retard influence on torque production' cT is a constant which is roughly the same for engines with t

1
Karl Hedrick Tomoyuki Kaga Edward A. Lee Pannag
R Sanketi Jose Carlos Zavala Jurado Haiyang Zheng
Automotive Engine Modeling for Control
http//chess.eecs.berkeley.edu
Overview The development of control systems
requires extensive use of models to represent the
behavior of a physical plant. In particular,
automotive systems can be expressed as hybrid
systems in different modeling environments. In
this presentation, we describe an automotive
engine model created in Simulink and further
present an engine model created in Ptolemy,
outlining advantages and disadvantages of the
models and the respective modeling
environments. The final objective of this engine
model is the synthesis of a controller, which
would eventually be implemented on experimental
facilities. The model must be accurate over the
operating ranges of its inputs and be as simple
as possible. The simplicity of model aids the
development of control laws. Simulink Engine
Model
Ptolemy Approach with Hybrid Systems
retard influence on torque production. cT is a
constant which is roughly the same for engines
with the same compression ratio. Fuel
Dynamics Fueling delays are important to the
overall in-cylinder air fuel ratio. When fuel is
injected into the intake ports, part of the fuel
deposits on the intake manifold as a liquid, part
of the fuel vaporizes, and part of the fuel
becomes droplets. A portion of the fuel deposited
on the intake manifold wall will later enter the
air stream and affect the in-cylinder air-fuel
ratio. A model of these dynamics is given by
Why hybrid systems?
Hybrid systems are a natural way to describe the
modal behavior of automotive dynamics and the
corresponding control laws. For example, the
ignition controller has to delicately control the
spark timing of each cylinder of an engine at
different operation modes, such as to achieve the
best usage of fuel or to limit the HC emission.
How to use hybrid systems?
(? is the part of the fuel that enters the
cylinder directly as vapor)
Hybrid systems are modeled with Modal Models,
such that continuous-time models are hierarchical
nested with Finite State Machine as the
middleware to activate and deactivate certain
models.
(Taken from Jay Bartons Thesis)

• Performance Criteria
• Accuracy, for a given set of inputs and initial
  conditions, we would like to know how similar
  the behavior of the model with respect to the
  physical engine is
• Utilized resources. What is the execution time
  and what resources are required to simulate the
  model and execute the control law
• Correctness. The question to answer is whether
  there is any tool inside the model or the
  modeling environment that detects flaws in the
  execution of the simulation

Ptolemy provides an operational semantics for
simulating hybrid systems, which gives a
well-defined and deterministic behavior for
complex interactions between simultaneous events
and continuous dynamics.
We present here a mean value model that comprises
states for the air flow, fuel flow and rotational
speed of the engine. It is particularly suitable
for fuel injection control.
Start-up phase accounts for - Most of HC
emission - Considerable calibration effort
Conclusion
Principal Subsystems

• Comments on differences in models
• Simulink model assumes the existence of a
  throttle angle controller and does not consider
  throttle dynamics at present, whereas the Ptolemy
  model associates second order dynamics with the
  throttle angle.
• Intake manifold air submodel is similar in both
  the models, the only difference being that the
  Simulink uses manifold air as the state, whereas
  the Ptolemy model uses manifold pressure as the
  state.
• Torque generation in Simulink is mean-value and
  simple. One in the Ptolemy model is event-based
  and complex, taking into account the torque
  produced by each cylinder separately.
• Differences in Modeling Environments
• Simulink 1. Well developed software structure to
  interface to an embedded controller 2. Control
  libraries a part of the software 3. Does not
  support Hybrid systems very well. Can give
  non-deterministic delays in stateflow systems. 4.
  Easy user interface.
• Ptolemy 1. Synthesis of embedded controllers
  needs to be developed 2. Good for modeling event
  based systems 3. More control libraries need to
  be developed. 4. Theory for developing
  controllers for hybrid systems is still under
  development.

Air Intake The mass air flow through the intake
manifold is the difference between the intake and
exit
Comparing model with experimental data
For different subsystems, plots with the
differences between the model results and the
experiment results are shown.
Torque Production The indicated torque, Ti, is
modeled as a scaled
function of the mass of air per cylinder
where AFI(?) is the normalized air-fuel ratio
influence on torque production and SPI(?) is the
normalized spark advance or
November 18, 2004