This action might not be possible to undo. Are you sure you want to continue?
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright © 2003 by ABCM OctoberDecember 2003, Vol. XXV, No. 4 / 329
L. C. Passarini and
P. R. Nakajima
Universidade de São Paulo
EESC Escola de Engenharia de São Carlos
13566590 São Carlos, SP. Brazil
luca@sc.usp.br
Development of a HighSpeed
Solenoid Valve: an Investigation of
the Importance of the Armature Mass
on the Dynamic Response
Traditional design criteria for electromagnetic valves are discussed. Performance criteria
for them are also shown. A method for investigating the armature mass importance on the
EFI performance is proposed. This method is based on energy losses of a massspring
damper system (MKsB). It was found a range of values in which the dynamic response can
be improved. Results are verified and discussed.
Keywords: Solenoid fuel injector, mathematical model, electromagnetic actuator
Introduction
Otto Cycle engines with fuel injection systems are equipped
with one or more electromagnetic fuel injectors (EFIs) mounted at
the intake manifold (Fig. 1). The EFI injects the fuel into the intake
valve in a way that the delivered fuel mixes with the intake air,
forming the airfuel (A/F) mixture that will be burned within the
cylinders. The EFI functioning is well known and it can be
explained with a little help of Fig.2. When an electrical driving
command is applied to its terminals, the EFI coil is energized. Then
the armature that is pressed against the valve seat is attracted by the
solenoid poles and moves up against them. This movement causes
an opening at the metering section where the fuel, under pressure,
passes and as soon as it leaves the valve is pulverized into fine
droplets. When the electrical excitation stops, the attracting force
diminishes very quickly because the coil is deenergized and the
armature is pushed against the valve seat by the action of the return
spring, closing the fuel passage. So, it can be said that at the EFI, the
released fuel per pulse Q, is a function of the electrical exciting
pulse width t applied to the solenoid of the EFI.
1
The dynamic response if the EFI solenoid is of primary
importance. For example, an important requirement for small
engines is that the EFI should present good linearity and precision
under the smallest pulses, and with minimum opening and closing
times. In fact, Greiner at al. (1987) affirm that linearity in the small
pulses is especially important for small engines where it is
demanded that fuel metering is done with high precision under the
smallest pulses. These requirements have motivated to investigate
the EFI design, particularly the influence of the armature mass on
the EFI dynamic response, in order to find a way that meets the
performance requirements and improve the EFI dynamic response.
The way by which it was done was by modelling the EFI’s
mechanical system and looking for clues that point to a satisfactory
design.
StateofArt of the Traditional Approaches to Designing
EFIs
According to Kushida (1985), the conditions for optimization of
an EFI are those that relate highspeed response and high power
dissipation. That is, it is necessary to allow high energy input and to
convert this into kinetic energy in an intended direction for
efficiency. This is achieved by:
Paper accepted August, 2003. Technical Editor: Atila P. Silva Freire.
 a) dissipating the heat sufficiently to permit high input of
electrical energy;
 b) concentrating the magnetic flux in an objective magnetic
field;
 c) limiting the generated magnetic force into an objective
direction and
 d) decreasing the moving mass as much as possible.
KajimaKamamura (1995) affirm that high speed operation can
be obtained by actuating
 a) increasing the solenoid force to actuate the armature;
 b) reducing the resisting force due to reactions, and/or;
 c) reducing the weight of moving parts.
It can be noticed that there is little difference between Kushida’s
and KajimaKamamura’s proposals. In this article the approach of
optimizing the EFI performance by decreasing the moving mass as
much as possible will be discussed and it will be demonstrated that
it is not completely true since it is possible to find a particular
moving mass that permit a high performance without necessarily
being the smallest possible.
Subsidies for an Analysis of the EFI ’s Performance:
In order to evaluate and to compare the results better, it becomes
opportune to present some of the criteria generally adopted when
analyzing the EFI performance. Those criteria were defined by the
EFI makers and by the designers of engine electronic control
systems. The main criteria and those that are the most relevant for
this article are these: static flow, dynamic flow, calibration, linearity
and dynamic range.
According to DeGraceBata (1985), the static flow, also called
full opening, is achieved when the EFI is energized with a steady
current. In general, it is measured in g/sec. The dynamic flow
reproduces the EFI operating condition better. It is the fuel flow
delivered when the EFI is pulsed with an electrical signal, usually
measured in milliseconds. This flow is usually given in milligrams
per pulse, or grams per 1,000 pulses. The duration of the pulse must
be given. Fig. 3 shows graphically these two definitions. It is
important to point out that the value of the static flow coincides with
that obtained by the slope of the dynamic flow line.
The static flow and the dynamic flow are important because, in
practice, they are the two only variables used to control the amount
of fuel released by the EFI of current control systems of internal
combustion engines. In fact, the best way to settle (or to calibrate)
an EFI, in accordance to De GraceBata (1985), is, from the static
flow, to take only one point from the dynamic fuel flow (in Fig. 3, it
is represented the value used in most cases, i.e., 2.5 msec at the time
L. C. Passarini and P. R. Nakajima
/ Vol. XXV, No. 4, OctoberDecember 2003 ABCM 330
axis) and to adjust the spring loading (preset) of the armature by
means of a locking screw until the desired flow is obtained at that
point. The spring loading adjusts the opening and closing time of an
EFI, but it does not affect the static flow.
According to DeGraceBata (1985) and Garret (1990), during
the operation of an EFI, the electromechanic interactions promote
significant delays (as that between the instant when the exciting
pulse begins and the instant at which the EFI begins releasing the
fuel; or as the delay between the finish of the exciting pulse and the
moment at which the EFI stops releasing fuel). The delays promote
a displacement of the point where the fuel flow is zero in relation to
the origin of the time pulse width axis (Fig. 3) creating the
appearance of a “dead zone ” popularly known as offset .
In agreement with Garret (1990), for an electronic control meter
the fuel supply to the engine with precision, there must be a linear
relationship between the quantity of fuel delivered, Q and the pulse
width applied, tҏover the full delivery range. As DeGraceBata
(1985) say, linearity is usually expressed in terms of the lowest
pulse width or fuel flow at which Flow vs. Input Pulse Width lies on
a straight line within a specified percentage of the theoretical flow
(Fig. 3). About this there is not a strict consensus: according to
Toyoda et al. (1982) the EFI linearity should stay within ±2%. De
GraceBata (1985) are more tolerant and extend this tolerance to
±2.5%. Matsubara et al.(1986) prefer using ±1.5%. For this article,
the criterion of ±2% was adopted. Linearity in the smallest pulses is
specially important because, as Greiner et al. (1987) say, the
increasing use of injection in smaller engines demands high
metering accuracy with minimal pulse times. This in turn requires
minimizing injector opening and closing times. Linearity of
injection improves with both the ratio of the actuation force to the
mass of the delivery valve (armature) and the appropriate matching
with it of the electrical and time constants of the solenoid circuit.
Since the time required to close the injector is a function of the
current in the solenoid at the point at which its circuit is broken,
delivery errors are difficult to avoid when the pulse width is shorter
than that necessary for the solenoid current to reach a stable value.
The dynamic range is defined as the ratio of maximum to
minimum duration of injection (Fig. 3). Garrett (1990) defines the
minimum linear duration of injection, Q
min
, as the difference
between the shortest linear pulse width and the offset. In order to
maximize the dynamic range, EFI makers fight to keep Q
min
as small
as practicable because the minimum duration of injection is
determined to a major degree by injector design and the consequent
offset of the delivery characteristics (Garret, 1990). So, it can be
said that the moving mass affects directly Q
min
. Matsubara et al.
(1986) affirm that reducing the armature weight and a higher spring
preset load are effective to get wide dynamic control range. On the
other hand, Garret (1990) says that the maximum duration of
injection, Q
max
, is primarily a function of the rotational speed of the
engine and whether the EFI is fired once or twice per revolution.
Because of the non linear flow characteristics at both ends, the EFI
can be utilized only between Q
min
e Q
max
where linearity is within
the tolerance adopted, whatever it is (Toyoda et al., 1982).
yoke
coil
adjustment
mechanism
seat
armature
return spring
stop
metering section
terminals
outlet
power
inlet
needle
seat section
Figure 1. A set of 3 EFIs (pointed by the light arrow) mounted at an intake manifold. An EFI injects fuel directly into the intake valve.
Figure 2. Nomenclature and schematic drawing of a generic EFI.
Development of a HighSpeed Solenoid Valve: an Investigation of the …
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright © 2003 by ABCM OctoberDecember 2003, Vol. XXV, No. 4 / 331
Figure 3. A characteristic curve of an EFI: Fuel Flow x Time Pulse Width.
Nomenclature
B = spring’s damping coefficient
F = force
j = \1
K
s
= spring’s stiffness elasture coefficient
M = armature’s mass
q = ratio M·K
s
/B
2
Q = mass fuel flow
s = Laplace’s complex variable
t = time
T = period between exciting pulses
x = armature’s displacement
e
n
= undamped natural frequency
t = exciting pulse width that is applied to the EFI ’s solenoid
coil
, = damping ratio
Index
min = minimum
mag = magnetic
max = maximum
ss = steady state
Modelling the EFI’s Mechanical System
As mentioned above, the way an EFI functions is relatively
simple, however the complexity of the mathematical description of
the physical actions and of the equations involved is considerable.
Several forces and effects appear and disappear during the EFI
functioning as, for example, a shock between the armature and the
valve seat during the closure of the EFI or a shock between the
armature and the limiter at the end of the course during the EFI
opening. If some effects are not significant at some times, at other
times they contribute significantly. Authors like SmithSpinweber
(1980), KaridisTurns (1982), MacBain (1985), Kushida (1985),
PawlackNehl (1988), Lesquesne (1990a), YuanChen (1990),
KawaseOhdachi (1991), Ohdachi et al. (1991), Kajima (1992 and
1993), Passarini (1993), Rahman et al. (1996), Ando et al. (2001),
SzenteVad (2001) and PassariniPinotti Jr. (2003) diverge when
describing the armature movement of the of the EFI and as to which
of the above mentioned phenomena should be considered and which
should be neglected. It seems that most of these authors consider it
sufficient to calculate precisely the magnetic force attraction, F
mag
to
accurately define the armature movement because basically their
mathematical model is described by only three forces, magnetic,
elastic (Hook’s law) and viscous as shown in Eq. (1):
x . B x . K F x . M
s mag
÷ ÷ = (1)
For a more complete discussion on modelling an EFI’s
mechanical system, please refer to PassariniPinotti Jr. (2003). In
this article this traditional basically linear massspringdamper
(MKsB) governed by Eq. (1) and represented in Fig.4 was adopted
because:
Figure 4. Linear MKsB model adopted to analyze the EFI mechanical
system.
a) it allows the use of Laplace’s transforms and consequently,
the application of the rootlocus method (very much used in
the analysis of linear systems);
b) even when such a complete and sophisticated model is not
used, the chosen model supplies satisfactory results
[because only a part of the problem is studied: that in which
the nonlinearities are small. It will be shown latter that the
simulation done with a much more complex and complete
model (Passarini, 1993) has confirmed the results obtained
with this simpler approach ].
By hypothesis it was assumed that all elements are pure and
ideal. The characteristic equation of a MKsB system is shown in Eq.
(2):
M s
2
+ B s + K
s
= 0 . (2)
The actual spring exhibits the characteristics illustrated in the
graph of Fig.5. The linearized model was obtained according to
Passarini (1993) and the parameters K
s
and B were found.
Analysis of the Problem
The rootlocus method is very powerful in the dynamic analysis
of linear systems (Ogata, 1993). Applying it to the characteristic
equation Eq. (2), the influence of M on the transient response of the
MKsB system could be investigated. For convenience, the ratio q
was defined as:
q =
M K
s
B
2
(3)
L. C. Passarini and P. R. Nakajima
/ Vol. XXV, No. 4, OctoberDecember 2003 ABCM 332
Figure 5. Elastic characteristic of the spring used in the EFI prototype
montage.
Figure 6 shows the rootlocus drawn as a function of the ratio q.
It can be verified that:
1. MKsB systems with very large moving mass (q » 0.25)
oscillate very much and are sluggish. Therefore ,they are
not appropriate for application in this case.
2. MKsB systems with very small moving mass (q « 0.25) do
not oscillate, however, when excessively decreasing the
moving mass, i. e. q ÷ 0, will not bring any significant gain
either on the system time constant or response speed of the
system
~
K
s
B

\

.
, due to the influence of the pole located at
s = ÷
K
s
B
. Therefore, this dismantles the previous idea:
“the smaller the magnitude of the moving mass, the better”.
3. The best response and therefore, the best adjustment of M
appears to occur when the roots are produced close to the
breakaway point on the real axis located at s = ÷2
K
s
B
(when q = 0.25). The MKsB system exhibits a behavior that
is practically nonoscillatory and will still have one of the
smallest setting times possible.
Confirming the Results Obtained with the MKsB Root
Locus
To confirm the result obtained with the rootlocus method, the
dynamic impulsive response of the MKsB system was used to
analyze the system performance. This analysis of the impulsive
response is quite useful because it represents very well the system
behavior under impacts.
Figure 6. The MKsB system rootlocus as a function of the ratio q.
When an EFI is opening or closing its armature collides several
times against the stopper or against the valve seat, respectively, over
and over again until all the mechanic energy present in it is
dissipated (please, refer to Fig. 7). This phenomenon is known as
armature bounce or rebound. Although rebounds also happen at the
shocks of the armature with the stop, its influence is important only
when the moving mass is very large, i. e., when q >0.75.
While the armature rebounds one or more times, some amount
of released fuel is affected, mainly during the closing of the EFI
(please refer again to Fig. 7). The surrounding medium and the
presence of vibrations introduce randomness to the rebound of the
armature. For that reason, the shocks are not uniform but randomly
and so, its influence on the fuel flow is not systematic as was
thought. This fact contributes to decrease the EFI linearity and
repeatability, mainly during operation under short pulse widths. In
fact, DeGraceBata (1986) affirm that the main cause for deviations
of the linearity in short pulse widths are the variations at the opening
and closing times. These authors observed that:
1) .significant flow linearity errors can be found even at a
pulse width beyond the time when the armature is fully
open and the armature bounce has subsided, and
2) this flow nonlinearity can be directly correlated to closing
of the armature alone (i.e., opening times are not affected).
In fact, the simulation shown in Fig. 7 confirmed this last
observation.
In order to improve this EFI transient response the criterion used
was, by taking a impulsive response of a MKsB system with null
mass as reference, to determine the instant when 99% of its
mechanical energy was dissipated. It was found that this instant
corresponds to t ~ 2.3·B/K
s
. After this, the dissipated power within
this period of time was computed for the range 0 < q < 2. The graph
is shown in Fig.8. It can be noticed that out of the area in which
q ~ 0, the relative dissipated power stays practically constant and
very close to 100% when 0.2 < q < 0.23. From this range above, the
dissipated power begins to drop, confirming our expectation from
the root locus. According to this criterion, the best q ratio occurred
when q = 0.21. Still examining Fig. 8, the detail shows the root
locus when 0.2 < q < 0.23
Development of a HighSpeed Solenoid Valve: an Investigation of the …
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright © 2003 by ABCM OctoberDecember 2003, Vol. XXV, No. 4 / 333
Figure 7. EFI’s time response: armature displacement and the corresponding instantaneous fuel flow versus time (t = 1 msec and T = 9 msec). (a) M =
354 mg (q = 0.19), (b) M = 1500 mg (q = 0.81). Notice that in the fuel flow curve (b) shown at right. The exceeding amount of released fuel is almost
exclusively due to the armature bouncing against the valve seat. In addition to this, observe how the EFI operation time jumps from ~3,5 msec to ~6,0
msec.
Results:
All the previous theoretical analysis led us to obtain the
following results:
Figure 8. Dissipated power graph (compared to the MKsB system when
M = 0) as a function of ratio q. The detail shows the rootlocus with the
roots that produced the indicated landing.
1) the root locus method pointed to an optimum ratio q
optimum
close to 0.25;
2) a refinement came when examining the dissipation of the
mechanical energy present in the MKsB system relative to
the KsB (when M = 0). This criterion revealed a q
optimum
=
0.21. It could be considered that 0.20 < q
optimum
< 0.23 since
at this range the power dissipation curve stays practically
constant and closely to 100%.
Verification of the Results
The results described above were verified under simulation,
since the model used is quite reliable (Passarini, 1993). The
simulation was applied to a wellknown EFI prototype. The exciting
pulse width was t = 1 ms with a T = 9 ms period long enough to
guarantee that the EFI would come back to its rest condition. Fig. 7
shows the EFI response for two different constructive conditions,
one using a armature which produced q ~ 0.19. In the other, we used
a heavier moving mass very close to those found in the traditional
commercial models (q ~ 0.80). Observe that the simulation has
confirmed and even illustrated the expectation that rebounds
contribute to liberate extra fuel. Fig.9 exhibits the contribution of
the armature mass on the amount of released fuel in the same
appraised situation (t = 1ms, T = 9ms). Finally, Fig. 10 highlights
the loss of the linearity as a function of the increase of the armature
mass (t = 1ms, T = 9ms). It is interesting to note that, the 2% limit is
reached at q ~ 0.24 while the 1.5% limit is reached when q ~ 0.20.
L. C. Passarini and P. R. Nakajima
/ Vol. XXV, No. 4, OctoberDecember 2003 ABCM 334
Figure 9. The fuel released by the EFI up to the instant that the armature
touches the valve seat is practically constant, however systems with large
moving mass have more difficulty in dissipating all the mechanical energy
at the stopping shock. Because of the successive armature shocks
against the valve seat, some extra quantity of fuel is released promoting
the loss of accuracy of the EFI and consequently a loss of linearity.
The following Table 1 shows the physical characteristics of the
studied prototype.
Table 1. Some characteristics of the EFI prototype.
Coil Characteristics
 excitation pulse (voltage mode)
 electrical resistance
 inductance
 copper wire diameter (AWG)
 number of coil turns
11.88 V x 1 msec
0.9348 (ohms)
0.5057 (mH)
0.4 (AWG 26)
225
Mechanical Characteristics
 return spring
 spring coefficient (Ks)
 damping coefficient (B)
 surrounding medium
 material (armature and yoke)
3313 (N/m)
2,47 (Ns/m)
kerosene
SAE 405
Figure 10. Verification of the influence of the armature mass on linearity
(when tҏ= 1 ms and T = 9 ms).
Conclusions
From the above results, it can be concluded that:
1) the adopted linear MKsB model was relatively suitable to
analyze the physical problem, even though it is known that
an EFI is complex and exhibits several nonlinearities;
2) significant flow nonlinearity that can be directly correlated
to closing of the armature alone can be significantly reduced
if the choice of moving mass and return spring
characteristics respects the following relationship:
0.2 <
M K
s
B
2
< 0. 23
This guarantees that the dissipation of power will be effective
improving the EFI linearity at shorter pulses. Another consequence
is that the EFI operation time is relatively reduced (please, refer to
Fig. 7) and so, the EFI dynamic range is increased. This is
particularly interesting for those who want to build up an injection
strategy.
Although the exciting pulse is frequently related to the released
fuel flow, it is significantly shorter than the EFI operation time as it
could be seen at Fig. 7. It means that the engine control system
designer should keep in mind that as the available time for fuel
injection is a function of the rotational speed. At high engine speed
conditions, this difference could become critical and some transient
problems, as engine instability or poor transient response, could
prejudice the engine performance.
Acknowledgment
We would like to register our gratefulness to FAPESP for the
financial support given to our IC project, process n.97/081424.
References
Ando, R.; Koizumi, M.; Ishikawa, T., 2001, “Development of a
Simulation Method for Dynamic Characteristics of Fuel Injector”, IEEE
Transactions on Magnetics, v. 37, n. 5, pp. 37153718.
DeGrace, L. G., Bata G.T., 1985, “The Bendix DEKA¹ Fuel Injector
SeriesDesign and Performance” SAE Technical Papers Series, n. 850559
(SP609), pp. 5763.
Garret, K., 1990, “Petrol Injector and Exhaust Catalyst Developments”,
Automotive Engineer, v. 15, n. 3, pp. 6263.
Greiner, M., Romann, P., Steinbrenner, U., 1987, “Injectors
Developments Examined”, Automotive Engineering, v. 95 n. 2, pp. 4347.
Kajima, T., 1993, “Development of a HighSpeed Solenoid Valve:
Investigation of the Energizing Circuits”, IEEE Transactions on Industrial
Electronics, v. 40, n. 4, pp. 428435.
Kajima, T.; Kawamura, Y., 1995, “Development of a HighSpeed
Solenoid Valve: Investigation of Solenoids”, IEEE Transactions on
Industrial Electronics, v. 42, n. 1, pp. 18.
Kajima, T.; Nakamura, Y.; Sonoda, K., 1992, “Development of a High
Speed Solenoid Valve: Investigation of the Energizing Circuits”,
Proceedings of the 1992 International Conference on Industrial Electronics,
Control, Instrumentation and Automation ”, v. 01, pp. 564569.
Karidis, J.P., Turns, S.R., 1982, “Fastacting Electromagnetic Actuators:
Computer Model Development and Verification”, SAE Technical Papers
Series, n. 820202, pp. 1125.
Kawase, Y.; Ohdachi, Y., 1991, “Dynamic Analysis of Automotive
Solenoid Valve Using Finite Element Method”, IEEE Transactions on
Magnetics, New York, v. MAG27, n. 5, pp.39393942.
Kushida, T., 1985, “High Speed Powerful and Simple Solenoid Actuator
“DISOLE” and Its Dynamic Analysis Results”, SAE Technical Papers
Series, n. 8503763, pp.3. 1273.136.
Lesquesne, B., 1990a, “Dynamic Model of Solenoids Under Impact
Excitation, Including Motion and Eddy Currents”,.IEEE Transactions on
Magnetics, New York, v. MAG26, n. 2, pp. 11071116.
Development of a HighSpeed Solenoid Valve: an Investigation of the …
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright © 2003 by ABCM OctoberDecember 2003, Vol. XXV, No. 4 / 335
Lesquesne, B., 1990b, “Fastacting, Longstroke Solenoids With Two
Springs”, IEEE Transactions on Industry Applications, New York, v. 26,
n. 5, pp. 848856.
MacBain, J.A., 1985, “Solenoid Simulation With Mechanical Motion”,
International Journal for Numerical Methods in Engineering, New York,
v. 21, pp. 1318.
Matsubara, M., Ando, T., Takada, S., Takeuchi, H., 1986 “Aisan Fuel
Injector for Multipoint Injection System”, SAE Technical Papers Series,
n. 860486, pp. 125130.
Passarini, L.C., 1993, “Análise e Projeto de Válvulas Eletromagnéticas
Injetoras de Combustível: Uma Nova Proposta”, Doctoral thesis,
Universidade de São Paulo, EESC, São CarlosSP, Brasil.
Passarini, L.C.; Pinotti Jr., M., 2003, “A New Model for Fast Acting
Electromagnetic Fuel Injector Analisys and Design”, Journal of the Brazilian
Society of Mechanical Sciences, v.25, n. 3, pp.97108.
Pawlack, A.M.; Nehl, T.W., 1988, “Transient Finite Modelling of
Solenoid Actuators: the Coupled Power Electronics, Mechanical, and
Magnetic Field Problem”, IEEE Transactions on Magnetics, v. MAG24,
n. 1, pp. 270273.
Ohdachi, Y.; Kawase, Y.; Murakami, Y.; Inaguma, Y., 1991, “Optimum
Design of Dynamic Response in Automotive Solenoid Valve”, IEEE
Transactions on Magnetics, v. 27, n. 6, pp. 52265228.
Ogata, K., 1993, “Engenharia de Controle Moderno”, 2a.Ed., Editora
Prentice/Hall do Brasil, Rio de Janeiro, Brasil.
Rahman, M.F.; Cheung, N.C.; Lim, K.W., 1995, “Position Estimation in
Solenoid Actuators”, Conference Record of the 1995 IEEE Thirtieth IAS
Annual Meeting (IAS ’95), 1995 Industry Applications Conference, v. 1,
pp. 476483.
Szente, V.; Vad, J., 2001, “Computational and Experimental
Investigation on Solenoid Valve Dynamics”, Proceedings 2001 IEEE/ASME
International Conference on Advanced Intelligent Mechatronics, v. 1,
pp. 618623.
Toyoda, T., Inoue, T., Aoki, K., 1982, “Single Point Electronic Injection
System” SAE Technical Papers Series, n. 820902, pp. 8387.
Yuan, K.; Chen, S.A, 1990, “New Algorithm for Coupled Solutions Of
Electric, Magnetic, and Mechanical System in Dynamic Simulation of
Solenoid Actuators”, IEEE Transactions on Magnetics, v. MAG26, n. 3, pp.
11891197.
yoke coil adjustment mechanism terminals seat section power inlet outlet metering section return spring stop armature needle Figure 1. Garret (1990) says that the maximum duration of injection. As DeGraceBata (1985) say. 1982). 3). the increasing use of injection in smaller engines demands high seat metering accuracy with minimal pulse times. there must be a linear relationship between the quantity of fuel delivered. 330 / Vol. (1982) the EFI linearity should stay within ±2%. it can be said that the moving mass affects directly Qmin. (1987) say. Nakajima axis) and to adjust the spring loading (preset) of the armature by means of a locking screw until the desired flow is obtained at that point. On the other hand. the EFI can be utilized only between Qmin e Qmax where linearity is within the tolerance adopted. Input Pulse Width lies on a straight line within a specified percentage of the theoretical flow (Fig.L. Q and the pulse width applied. Figure 2. For this article. OctoberDecember 2003 ABCM . linearity is usually expressed in terms of the lowest pulse width or fuel flow at which Flow vs. Since the time required to close the injector is a function of the current in the solenoid at the point at which its circuit is broken. Passarini and P. De GraceBata (1985) are more tolerant and extend this tolerance to ±2. R. Nomenclature and schematic drawing of a generic EFI. during the operation of an EFI. Garrett (1990) defines the minimum linear duration of injection. No. This in turn requires minimizing injector opening and closing times. as Greiner et al. About this there is not a strict consensus: according to Toyoda et al. is primarily a function of the rotational speed of the engine and whether the EFI is fired once or twice per revolution. 1990). (1986) affirm that reducing the armature weight and a higher spring preset load are effective to get wide dynamic control range. The spring loading adjusts the opening and closing time of an EFI. delivery errors are difficult to avoid when the pulse width is shorter than that necessary for the solenoid current to reach a stable value. An EFI injects fuel directly into the intake valve. A set of 3 EFIs (pointed by the light arrow) mounted at an intake manifold. Linearity in the smallest pulses is specially important because. Because of the non linear flow characteristics at both ends. So.5%.. the criterion of ±2% was adopted. 3) creating the appearance of a “dead zone ” popularly known as offset . Qmin. Linearity of injection improves with both the ratio of the actuation force to the mass of the delivery valve (armature) and the appropriate matching with it of the electrical and time constants of the solenoid circuit. According to DeGraceBata (1985) and Garret (1990). or as the delay between the finish of the exciting pulse and the moment at which the EFI stops releasing fuel). In order to maximize the dynamic range. over the full delivery range. The delays promote a displacement of the point where the fuel flow is zero in relation to the origin of the time pulse width axis (Fig. C. Matsubara et al. Qmax. the electromechanic interactions promote significant delays (as that between the instant when the exciting pulse begins and the instant at which the EFI begins releasing the fuel.5%. The dynamic range is defined as the ratio of maximum to minimum duration of injection (Fig.(1986) prefer using ±1. as the difference between the shortest linear pulse width and the offset. 4. but it does not affect the static flow. 3). In agreement with Garret (1990). for an electronic control meter the fuel supply to the engine with precision. XXV. EFI makers fight to keep Qmin as small as practicable because the minimum duration of injection is determined to a major degree by injector design and the consequent offset of the delivery characteristics (Garret. Matsubara et al. whatever it is (Toyoda et al.
the application of the rootlocus method (very much used in the analysis of linear systems). (2003). Soc. a shock between the armature and the valve seat during the closure of the EFI or a shock between the armature and the limiter at the end of the course during the EFI opening. (2): M s 2 B s Ks 0. elastic (Hook’s law) and viscous as shown in Eq. 1993) has confirmed the results obtained with this simpler approach ]. & Eng. Linear MKsB model adopted to analyze the EFI mechanical system. 1993). (2001).Development of a HighSpeed Solenoid Valve: an Investigation of the … should be neglected. Kushida (1985). Authors like SmithSpinweber (1980).x Fmag K s . x (1) For a more complete discussion on modelling an EFI’s mechanical system. Several forces and effects appear and disappear during the EFI functioning as. MacBain (1985). Copyright The actual spring exhibits the characteristics illustrated in the graph of Fig. b) even when such a complete and sophisticated model is not used. PawlackNehl (1988). the influence of M on the transient response of the MKsB system could be investigated. (2) Modelling the EFI’s Mechanical System As mentioned above. (1) and represented in Fig. It seems that most of these authors consider it sufficient to calculate precisely the magnetic force attraction. Fmag to accurately define the armature movement because basically their mathematical model is described by only three forces. Sci. If some effects are not significant at some times. a) it allows the use of Laplace’s transforms and consequently. Ando et al. Analysis of the Problem The rootlocus method is very powerful in the dynamic analysis of linear systems (Ogata. the chosen model supplies satisfactory results [because only a part of the problem is studied: that in which the nonlinearities are small. SzenteVad (2001) and PassariniPinotti Jr. Vol. (1): M. Applying it to the characteristic equation Eq. however the complexity of the mathematical description of the physical actions and of the equations involved is considerable. The linearized model was obtained according to Passarini (1993) and the parameters Ks and B were found.4 was adopted because: Figure 3. magnetic. (1996). A characteristic curve of an EFI: Fuel Flow x Time Pulse Width. Lesquesne (1990a). In this article this traditional basically linear massspringdamper (MKsB) governed by Eq. (2). for example. YuanChen (1990). Rahman et al. Kajima (1992 and 1993). x B. It will be shown latter that the simulation done with a much more complex and complete model (Passarini. the way an EFI functions is relatively simple. The characteristic equation of a MKsB system is shown in Eq. Nomenclature B = spring’s damping coefficient F = force j = 1 Ks = spring’s stiffness elasture coefficient M = armature’s mass q = ratio M·Ks/B2 Q = mass fuel flow s = Laplace’s complex variable t = time T = period between exciting pulses x = armature’s displacement n = undamped natural frequency = exciting pulse width that is applied to the EFI ’s solenoid coil = damping ratio Index min = minimum mag = magnetic max = maximum ss = steady state Figure 4. Ohdachi et al. please refer to PassariniPinotti Jr. KawaseOhdachi (1991). 4 / 331 . (2003) diverge when describing the armature movement of the of the EFI and as to which of the above mentioned phenomena should be considered and which J. (1991). No. For convenience. the ratio q was defined as: q M Ks B2 (3) 2003 by ABCM OctoberDecember 2003. of the Braz. at other times they contribute significantly. KaridisTurns (1982). Passarini (1993).5. of Mech. By hypothesis it was assumed that all elements are pure and ideal. XXV.
21. 2. Figure 6 shows the rootlocus drawn as a function of the ratio q. will not bring any significant gain either on the system time constant or response speed of the Ks system .3·B/Ks. its influence is important only when the moving mass is very large. its influence on the fuel flow is not systematic as was thought. OctoberDecember 2003 . e. i. After this. mainly during the closing of the EFI (please refer again to Fig. over and over again until all the mechanic energy present in it is dissipated (please. From this range above. MKsB systems with very large moving mass (q » 0.. It was found that this instant corresponds to t 2. Still examining Fig. the detail shows the rootlocus when 0. R.25). The graph is shown in Fig. This analysis of the impulsive response is quite useful because it represents very well the system behavior under impacts. however.. confirming our expectation from the root locus. 8. For that reason. 7).2 < q < 0. Therefore . This phenomenon is known as armature bounce or rebound. The best response and therefore. It can be noticed that out of the area in which q 0. the better”.75. In fact. and 2) this flow nonlinearity can be directly correlated to closing of the armature alone (i. by taking a impulsive response of a MKsB system with null mass as reference. XXV. While the armature rebounds one or more times. when q >0. Nakajima Figure 5. respectively. Although rebounds also happen at the shocks of the armature with the stop. the best adjustment of M appears to occur when the roots are produced close to the K 2 sB breakaway point on the real axis located at s (when q = 0. MKsB systems with very small moving mass (q « 0. s Ks B . DeGraceBata (1986) affirm that the main cause for deviations of the linearity in short pulse widths are the variations at the opening and closing times. mainly during operation under short pulse widths. the relative dissipated power stays practically constant and very close to 100% when 0.e.8. According to this criterion. refer to Fig. The MKsB system rootlocus as a function of the ratio q. This fact contributes to decrease the EFI linearity and repeatability. Therefore. some amount of released fuel is affected.25) do not oscillate. q 0. the dynamic impulsive response of the MKsB system was used to analyze the system performance.significant flow linearity errors can be found even at a pulse width beyond the time when the armature is fully open and the armature bounce has subsided. It can be verified that: 1. In fact. opening times are not affected). e. when excessively decreasing the moving mass. In order to improve this EFI transient response the criterion used was.they are not appropriate for application in this case. When an EFI is opening or closing its armature collides several times against the stopper or against the valve seat.L.23. the dissipated power within this period of time was computed for the range 0 < q < 2. the best q ratio occurred when q = 0. due to the influence of the pole located at B Figure 6. These authors observed that: 1) . 7 confirmed this last observation.23 ABCM Confirming the Results Obtained with the MKsB RootLocus To confirm the result obtained with the rootlocus method. 7). 3. to determine the instant when 99% of its mechanical energy was dissipated. the simulation shown in Fig. 332 / Vol. Elastic characteristic of the spring used in the EFI prototype montage. i. the dissipated power begins to drop.25) oscillate very much and are sluggish. No. C. the shocks are not uniform but randomly and so. The MKsB system exhibits a behavior that is practically nonoscillatory and will still have one of the smallest setting times possible. The surrounding medium and the presence of vibrations introduce randomness to the rebound of the armature. Passarini and P.2 < q < 0. 4. this dismantles the previous idea: “the smaller the magnitude of the moving mass.
10 highlights the loss of the linearity as a function of the increase of the armature mass ( = 1ms.5 msec to ~6. observe how the EFI operation time jumps from ~3. In the other. Soc. 1993).81).20. (a) M = 354 mg (q = 0. 2) a refinement came when examining the dissipation of the mechanical energy present in the MKsB system relative to the KsB (when M = 0). Sci. This criterion revealed a qoptimum = 0. The detail shows the rootlocus with the roots that produced the indicated landing.0 msec. Fig. 7 shows the EFI response for two different constructive conditions. T = 9ms). 4 / 333 .5% limit is reached when q ~ 0. It could be considered that 0. EFI’s time response: armature displacement and the corresponding instantaneous fuel flow versus time ( = 1 msec and T = 9 msec). J. Fig. Verification of the Results The results described above were verified under simulation. Notice that in the fuel flow curve (b) shown at right. Vol. It is interesting to note that. The exceeding amount of released fuel is almost exclusively due to the armature bouncing against the valve seat.80). Copyright 2003 by ABCM OctoberDecember 2003. of the Braz. Observe that the simulation has confirmed and even illustrated the expectation that rebounds contribute to liberate extra fuel. T = 9ms).21. Figure 8. one using a armature which produced q ~ 0.9 exhibits the contribution of the armature mass on the amount of released fuel in the same appraised situation ( = 1ms. Results: All the previous theoretical analysis led us to obtain the following results: 1) the root locus method pointed to an optimum ratio qoptimum close to 0. (b) M = 1500 mg (q = 0. the 2% limit is reached at q ~ 0.19.Development of a HighSpeed Solenoid Valve: an Investigation of the … Figure 7.20 < qoptimum < 0. Fig. The exciting pulse width was = 1 ms with a T = 9 ms period long enough to guarantee that the EFI would come back to its rest condition.25. XXV. Dissipated power graph (compared to the MKsB system when M = 0) as a function of ratio q. In addition to this.24 while the 1. of Mech. we used a heavier moving mass very close to those found in the traditional commercial models (q ~ 0. since the model used is quite reliable (Passarini.23 since at this range the power dissipation curve stays practically constant and closely to 100%. Finally. No. & Eng.19). The simulation was applied to a wellknown EFI prototype.
IEEE Transactions on Industrial Electronics. “High Speed Powerful and Simple Solenoid Actuator “DISOLE” and Its Dynamic Analysis Results”. 1982. U. Automotive Engineering. refer to Fig. “Fastacting Electromagnetic Actuators: Computer Model Development and Verification”.. pp. Garret.. Y. It means that the engine control system designer should keep in mind that as the available time for fuel injection is a function of the rotational speed. could prejudice the engine performance.3. pp. New York.136. S. This guarantees that the dissipation of power will be effective improving the EFI linearity at shorter pulses. n.97/081424. 1990a. Coil Characteristics 11. however systems with large moving mass have more difficulty in dissipating all the mechanical energy at the stopping shock. 8503763. Passarini and P. n. Nakajima Conclusions From the above results. 1992. Ishikawa.. K. pp. “Development of a HighSpeed Solenoid Valve: Investigation of Solenoids”. Steinbrenner. n.4 (AWG 26) copper wire diameter (AWG) 225 number of coil turns Mechanical Characteristics return spring 3313 (N/m) spring coefficient (Ks) 2. XXV. Kushida. 1125. Automotive Engineer. Nakamura. Greiner. P.9348 (ohms) electrical resistance 0. process n.. “Petrol Injector and Exhaust Catalyst Developments”. IEEE Transactions on Magnetics. Although the exciting pulse is frequently related to the released fuel flow. C. n. v.. n. 40. L. T. 42. pp. n.5057 (mH) inductance 0. “Development of a HighSpeed Solenoid Valve: Investigation of the Energizing Circuits”. 11071116. MAG27. 95 n. 428435. “Development of a Simulation Method for Dynamic Characteristics of Fuel Injector”. R. 7. it can be concluded that: 1) the adopted linear MKsB model was relatively suitable to analyze the physical problem. Another consequence is that the EFI operation time is relatively reduced (please. some extra quantity of fuel is released promoting the loss of accuracy of the EFI and consequently a loss of linearity. 2. Koizumi. “The Bendix DEKA Fuel Injector SeriesDesign and Performance” SAE Technical Papers Series. 2. pp. IEEE Transactions on Magnetics. 1993.. 15. Table 1.. “Dynamic Model of Solenoids Under Impact Excitation. Verification of the influence of the armature mass on linearity (when = 1 ms and T = 9 ms).. v. v. T. Y. Including Motion and Eddy Currents”. Kajima. pp. New York. 1985. “Injectors Developments Examined”. 2001.. Proceedings of the 1992 International Conference on Industrial Electronics. pp. Kawase. DeGrace. v. SAE Technical Papers Series. 820202. 2) significant flow nonlinearity that can be directly correlated to closing of the armature alone can be significantly reduced if the choice of moving mass and return spring characteristics respects the following relationship: 0. v. n. 3. The following Table 1 shows the physical characteristics of the studied prototype. 01. 4. Kajima. Sonoda. v. Some characteristics of the EFI prototype. 5. G. R. Bata G. even though it is known that an EFI is complex and exhibits several nonlinearities. Figure 10. 18. Control. the EFI dynamic range is increased. 564569... Lesquesne. 1987. T. References Ando.2 M Ks B2 0.. Because of the successive armature shocks against the valve seat. MAG26.L. 4.. n. 1991. 1985.88 V x 1 msec excitation pulse (voltage mode) 0.39393942. 37. 1995. 5. At high engine speed conditions. 1.IEEE Transactions on Magnetics. 1273. 1990. 6263. it is significantly shorter than the EFI operation time as it could be seen at Fig. The fuel released by the EFI up to the instant that the armature touches the valve seat is practically constant. OctoberDecember 2003 ABCM .. pp. 37153718. 850559 (SP609).. Karidis. pp.. pp.. Y. this difference could become critical and some transient problems. “Development of a HighSpeed Solenoid Valve: Investigation of the Energizing Circuits”. No. v. 334 / Vol. T.47 (Ns/m) damping coefficient (B) kerosene surrounding medium SAE 405 material (armature and yoke) Acknowledgment We would like to register our gratefulness to FAPESP for the financial support given to our IC project. “Dynamic Analysis of Automotive Solenoid Valve Using Finite Element Method”. M. 4347. T. v. as engine instability or poor transient response. IEEE Transactions on Industrial Electronics. M. Kawamura. Romann.. Kajima.. K. Ohdachi. pp. Turns..P.R. 7) and so. J. Y. 5763. n.23 Figure 9.T. Instrumentation and Automation ”. SAE Technical Papers Series. This is particularly interesting for those who want to build up an injection strategy... B.
21.. Toyoda. and Magnetic Field Problem”..Ed. n. Journal of the Brazilian Society of Mechanical Sciences. “Position Estimation in Solenoid Actuators”. “A New Model for Fast Acting Electromagnetic Fuel Injector Analisys and Design”. 1995 Industry Applications Conference. MAG26. Y. 1985. “Análise e Projeto de Válvulas Eletromagnéticas Injetoras de Combustível: Uma Nova Proposta”.25. 3.A. Y. of Mech. J. XXV.W.. Longstroke Solenoids With Two Springs”. Ogata.. pp.M. Soc. IEEE Transactions on Magnetics. Nehl.Development of a HighSpeed Solenoid Valve: an Investigation of the … Lesquesne. “Fastacting.A. T. pp. v. 1982. MAG24. Chen. Magnetic..97108. 618623. Passarini. T. Ando. Y. Lim... “Computational and Experimental Investigation on Solenoid Valve Dynamics”.. 5. “Optimum Design of Dynamic Response in Automotive Solenoid Valve”. 820902.. São CarlosSP. Cheung. N. pp. 125130. v. Ohdachi. 2a. Mechanical. H.. Editora Prentice/Hall do Brasil. M. K.W. Rio de Janeiro. Passarini. Takeuchi. 1986 “Aisan Fuel Injector for Multipoint Injection System”. B. pp. Vol. 11891197. n. 860486. No... M. 1. “New Algorithm for Coupled Solutions Of Electric.. and Mechanical System in Dynamic Simulation of Solenoid Actuators”. Matsubara. S. n. 26. Conference Record of the 1995 IEEE Thirtieth IAS Annual Meeting (IAS ’95). Aoki. IEEE Transactions on Magnetics. Brasil. New York. pp. V. pp. International Journal for Numerical Methods in Engineering. 2001..F. Kawase. 1993. 1.. Szente. 1988. 1990b.. Rahman. 27. IEEE Transactions on Magnetics. 2003. pp. Murakami. pp. Copyright 2003 by ABCM OctoberDecember 2003. n. J. Inoue. Yuan. 52265228. 270273.. T.. 4 / 335 . pp. v. v.. L.. v.. 1993. A. v. Takada. v. Y. v. & Eng. S. K. T. K.C. n. “Single Point Electronic Injection System” SAE Technical Papers Series. MacBain.C. pp... New York. 3. 848856.. 8387. SAE Technical Papers Series. 6. Doctoral thesis. 1318. 1. L. 1995. “Solenoid Simulation With Mechanical Motion”. M. EESC. Brasil. 1991.. Pinotti Jr. Pawlack.. n. Inaguma. 1990. of the Braz. K.C. Proceedings 2001 IEEE/ASME International Conference on Advanced Intelligent Mechatronics. IEEE Transactions on Industry Applications. Universidade de São Paulo.. J. n. Sci. “Transient Finite Modelling of Solenoid Actuators: the Coupled Power Electronics. Vad. “Engenharia de Controle Moderno”. 476483.