SAE04-454
Feasibility of Reusable Vehicle Modeling:
Application to Hybrid Vehicles
A. Rousseau, P.Sharer, F. Besnier
Argonne National Laboratory
Copyright © 2004 SAE International
ABSTRACT
Many of today’s vehicle modeling tools are good for
simulation, but they provide rather limited support for
model building and management. Setting up a
simulation model requires more than writing down state
equations and running them on a computer. The role of
a model library is to manage the physics of the system
and allow users to share and reuse component models.
In this paper, we describe how modern software
techniques can be used to support modeling and design
activities; the objective is to provide better system
models in less time by assembling these system models
in a “plug and play” architecture. With the introduction of
hybrid electric vehicles, the number of components that
can populate a model has increased considerably, and
more components translates into more drivetrain
configurations. To address these needs, we explain how
users can simulate a large number of drivetrain
configurations. The proposed approach could be used to
establish standards within the automotive modeling
community.
INTRODUCTION
In a world of growing competitiveness, the role of
simulation in vehicle development is constantly
increasing. Because of the number of possible advanced
powertrain architectures — such as hybrid or fuel cell —
that can be employed, the development of the next
generation of vehicles will require accurate, flexible
simulation tools. Such tools are necessary to quickly
narrow the technology focus to those configurations and
components that are best able to reduce fuel
consumption and emissions. The simulation tools must
be flexible enough to encompass a wide variety of
components and drivetrain configurations.
With improvements in computer performance, many
researchers started developing their own vehicle
models. But often, computers in simulation are used only
to “crunch numbers.” Moreover, model complexity is not
the same as model quality. Using wrong assumptions
can lead to erroneous conclusions; errors can come
from modeling assumptions or from data. To answer the
right questions, users need to have the right modeling
tools modeling. For instance, one common mistake is to
study engine emissions by using a steady-state model or
to study component transient behavior by using a
backward model. Indeed, specific component models
and modeling philosophies should be used for specific
applications.
In this article, we describe how a graphical user interface
(GUI), combined with an innovative software
architecture, can be used to support powertrain
modeling. It is important to separate modeling from
simulation: We will focus on component model
management and powertrain building management. that
the paper will address ways in which component model
management involves much more than assigning
specific folders for each component and discuss how
powertrain building management is more complicated
than just manually connecting components together. The
Powertrain System Analysis Toolkit (PSAT) developed
at Argonne National Laboratory will be used to explain
the methodology.
PSAT INTRODUCTION
PSAT [1, 2] is a powerful modeling tool that allows users
to realistically evaluate not only fuel consumption but
also vehicle performance. One of the most important
characteristics of PSAT is that it is a forward-looking
model — meaning that PSAT allows users to model real-
world conditions by using real commands. For this
reason, PSAT is called a command-based model. A
driver model estimates the wheel torque necessary to
achieve the desired vehicle speed. The powertrain
controller then sends real commands to the different
components: throttle for engine, displacement for clutch,
gear number for transmission, or mechanical braking for
wheels to achieve the desired wheel torque. Because
the components react to the commands as they would
under real-world conditions, researchers can implement
advanced component models (based on physics rather
than lookup tables), take into account transient effects
(e.g., engine starting, clutch
engagement/disengagement or shifting), or develop
realistic control strategies (which can be used later to
control hardware).