than in case of a pre-
scribed drive-sprocket torque. Thus, predictions of the effect of system
dynamics on driver fatigue, electronics, and mechanical component dura-
bility are more reliable.
• High frequency fluctuations of engine speed and high rates of change of
fueling rates during the run over a bump course are likely to cause fuel
control, combustion stability, emissions, and noise problems.
MODELING AND SIMULATION OF MI ABRAMS TANK 503

• Periods of transient operation of the torque converter, characterized by


large internal slip ratios, could create cooling problems.
The numerical examples presented illustrate coupling effects between the
powertrain system and the vehicle. Results clearly demonstrate the benefits of
using the integrated system simulation.
Finally, a major enhancement of this complete tracked vehicle model would
follow from the inclusion of a track-terrain interface model that captures the
effects of track bridging and a soil pressure-shrinkage relationship. The resulting
model, which is presently being developed, may support rapid estimates of track
vibration response and prove useful for efficient vehicle simulation.

ACKNOWLEDGMENTS

The authors would like to acknowledge the technical and financial support of
the Automotive Research Center (ARC) by the National Automotive Center
(NAC) located within the US Army Tank-Automotive Research, Development
and Engineering Center (TARDEC) in Warren, Michigan. The ARC is a US
Army Center of Excellence for Automotive Research at the University of Michi-
gan, under contract number DAAE07-94-C-R094. Current ARC partner univer-
sities include the University of Alaska-Fairbanks, Clemson University, the Uni-
versity of Iowa, Oakland University, the University of Tennessee, Wayne State
University, and the University of Wisconsin-Madison.
The authors gratefully acknowledge Mike Saxon (formerly of US Army
TACOM) for providing the experimental results; and John Weller (US Army
TACOM), Jim Ajlouny (General Dynamics), and T. C. Lin (General Dynamics)
for providing MI tank modeling data and useful discussions. The authors also
gratefully acknowledge Professor John Moskwa, Scott Munns, and Zachary Ru-
bin of the ARC research team at the University of Wisconsin-Madison for pro-
viding the flexible powertrain simulation structure in MATLAB-Simulink, as
well as the engine ancillary component models and the drive line module used
in this work.

REFERENCES

I. P. Wheeler, Tracked vehicle ride dynamics computer program. SAE Paper No. 770048, SAE.
Warrendale, PA. 1977.
2. M. R. Gamich and T. R. Grimm, Modeling and simulation of a tracked vehicle. In Proc.
of the ASME International Computers in Engineering Conference and Exhibit on Advanced
Automat;on, Vol. 2, 591-600, 1984.
3. M. D. Bennett and P. H. G. Penny. The assessment of tracked vehicle suspensions using
computer simulation techniques. In /Mech Conference Publications /985-5, Vol. CII2I85,
103-117, 1985.
504 ASSANIS ET AL.

4. M. K. McCullough and E. 1. Haug, Dynamics of high mobility track vehicles. ASME Journal
of Mechanisms, Transmissions, and Automation in Design, 85-DET-95:t-8 (1985).
5. A. Dhir and S. Sankur, Ride dynamics of high-speed tracked vehicles: Simulation with field
validation. Vehicle System Dvnamtcs 23:379-409 (1994).
6. J. H. Choi. Usc of Recursive and Approximation Methods in the Dynamic Analysis of Spatial
Tracked Vehicles, Ph.D. thesis, University of Illinois at Chicago, Chicago, IL, 1996.
7. J. P. Wilcox. Documentation for the trackdyne ii program, Technical Report. U.S. Army Engi-
neer Waterways Experiment Station. 1989.
8. J. P. Wilcox. Documentation for the trackdrive program, Technical Report, U.S. Army Engi-
neer Waterways Experiment Station. 1994.
Y.· A. G. Gnlaitsis. Traxion: A model for predicting dynamics track loads in military vehicles.
ASME Journal of vibration, Acoustics. Stress. and Reliability in Design 106:286-291 (1984).
10. S. A. Munn, Computer simulation of powertrain components with methodologies for general-
ized system modeling. M.S, thesis. University of Wisconsin-Madison, Madison, WI, 1996.
II. Z. J. Rubin. S. A. Munns. and J. 1. Moskwa, The development of vehicular powertrain system
modeling methodologies: Philosophy and implementation, SAE Paper 971089, SAE, Warren-
dale. PA. 1997.
12. J. Anthony. J. Moskwa. and E. Danielson. Powertrain simulation of the rnl a l abrams using
modular model components, SAE Paper 980926. SAE, Warrendale, PA, 1998.
13. H. M. Irvine and T. K. Caughey, The linear theory of free vibrations of a suspended cable.
Proc. of the Royal Society, London A-341 :299-315 (1974).
14. N. C. Perkins. Modal interactions in the nonlinear response of elastic cables under parametricl
external cxcnmlon. lnternatlonal Journal of Non-linear Mechanics. 27(2):233-250 (1992).
15. C. Scholar and N. C. Perkins. Efficient vibration modeling of elastic vehicle track systems.
Journal of Sound and Vibration, in press.
16. C. Scholar and N. C. Perkins. Low order vibration models for tracked vehicles. In ASME
Design Engineering Technical Conference. Las Vegas, NV, September 1999. in press.
17. R. R. Craig, Substructure methods in vibration, ASME Journal of Mechanical Design 1178:
207-213, 1995.
18. P, Scshu. Substructuring and component mode synthesis, Shock and Vibration 4(3):J99-210
(1997).
19. C. Scholar and N. C. Perkins. Longitudinal vibration of elastic vehicle track systems. SA£
Papa No. 971090, SAE, Warrendale, PA. 1997.
20. Z. Rubin, S. A. Munns. and J. J. Moskwa, Development of the automotive research center
(arc) powertrain system dynamic models, In Proc. ofASME-ICE Spring Technical Conference,
Vol. 2S-1. Fort Collins, CO, April 1997.
21. D. N. Assanis and J. B. Heywood, Development and use of a computer simulation of the
rurbocompoundcd diesel system for engine performance and component heat transfer studies,
SAl' Paper 860329, SAE, Warrendale, PA, 1986.
22. R. R. Poola, R. Sckar. D. N. Assanis, and G. R. Cataldi, Study of oxygen-enriched combustion
air for locomotive diesel engines, In Proc. of ASME-ICE Fall Technical Conference, Vol. 27-
4, Fairborn, OH, October 1996.
23. N. Watson. A. D. Pilley. and M. Marzouk, Combustion correlation for diesel engine simula-
tion. SAE Paper IW0029, SAE, Warrendale, PA. 1980.
24. M. Tcnnckcs and J. L. Lumley, A First Course in Turbulence. MIT Press, Cambridge, MA,
1972.
25. S. V. Bohac. D. M. Baker. and D. N. Assanis, A global model for steady state and transient
s.i. engine heat transfer studies, SAE Puper 960073, SAE, Warrendale, PA, 1996.
26. Z. S. Filipi and D. N. Assanis, A non-linear. transient. single-cylinder diesel engine simulation
for predictions of instantaneous engine speed and torque, In Proc. of ASME-ICE Spring Tech-
nical Conference. Vol. 28-1, Fort Collins, CO. April 1997.
MODELING AND SIMULATION OF Ml ABRAMS TANK 505

27. G. Zhang, Z. S. Filipi, and D. N. Assanis. A flexible, reccnfiguruble. transient multi-cylinder


diesel engine simulation for system dynamics studies, Mechanics of Structures and Machines
25(3):357-378 (1997).
28. B. W. Millington and E. R. Hartles, Frictional losses in diesel engines. SA£ Paper 680590
SAE Trails .. 77 (1968).
29. Y. Shiao. C.-H. Pan. and J. J. Moskwa. Advanced dynamic spark ignition engine modeling
for diagnostics and control, International Journal of Vehicle Design, 15(6) (1994).
30. N. Watson and M. S. Janota, Turbocharging the Internal Combustion Engine. The Macmillan
Press Ltd.. London, England, 1982.
31. Z, Filipi, D. Assanis, D. Jung, G. Delagrammatikas. J. Liedtke, D, Reyes, D. Rosenbaum, and
A. Sales, Enhancing flexibility and transient capability of the diesel engine system simulation.
In 4th ARC Conference on Critical Technologies for Modeling and Simulation of Ground
Vehicles Ann Arbor. MI, May 1998 (http://arc.engin.llwich.edu/arc/conj98/filipi.pdj).
32. D. Assanis, W. Bryzrk. N. Chalhoub, Z. Filipi. N. Henein, D. Jung. X. Liu. L. Louca. J.
Moskwa, S. Munns, J. Overholt. P. Papalambros, S. Riley. Z. Rubin, P. Sendur, J. Stein, and
G. Zhang. Integration and use of diesel engine, driveline and vehicle dynamics models for
heavy duty truck simulation, SAE Paper 1999-01-0970, SAE, Warrendale, PA, 1999,
33. K. J. Astrom and T. Hagglund. PID Controllers: Theory, Design. and Tuning, 2nd Ed.. Instru-
ment Society of America. Research Triangle Park. NC, 1995.
34. R. K. Jurgen, ed. AUlomolive Electronics Handbook, 2nd Ed., McGraw-Hili Co.. Inc.. New
York, 1999.
35. T. Denton, ed.. Automobile Electrical and Electronic Systems, 3rd Ed.. Edward Arnold, Lon-
don, England, 1997,
36. D. N. Assanis, A. Atreya. C. Borgnakke. W. Bryzik, D. R. Dowling, Z. S. Filipi, S. Hoffman,
S. Homsy, F. Kanafani, K. Morrison, D. Patterson, M. Syrimis. D. Winton, and G. Zhang.
Development of a modular, transient, multi-cylinder diesel engine simulation for system per-
formance and vibration studies. In Proc. of ASME-ICE Spring Technical Conference, Vol. 28-
I, Fort Collins, CO, April 1997.

Received April /999

View publication stats

You might also like