Real Time Simulation For Control Application
Real Time Simulation For Control Application
11:00
*
Abstract
A nonlinear port fuel injected automotive engine model
is developed which can be used for the design of engine
control algorithms, or within an algorithm for real time
control. The model is being validated using a sequential
fuel injection V-6 engine, and includes:
. intake manifold dynamics,
* fuel dynamics, and
. process delays inherent in the four stroke engine
cycle.
The model is compact enough to run in real time and
can be used as an embedded model withim a control algorithm or an observer. Although developed for a specific
engine, the model can be adapted to represent various types
of automotive engines with a limited amount of engine data.
Preliminary validation results look very promising.
Introduction
Constant improvements in microprocessor technology
have led to their increased use in automotive control applications. Controlling the engine has been a major thrust in
this control effort. The main goals of engine control development to date has been to:
. control idle speed,
. constrain the range of air/fuel ratio,
* limit knock, and
* optimize spark advance.
An important area of research goimg on today is total powertrain control. Tsangarides et.al. (1) have simulated total powertrain response using a steady-state engine model
along with a transmission/driveline model. Another name
for this approach would be torque control since managing
torque production and delivery is at the heart of this approach. Some of the overall goals in this approach are to:
. improve shift quality by optimizing engine torque
production and shift timing, and
* improve the efficiency and driver control of the automobile by constraining wheel slip to some optimal
region (commonly known as traction control.
To design control algorithms to meet these goals requires
an automotive engine model that can prodict torque production as a function of throttle angle, fuel injector mass
flow rate, spark advance, and exhaust gas recirculation
* This research is
sponsored under a contract from
General Motors Research Laboratories, Power Systems
Research Department.
t Ph.D. candidate
tt Professor of Mechanical Engineering, Director of Vehicle
Dynamics Laboratory, M.I.T.
341
of the flow of air out of the intake manifold and into the
cylinder. The model does not attempt to predict an accurate instantaneous flow rate through the respective intake
valve. Instead, it predicts an averaged flow which is useful
for fueling calculations. This is sufficient since fuel control
is not instantaneous but rather a stepwise process. So, in
effect, the modeled air flow should be an approximation of
the integrated actual airflow over the period between fuel
) ~{km1
({C)cs
}A}
injections.
To lend confidence to the engine model, a series of experiments are currently being conducted at the General
Motors Research Laboratories, Power Systems Research
Department. A V-S sequential fuel-injected engine was used
in these validation tests. The goals of these tests are to estimate model parameters and validate the overall structure
of the engine model. Tests include both quasi-static and
dynamic throttle, fuel, spark advance, and E.G.R. tests.
= MA
(6)
TC PRI
where MA is the maxmum mass air flow through the throttle body (i.e. maum area and choked flow) and TC is
the throttle characteristic:
Cgl(a) Ath(Ct)
TC =
Model Structures
The engine model can be divided into five sub-systems
which are the:
. throttle body,
* intake manifold,
* fuel injection,
* combustion and torque production, and
* rotational dynamics.
The throttle body is modeled as an instantaneous device
with respect to the throttle angle and the pressure ratio
across the device. The basis of the throttle body model is
the one-dimensional compressible flow equation across an
(7)
Ati.(a,,,,)'
=iJ
+
-~2 D2[1eAt+a)
_ (Den
2 (co(a)
orifice.
\
I
dcos(a0)
n-1{l
d a1)
si(
]
D co(( +a))
a(
Harrington and Bolt (7) have shown that due to manufacturing tolerances, the term cos(ao) becomes cos(a:) where:
mhj= Cd All,){k1[(p)]
1k- i
cos(ac)
coa(O.91 * a0 -2.59)
(9)
Pto
2 \I*
\k+1/
(2)
PRI=
.ck2I [1-(p)
Ca (*.)
a0Io
I9
( '-2 ) #
P6+1
2
]}It
P
>( 2
P. c
)4r;(10
(^+1
(3)
(4)
Pt so P,=if ad = P.
A further simplification can be made if it is assumed that
the discharge coefficient does not have any cross coupled
terms between throttle area and pressure ratio, If this is
the case, then the discharge coefficient can be split into
two uncoupled coefficients, and the mass flow relation can
be written as:
342
To study the dynamics of the intake manifold, the following assumptions are made:
* contents of the manifold obey the Ideal Gas Law,
and Dalton's Law of Nonreacting Mixtures,
* homogeneity of temperature and pressure, and
* complete mixing of air and E.G.R.
Bear in mind that the purpose of the intake manifold model
is to estimate the mass air flow out of the intake manifold
and into the cylinder for fueling calculation. The conservation of mass equations for air and E.G.R. are given as:
ein. t
rt4,
V M T _ VM
__
(11)
dta=
Ttdp
TF(-a
Me)
Pa+ Mr
P]
d;P4
.d
Ml.)
Ma )
J7
M
l
MW)
+
A
Pi
M.
-P
(rhm + Thg")
(,a + regri) + KW
w1l steady-sate +
tll
P(21)
(22)
MV
(16)
28.838 g/mole
air
Mcr
= product.
o
ni Mi
IN
28.884g/nolc.
(17)
produJt
Because these molecular weights are essentially equal, equation (14) reduces to the volumetric efficiency relation of
equation (13) and is not useful in estimating P.. Consider
the consequence of the complete mixng assumption:
'ro
Mao0
male
(20)
(s
trcLneicnt
dtP
(15)
7C02+6.58H20 + 38.82N2.
Solving for the molecular weights yields:
=____
(19)
(14)
C7HI3.16+i0.29(02 + 3.773N2) =
Eni-M
V rni
The first term represents the steady-state volumetric efficiency typically calculated from steady-state test cell data
by measuring P, T,wc, and vJr+ Mih The second term is
the dynamic correction term for volumetric efficiency. The
steady-state volumetric efficiency relation can be inserted
into equation (19) to yield:
'm=
R-T
P+ m
v
T
T K-we-i7v
-W
R T
M; (T
MU
K=
(13)
Pa+
ciency relation:
0I
KV ,T ((Ma
+ irc)
(12)
'RCombining these two equations with the volumetric effi-
(Th
[p P-V
gro = egri + ( -)
R* 72TV
P.
Combining equations (11), (12), and (18) yields the differential equation in P.:
( P-P
)(
Mr
)\
mu1 PI )Ml'
IV02
(18)
IVCI
IVOS
IVC2
time
The intake flow starts when the intake valve opens (02)
and increases as a sinusoid until it reaches a magnitude of
rh., when the previous intake valve closes (IVC1). The flow
is equal to hi until the next intake valve opens (1Y03),
since the # 2 intake valv is the only valve which is open at
this time.The flow then decreases as a sinusoid to zero when
the intake valve closes (IVC2). The cosine function is used
so that the flow remains smooth. Other simpler functions,
such as a linear function, can be used but the first derivative
becomes discontinuous at the beginning and ending of valve
events. The expression for this flow is given as:
{I-coo[t
I1!
in.,=
2
o0,
f;_
1+ c[to's
Q,]},
[t$
'ff
te [txvo3,trc2);
are
negligible during
an
an
additional
seconds.
intake event.
The two nice properties of these flows are that each of the
individual flows are smooth with no abrupt changes, and:
as
(24)
E Ffnzok(t) =mh,
k=1
This analysis, while not an exact representation of the indiv'idual flows out of the intake manifold, gives meaning to
the quantity calculated in equation (11), and the control
engineer can visualize the intake process.
'y
't
1,
*f
if LPW <Sini-.vc
if LPW > zinj-fVC
or:
r,
number of
cylinders. vhrj
*n4j-.i.v.c.-
(27)
(25)
f1e-tdt
(26)
344
RBOMS
Although the dynamic validation results are not complete, the preliminary engine model's predictive capabilities
appear to be quite good. The inclusion of the E.G.R. dynamics in the intake manifold allows the engine model to
predict the mass air flow out of the intake manifold with
various E.G.R. flows. It also shows the errors of calculating
fueling command using a simple speed density calculation.
Figure 6 shows simulation results for a 2% to 10% step in
E.G.R..
.. in.,
characterize, however,
co/
m \
*l
/\
as
*,
by
decreasming
-II
I
ime
time
closes.
}.fTidt
- -
manifold,and
W=
~~~~~~~~~~~/
On the left are the controls, which are throttle angle and
E.G.R. flow rate. On the right are the predictions of percent
E.G.R. into and out of the intake manifolC. The percen,ctge
of E.G.R. can increase dramatically, and E.G.R. valve dyeffect
on
can
have
a
namics
significant
torque production. The effects of spark changes on torque
are modeled as being instantaneous. Peliminary validation
results show this to be essentially true, except that the in-
(28)
--
r -lI
~ri
'4
03U
0,
tim
throttle s
t ime.
time
Nomgnclature
A,,h cross-sectional flow area of throttle body
AFI air/fuel ratio influence on torque
C carbon
Cd throttle plate discharge coefficient
Cdl discharge coefficient (fiuction of throttle angle)
C,m discharge coefficient (function of pressure ratio)
CO carbon monoxide
C02 carbon dioxide
d diameter of throttle rod
D diameter of throttle bore
E.G.R. exhaust gas recirculation
H hydrogen
H20 water
MC hydrocarbons
I moment of inertia
rvCi intake valve closes on cylinder i
lVOi intake valve opens on cylinder i
k ratio of specific heats
Refemees
1. Tsangarides, M. C., W. E. Tobler, C. R. Heermann, "Interactive Computer Simulation of Drivetrain Dynamics",
Proceedings of Noise and Vibration Conference, Traverse
City, Mich., May 1985.
2. Dobner, D. J., "Dynamic Engine Models for Control
Development Part I: Nonlinear and Linear Model Formulation', International Journal of Vehicle Design,1982.
1982.
N2 nitrogen
346