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

net/publication/308849328

A new model of electromechanical relays for predicting the motion and

electromagnetic dynamics

Conference Paper · October 2015

DOI: 10.1109/IAS.2015.7356818

CITATIONS READS
5 836

3 authors:

Edgar Ramirez-Laboreo C. Sagues

University of Zaragoza University of Zaragoza
22 PUBLICATIONS   108 CITATIONS    218 PUBLICATIONS   2,752 CITATIONS   

SEE PROFILE SEE PROFILE

Sergio Llorente
BSH Bosch und Siemens Hausgeräte GmbH
62 PUBLICATIONS   864 CITATIONS   

SEE PROFILE

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

Thermal Analysis through Lumped Element Models View project

Modeling & Control of Switch-Type Reluctance Actuators View project

All content following this page was uploaded by Edgar Ramirez-Laboreo on 05 October 2016.

The user has requested enhancement of the downloaded file.

 A new model of electromechanical relays for
predicting the motion and electromagnetic dynamics

Edgar Ramirez-Laboreo Carlos Sagues Sergio Llorente

Departamento de Informática e Departamento de Informática e Research and Development Department,
Ingenierı́a de Sistemas (DIIS) and Ingenierı́a de Sistemas (DIIS) and Induction Technology,
Instituto de Investigación en Instituto de Investigación en Product Division Cookers,
Ingenierı́a de Aragón (I3A), Ingenierı́a de Aragón (I3A), BSH Home Appliances Group,
Universidad de Zaragoza Universidad de Zaragoza Zaragoza 50016, Spain
Zaragoza 50018, Spain Zaragoza 50018, Spain sergio.llorente@bshg.com
ramirlab@unizar.es csagues@unizar.es

Abstract—In this paper, a novel multiphysics and nonlinear the very first attempts to study the contact bounce phenomenon
model for electromechanical relays is presented. The electromag- was [1], where an energy-based analysis and some simple
netic dynamics is analyzed by calculating the total reluctance of spring-mass systems were proposed to explain the process
the magnetic equivalent circuit, which is composed of a fixed dynamics. A good review of electrical contacts including the
length iron core and an angular air gap. Magnetic saturation influence of the arc current was made years later [2], and one
and angular dependency of the reluctance are considered in the
calculation. Then, an energy balance applied to the electromag-
of the first works regarding the control of contact bounce was
netic components of the system is used to obtain the torque which presented in [3]. With respect to modeling, an early mechanical
acts on the movable armature. A planar mechanism of four rigid and electromagnetic coupled model including contact bounce
bodies, including spring-damping torques that restrict the motion was already presented in [4], but the authors themselves
and model the contact bounces that occur in the switchings, is suggested that some aspects should be studied deeply.
proposed to explain the dynamics of the movable components.
Experimental tests show the accuracy of the model both in the
Many researches have been presented in the recent years
electromagnetic and the mechanical parts. for predicting the electromagnetic dynamics and the motion
of relays and contactors. Different approaches have been
Index Terms—Electromechanical relay, Modeling, Magnetic considered for the electromagnetism, ranging from models
Equivalent Circuit, Planar mechanism, Contact bounce. with low computing requirements based on Magnetic Equiv-
alent Circuits (MECs) to comprehensive and time consuming
I. I NTRODUCTION Finite Element Method (FEM) analyses. Concerning the first
Electromechanical devices such as relays or contactors are approach, a detailed framework for building three-dimensional
widely used in domestic and industrial applications. Although electromagnetic MEC-based models can be found in [5]. An
their operating principle comes from the nineteenth century, electromagnetic MEC model including magnetic saturation is
these devices still offer many advantages over their semicon- proposed for AC contactors in [6], while the motion dynamics
ductor counterparts: they have lower energy losses, are able to and a position estimator are presented by the same authors
block current and to withstand voltage in both directions, do in a previous paper [7]. A MEC model is also used in [8]
not require drivers to be activated and are slightly cheaper. to optimize the geometry of a permanent-magnet contactor.
For these reasons, they are extensively used in products Some works combining the two methods can also be found.
needing some type of electric or electronic regulation. Relays See, e.g., the research in [9], where a MEC-based model is
and contactors can be found, e.g., in cars, home appliances, improved by using a correction factor calculated from FEM
communications systems or industrial machines. simulations, or [10], where an one-dimensional MEC model
and a two-dimensional FEM analysis are compared. On the
Despite these advantages, electromechanical devices also other hand, a pure FEM-based model is proposed in [11] for
present some drawbacks related to their operation mode. They calculating the attractive torque of a permanent-magnet relay.
are considerably slower than solid-state switches and, since
Different options have been considered also for the motion.
there are moving components, they are continuously subject
A mechanical model based on the Euler-Bernoulli theory for
to wear and their service life is shorter. Additionally, strong
beams is proposed to predict the motion of a relay in [12],
bounces are generated in each switching when the movable
where a viscoelastic spring-damper contact model is also
parts hit the fixed ones, originating additional wearing and a
presented for explaining the contact bounces. The beam theory
noise which is a problem in many applications. As a result,
is used in other researches [10], but the most widespread
electromechanical devices are often discarded in favor of solid-
approach is the use of spring-damper models with a single
state switches.
degree of freedom [4], [7], [8]. Some research has also been
In order to reduce the problems and increase the benefits, made regarding the control of electromechanical devices [7],
several research groups have been working on models and [8]. Nevertheless, the problem is still not solved because of the
control strategies for these electromechanical devices. One of highly nonlinear dynamics and the partially stochastic behavior
of these devices. In [13], a control algorithm which limits the III. E LECTROMAGNETIC MODEL
current through the coil of a contactor is described for reducing
contact bounce. Control strategies can be found also for other A. Current and magnetic flux dynamics
electromechanical devices, e.g., for an electromagnet excited The electromagnetic components of the relay are schemat-
pendulum [14] or for a solenoid valve [15]. ically represented in Fig. 2(a). The coil has N turns and
electrical resistance R. The magnetic circuit is composed of a
The research on electromechanical relays and contactors
soft iron core, with a fixed and a movable part, and an air gap.
has not been limited to predicting their motion or electro-
The geometry of the core is known and the size of the air gap
magnetic dynamics. See, e.g., the method proposed in [16] to
is determined by the angular position of the movable armature
evaluate the erosion of the electrical contacts of a contactor,
θ1 . When a voltage difference v is applied between the coil
or the FEM-based thermal model of sealed electromagnetic
terminals, the electrical behavior of the system is given by
relays developed in [17]. Besides, novel dynamic models are
also being developed for other types of electromechanical dφ
switches [18]. v = Ri + N , (1)
dt
In this paper, we present a novel model for electromechan- where i is the current flowing through the coil and φ is the
ical relays that (1) describes the electromagnetic dynamics magnetic flux through the core. Current i and magnetic flux
of the system taking into account both magnetic saturation φ are also related by means of the equation of the magnetic
and angular dependency, (2) studies the motion of a complex equivalent circuit,
mechanism with two degrees of freedom, (3) simulates the N i = φ<, (2)
contact bounces generated in the closure and the opening by
means of spring-damping torques and (4) requires a much where < is the total reluctance of the circuit and can be divided
lower computational cost than FEM-based models because it into the air gap reluctance <g and the magnetic core reluctance
is based on a magnetic equivalent circuit. These features make <c (Fig. 2(b)).
our model very appropriate for designing and validating control < = < g + <c . (3)
algorithms, e.g., for reducing contact bounce.
The first component of < is calculated from the geometry
The paper is structured as follows. First, in section II, the of the air gap,
electromechanical relay investigated in this work is presented lg rg θ1
and its operating mode explained. The electromagnetic dynam- <g = = , (4)
µ0 Ag µ0 Ag
ics of the system is described in section III by calculating
the reluctance of the magnetic equivalent circuit. Besides, an where lg and Ag are the length and the cross-sectional area
analysis based on the electromagnetic energy is made to obtain of the air gap, rg the average turning radius of the cross-
an expression for the torque that drives the motion of the relay. section, θ1 the rotation angle of the moving armature and µ0
Then, in section IV, a planar mechanism composed of four the vacuum magnetic permeability.
rigid bodies is proposed to describe the motion of the movable
On the other hand, the reluctance of the magnetic core
components. This mechanical model includes articular limits
is approximated by using an average path length lc and an
so that the motion of the mechanism is restricted to specific
average cross-sectional area Ac of the magnetic core,
positions. These bounds are achieved by means of spring-
damping torques that are also able to model the contact lc
bounces that usually exist in the operation of electromechanical <c = , (5)
µc Ac
devices. Experimental tests have been carried out to validate
the model and the results are presented and discussed in where µc is the magnetic permeability of the core material. The
section V. Finally, the conclusions of the research are presented relationship between magnetic flux density Bc and magnetic
in section VI. field intensity Hc within the magnetic core is considerably
nonlinear due to magnetic saturation. Consequently, µc cannot
be considered constant. To reflect this behavior, an alternative
II. S YSTEM DESCRIPTION version of the Froelich equation has been used,
The electromechanical switch used in this research is a c 1 Hc
general purpose PCB power relay which features a high elec- Bc = , (6)
1 + c2 |Hc |
trical insulation between the coil and the contacts (Fig. 1). This
is achieved by means of a mechanism with an intermediate where c1 and c2 are constants which have been experimentally
plastic part that completely separates the electromagnet and fitted. Although similar expressions have been already used in
the power terminals. Similar models are sold by the main previous works [6], [7] with good results, it must be noted
manufacturers in the market. Its operating mode is as follows. that the absolute value included in the denominator allows
When the coil is not energized, the movable contact stands still, the calculation also with negative values of Hc . Otherwise,
touching the normally closed contact. If the coil is energized, meaningless values of µc would be obtained and divisions by
the movable armature closes the magnetic circuit and pushes zero may appear in the equations.
the plastic part. At the same time, the opposite end of the According to (6), µc may be expressed either as a function
plastic component causes the movable contact to touch the of magnetic field Hc or of magnetic flux density Bc .
normally open contact. When the coil is de-energized, elastic
forces make the relay return to its original position, opening the Bc c1
µc = = = c1 − c2 |Bc | . (7)
magnetic circuit and closing the normally closed connection. Hc 1 + c2 |Hc |
 Now, if angular position θ1 is known, (11) can be solved by
numerical integration and then the current through the coil is
directly obtained by using (2). In addition, inductance L of the
coil can also be calculated by means of (10) and considering
that L = N 2 /<.

B. Magnetic torque calculation

In this section, an expression for the magnetic torque that
acts on the movable armature is obtained by applying an energy
balance to the electromagnetic components of the relay.
The electric energy supplied to the coil, We , is the only
energy input of the balance. This energy is transformed in
electromagnetic energy, Wem , mechanical work, Wm , and
energy losses, Wl . Hence, the balance can be expressed as
Fig. 1. Electromechanical relay.
We = Wem + Wm + Wl , (12)
or, in differential form, as
dWe = dWem + dWm + dWl . (13)

In the first place, the electric energy supplied to the system

in a differential time dt can be calculated as
dWe = vi · dt. (14)

By substituting (1) in the previous expression, it becomes

(a) (b) dφ
dWe = (Ri + N )i · dt = Ri2 · dt + N i · dφ, (15)
dt
Fig. 2. (a) Scheme of the electromagnetic components and (b) Magnetic
equivalent circuit. and, by using (2),
dWe = Ri2 · dt + φ< · dφ. (16)
Considering that φ = Bc Ac , the previous expression is
transformed to be in terms of magnetic flux.
On the second place, the total electromagnetic energy of a
c2
µc = c1 − |φ| . (8) system of volume V can be calculated as
Ac Z
1
Now, by substituting (8) in (5) and then in (3), the Wem = · B · H dV, (17)
2 V
expression for the total reluctance of the magnetic circuit is
obtained. where B and H are, respectively, the magnetic flux density and
rg θ1 lc the magnetic field intensity vectors, defined for every point in
<= + . (9)
µ0 Ag c1 Ac − c2 |φ| the space. If Bg and Bc are the average flux densities in the
air gap and the magnetic core, and Hg and Hc are the average
Note that reluctance < depends on angular position θ1 magnetic fields in the same components, the electromagnetic
and magnetic flux φ, hence including the variations caused energy stored in the relay can then be approximated by
by the motion of the mechanism and by magnetic saturation.
For simplicity in the notation, Eq. (9) will be expressed from 1 1
Wem = Wem,g +Wem,c = Bg Hg Ag lg + Bc Hc Ac lc , (18)
now on as 2 2
k2
< = k1 θ1 + , (10) where Wem,g and Wem,c are respectively the electromagnetic
1 − k3 |φ| energies in the air gap and the magnetic core. Given that
where the magnetic flux is constant along the magnetic circuit, i.e.,
rg lc c2 φ = Bg ·Ag = Bc ·Ac , Eq. (18) can be transformed into
k1 = , k2 = , k3 = .
µ0 Ag c1 Ac c1 Ac 1
Wem = φ (Hg lg + Hc lc ) , (19)
2
Substituting (10) in (2), and solving the system of (1)
and (2), the explicit equation for the dynamics of the magnetic and, by applying Ampere’s law, i.e., Hg lg + Hc lc = N i, an
flux is finally obtained. expression for the electromagnetic energy as a function of φ
  and i is obtained.
dφ v Rφ k2 1
= − 2 k1 θ1 + . (11) Wem = φN i. (20)
dt N N 1 − k3 |φ| 2
 4 θ4
The previous equation may now be expressed in terms of
the current through the coil or of the flux through the magnetic 2 G2
circuit by using (2). B H
C
G3
1 N 2 i2 1 θ1
Wem = = φ2 <. (21) G4
2 < 2 G1 θ3 E
Note that precisely one of these expressions is obtained if 1 3
y
inductance L of the coil is firstly calculated as L = N 2 /<
and then the expression for the energy stored in an inductor is x
A
used. D
1 1 N2 2
Wem = Li2 = i . (22)
2 2 <
Fig. 3. Mechanical model.
By using the form in terms of φ, the second term of the
balance is finally obtained.
1 2 exists in the bottom of the movable armature is also included
dWem = φ · d< + φ< · dφ. (23) in the mechanical model in joint A. To model the friction of
2 the whole mechanism, a viscous friction force is included in
In the third place, differential mechanical work dWm is joint H. Before analyzing the mechanism we introduce the
created by the electromagnetic torque τmag that acts on the notation for this section:
movable armature when it rotates a differential angle dθ1 .
xP , yP Coordinates of point P in the XY coor-
dWm = τmag · dθ1 . (24)   dinate system.
xP
~rP = Position vector of point P .
Finally, the fourth term of the balance is obtained by con- yP
sidering that the only losses in the electromagnetic components # »
PQ Vector from point P to point Q.
# » # »

are due to the Joule heating effect in the coil resistance. P Q = P Q Length of vector P Q.
dWl = Ri2 · dt. (25) g Gravity.
Gi Center of mass of rigid body i.
Thus, by substituting each term in the energy balance, mi Mass of rigid body i.
Ii Moment of inertia of rigid body i in the
Ri2 · dt + φ< · dφ = XY plane.
1 2 θ̇, θ̈ First, second time derivative of θ.
φ · d< + φ< · dφ + τmag · dθ1 + Ri2 · dt. (26)
2
Joints A, D and H are fixed and have known positions. If
The previous equation can now be solved to obtain the A is selected as the origin of the coordinate system, then
magnetic torque τmag applied to the movable armature.      
1 d< 0 xH xD
τmag = − φ2 . (27) ~rA = , ~rH = , ~rD = .
0 yH yD
2 dθ1

By deriving and substituting (10) in (27), the final expres- The position of all the other points of the mechanism
sion for τmag is obtained. depends on angular variables θ1 , θ3 and θ4 . Considering that
1 1 rg AG1 , BC, BG2 , DE, DG3 , and EG4 are known constant
τmag = − φ2 k1 = − φ2 . (28) lengths of the mechanism, the following vectors of the kine-
2 2 µ0 Ag
matic chain are calculated:
Note that the minus sign, together with the squared value of # » # »
   
−AG1 · sin(θ1 ) −yH · tan(θ1 )
φ, indicates that the torque acts always in the opposite direction AG1 = , AB = ,
AG1 · cos(θ1 ) yH
of θ1 .
# » # »
   
BG2 BC
IV. M ECHANICAL MODEL BG2 = , BC = ,
0 0
To describe the motion of the relay we propose the planar # »

DE · sin(θ3 )

# »

DG3 · sin(θ3 )

mechanism of Fig. 3. This mechanical model is composed DE = , DG3 = ,
DE · cos(θ3 ) DG3 · cos(θ3 )
of four rigid bodies connected by joints and numbered in
circles in the figure. Body 1 models the movable armature and # »
 
−EG4 · sin(θ4 )
body 2 corresponds to the plastic pusher. Since the movable EG4 = .
EG4 · cos(θ4 )
contact is made from a flexible copper sheet, it is modeled as
the union of two rigid bodies, 3 and 4, which are connected
by joint E. Spring torques that model the sheet rigidity are Then, by adding and subtracting these vectors, the position
included in joints D and E. An actual torsion spring that of centers of mass G1 , G2 , G3 and G4 and joints B, C and
E can be calculated as 1 4

FH
# » # » FB
~rG1 = AG1 , ~rB = AB, FcH H H G2
# » # » # » # » B
~rG2 = AB + BG2 , ~rC = AB + BC, m2g FC C
# » # »
~rG3 = ~rD + DG3 , ~rE = ~rD + DE,
# » # » (a) Body 2.
~rG4 = ~rD + DE + EG4 .
3
Note that, although the previous equations are in terms of 4

θ1 , θ3 and θ4 , the mechanism has only two degrees of freedom G3 FC

FEy C
because the variables are intrinsically related. m3g
FEx
 
xE (θ3 ) − xC (θ1 ) E G4
θ4 = atan . (29)
yC (θ1 ) − yE (θ3 ) kE
m4g
lD
FDy FEx
E
Free body diagrams of the four components of the mech- D FDx kE

anism are presented in Fig. 4. All the internal and external FEy
kD
forces and torques that act on the solids are represented
in these diagrams by means of arrows, each pointing to (b) Body 1. (c) Body 3. (d) Body 4.
the direction which has been defined as positive. Revolute
joints A, D and E cause the apparition of internal forces of Fig. 4. Free body diagrams.
unknown direction which can be decomposed in the X and Y
coordinates, obtaining FAx , FAy , FDx , FDy , FEx and FEy .
Pin-slot joints B and C create forces FB and FC , which are where klA and klD are the stiffness constants and clA and clD
respectively perpendicular to the direction of solids 1 and 3, the damping coefficients of the spring-damping torques.
and prismatic joint H creates a force FH in the Y coordinate
and a torque τH in the Z coordinate. The weights of the four bodies and magnetic torque τmag ,
which acts on body 1 and is given by the electromagnetic
Spring torques τkA , τkD and τkE in joints A, D and E are model, are also included in the mechanism. Equations for
calculated by means of Hooke’s law: linear and angular accelerations are then obtained by applying
Newton’s second law. Three equations are presented below for
τkA = kA · (θ1 − θ1,0 ) , (30) each body, corresponding to the net forces in coordinates X
τkD = kD · (θ3 − θ3,0 ) , (31) and Y and the net torque in coordinate Z. For bodies 1 to 4,
τkE = kE · (θE − θE,0 ) , θE = θ3 + θ4 , (32) torques have been respectively calculated in points A, B, D
and E.
where kA , kD and kE are stiffness constants and θ1,0 , θ3,0
and θE,0 are the springs natural angles. With regard to viscous Equations for body 1:
friction force FcH , it is calculated as
FAx − FB · cos(θ1 ) = m1 ẍG1 , (36)
FcH = cH · ẋG2 , (33) FAy − FB · sin(θ1 ) − m1 g = m1 ÿG1 , (37)
where cH is a constant damping coefficient. FB ·AB +m1 g·AG1 ·sin(θ1 )+τlA −τkA+τmag = I1 θ̈1 . (38)
The motion of the mechanism is restricted to specific
ranges in variables θ1 and θ3 due to the position of the fixed Equations for body 2:
armature and the contacts (Fig. 1). To model these limits, FB · cos(θ1 ) − FC · cos(θ4 ) − FcH = m2 ẍG2 , (39)
torques τlA and τlD act in joints A and D when bodies 1
and 3 are beyond their boundaries. Besides of limiting the FB · sin(θ1 ) + FH − m2 g − FC · sin(θ4 ) = m2 ÿG2 , (40)
motion, these torques also model the bounces generated when FH · BH − m2 g · BG2 − FC ·BC ·sin(θ4 ) + τH = 0. (41)
the moving parts hit the fixed ones. This is achieved by means
of torsional spring-damper systems which emulate the Kelvin- Equations for body 3:
Voigt viscoelastic model [12]. Considering [θ1,min , θ1,max ]
and [θ3,min , θ3,max ] the ranges of θ1 and θ3 , τlA and τlD are FDx + FEx = m3 ẍG3 , (42)
given by the following expressions: FDy + FEy − m3 g = m3 ÿG3 , (43)
 ˙ FEy · DE · sin(θ3 ) − FEx · DE · cos(θ3 )
−klA ·(θ1 −θ1,min ) − clA · θ1 , if θ1 < θ1,min

− m3 g·DG3 ·sin(θ3 ) + τkD + τkE − τlD = −I3 θ̈3 . (44)
τlA = −klA ·(θ1 −θ1,max ) − clA · θ˙1 , if θ1 > θ1,max (34)


0, otherwise, Equations for body 4:
 ˙
−klD ·(θ3 −θ3,min ) − clD · θ3 , if θ3 < θ3,min
 FC · cos(θ4 ) − FEx = m4 ẍG4 , (45)
τlD = −klD ·(θ3 −θ3,max ) − clD · θ˙3 , if θ3 > θ3,max (35) FC · sin(θ4 ) − FEy − m4 g = m4 ÿG4 , (46)
m4 g · EG4 · sin(θ4 ) − FC · EC − τkE = I4 θ̈4 .


0, otherwise, (47)
 By substituting expressions (29) to (35) as well as the
position of the centers of mass, the system of Eqs. (36) to (47),
which has 12 unknown variables (FAx , FAy , FB , FH , τH , FC ,
FDx , FDy , FEx , FEy , θ1 and θ3 ), permits the resolution of
the mechanism.

V. M ODEL VALIDATION
A. Implementation for simulation
The electromechanical model of the relay has been im-
plemented in MATLAB Simulink to carry out simulations
and prove its validity. The electromagnetic equations have
been written in MATLAB code and included in Simulink by Fig. 6. Circuit diagram of the experimental test bench. Resistor Rs permits
means of a MATLAB Function block. On the other hand, the measurement of the current through the coil.
the mechanical part has been implemented using blocks of
the SimMechanics library. This tool, apart from resolving the

CONTROL
equations for the motion of the model, provides a graphical 1
interface which allows a 3D visualization of the system during
the simulation. Fig. 5 shows some images of the developed 0
model in the SimMechanics graphical interface. 0 10 20 30 40 50
Time (ms)

Fig. 7. Control signal used in the two tests.

Fig. 8 shows a comparison between the experimental coil

current during the 24 V test and the current obtained from
simulation. The nonlinear dynamics of the system are evident
during the closure of the armature (first 25 milliseconds)
and also in the opening (last 25 milliseconds). The current
predicted by the model fits well with the experimental data.
For the same test, Figs. 9 and 10 show, respectively, binary
graphs for the normally closed and the normally open contacts.
Fig. 5. SimMechanics graphical interface. In these charts, a value ’1’ is represented for a contact if
it is closed and a ’0’ if it is open. During the closure, it
is seen that the movable contact does not leave its normal
B. Experimental tests position immediately after the activation of the control signal,
but it takes about 7.2 milliseconds. This behavior is due to
An electronic circuit (Fig. 6), similar to those used for the dynamics of the magnetic torque, which is lower than the
relay switching in industry applications, has been designed elastic forces until a given value of the current through the
and implemented to run experimental tests on the device and coil is reached (28). Approximately 2.5 milliseconds later, the
validate the model. Two BJT transistors and a resistor network movable contact hits the normally open contact, causing some
permit the relay to be activated and deactivated by a computer- bounces. Altogether, the relay takes about 10 milliseconds to
generated control signal. A general purpose diode is included make the complete switching, i.e., to open the normally closed
for coil suppression, i.e., to limit the elevated induced voltages contact and close the normally open contact. The behavior
that appear when switching off the relay. Supply voltage in predicted by the model in this sense fits almost exactly with
the test bench can also be adjusted from 0 up to 30 V. Voltage the measurements.
and current through the coil are measured and registered by
means of a digital oscilloscope during the tests, and binary The procedure is reversed during the opening, causing the
measurements of the normally closed and the normally open movable contact to return to its original position, opening the
contacts are also obtained throughout the experiments. normally open contact and closing the normally closed contact.
In this process, the time needed to begin the motion is about
Several tests have been performed on the actual relay. Data 9 milliseconds, the free motion lasts about 1 millisecond and
from a first set of experiments have been used to adjust some the bounces take approximately 2.5 milliseconds. It is shown
of the model parameters which were partially undetermined. again that the prediction given by the model agrees very well
To be coherent, the results presented in this section correspond with the measurements.
to a second set, different from the one used for the adjustment.
Experimental and simulated results of relay switchings at The results from the test with supply voltage of 30 V are
supply voltages of 24 and 30 V are shown in this paper. The presented in Figs. 11 to 13. Although the system dynamics are
control signal of both tests consists of a 25 milliseconds ’on’ similar to those of the 24 V test, there are some differences
time followed by a 25 milliseconds ’off’ time (Fig. 7). due to the nonlinearities, specially during the closure. At this
 Coil current (mA) 20 20

Coil current (mA)

Measured Measured
15 Simulated 15 Simulated
10 10

5 5

0 0

0 10 20 30 40 50 0 10 20 30 40 50
Time (ms) Time (ms)

Fig. 8. Measured and simulated coil current during the 24 V experiment. Fig. 11. Measured and simulated coil current during the 30 V experiment.

1 1
NCC

NCC
Measured Measured
0 0

1 1
NCC

NCC
Simulated Simulated
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Time (ms) Time (ms)

Fig. 9. Measured and simulated normally closed contact (NCC) during the Fig. 12. Measured and simulated normally closed contact (NCC) during the
24 V experiment. 30 V experiment.

1 1
NOC

NOC

Measured Measured
0 0

1 1
NOC

NOC

Simulated Simulated
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Time (ms) Time (ms)

Fig. 10. Measured and simulated normally open contact (NOC) during the Fig. 13. Measured and simulated normally open contact (NOC) during the
24 V experiment. 30 V experiment.

TABLE I. M OTION STAGES DURING THE CLOSURE .

stage, the coil through the current reaches the steady state in Motion start Free motion Bounces
Test
(ms) (ms) (ms)
about 10 milliseconds, faster than in the 24 V experiment. Measured 7.16 2.43 0.57
24 V
Nonetheless, its behavior during the opening is more similar Simulated 7.20 2.51 0.45
to the previous test. Similar conclusions are drawn from Measured 4.72 1.15 0.65
30 V
Simulated 4.91 1.08 0.60
analyzing the measurements of the contacts. It can be seen
that the closure is faster, mainly because the motion is started
earlier, although the period of bounces is slightly longer. The TABLE II. M OTION STAGES DURING THE OPENING .
opening is however a little slower because of the higher current
Motion start Free motion Bounces
in the closure, which generates a more elevated magnetic Test
(ms) (ms) (ms)
torque (28) that takes more time to be overtaken by the elastic 24 V
Measured 9.22 0.82 2.54
forces. In spite of these differences, the model provides a very Simulated 9.06 1.44 2.58
Measured 9.93 1.39 2.97
good estimation of the relay dynamics also for this experiment. 30 V
Simulated 10.39 1.33 2.12

Tables I and II summarize the times of the main stages of

the motion during the closure and the opening. The process VI. C ONCLUSIONS
has been divided into three times: the time needed to start
the motion, the time of free motion and the time of bounces. In this paper, a new model for predicting the electromag-
The sum of these three times results in the total switching time. netic and motion dynamics of electromechanical relays has
The tables present the results from both experimental tests and been developed and validated. The electromagnetic dynamics
also from the simulations. The agreement between model and has been studied by obtaining the total reluctance of the
reality is proved once again. magnetic circuit. Then, the magnetic torque that acts on the
mechanism, which depends on the current through the coil and [4] H. Nouri, N. Larsen, and T. Davies, “Contact bounce simulation using
the position of the mechanism, has been calculated by means matlab,” in Proc. 43rd IEEE Holm Conf. Electrical Contacts. IEEE,
of a balance of the electromagnetic energy stored in the system. 1997, pp. 284–288.
Besides of the current and the magnetic flux, the instantaneous [5] M. Amrhein and P. T. Krein, “3-d magnetic equivalent circuit framework
for modeling electromechanical devices,” IEEE Trans. Energy Convers.,
value of the coil inductance is also obtained. vol. 24, no. 2, pp. 397–405, 2009.
A planar mechanism with two degrees of freedom has [6] J.-R. Riba Ruiz and A. Garcia Espinosa, “A novel parametric model
been proposed to explain the motion of the relay, and a com- for ac contactors,” IEEE Trans. Magn., vol. 44, no. 9, pp. 2215–2218,
2008.
prehensive theoretical description of the kinematics and the
dynamics of the system has been presented. Spring-damping [7] A. G. Espinosa, J.-R. Riba Ruiz, and X. A. Morera, “A sensorless
method for controlling the closure of a contactor,” IEEE Trans. Magn.,
torques based on the Kelvin-Voigt viscoelastic model have vol. 43, no. 10, pp. 3896–3903, 2007.
been included in order to restrict the motion of the mechanism [8] H. Lin, X. Wang, S. Fang, P. Jin, and S. Ho, “Design, optimization,
and to model the bounces generated in the switchings. and intelligent control of permanent-magnet contactor,” IEEE Trans.
Ind. Electron., vol. 60, no. 11, pp. 5148–5159, 2013.
Experimental tests have been carried out on the actual
relay to validate the model. It has been shown that it fits [9] Z. Guofo, W. Qiya, and R. Wanbin, “An output space-mapping algo-
rithm to optimize the dimensional parameter of electromagnetic relay,”
correctly to the nonlinear behavior of the system, providing IEEE Trans. Magn., vol. 47, no. 9, pp. 2194–2199, 2011.
good estimations for the current through the coil and the [10] D. Wattiaux and O. Verlinden, “Modelling of the dynamic behaviour
motion of the mechanism. The agreement between simulated of electromechanical relays for the analysis of sensitivity to shocks and
and measured data has been proved to be very high even vibrations,” Exp. Mech., vol. 51, no. 9, pp. 1459–1472, 2011.
at different supply voltages and both in the closure and the [11] Z. Guofu, W. Qiya, Y. Wenying, and L. Huimin, “Permanent-magnet
opening of the relay. equivalent model of calculating relay’s static attractive torque charac-
teristics by finite element method,” IEEE Trans. Magn., vol. 48, no. 9,
The developed model, since it is based on a magnetic pp. 2467–2471, 2012.
equivalent circuit, is much less time consuming than FEM [12] X. Jun, H. Jun-jia, and Z. Chun-yan, “A dynamic model of electro-
based models. This feature makes it very appropriate for the magnetic relay including contact bounce,” in 2008 Int. Conf. Electrical
Machines and Systems. IEEE, 2008, pp. 4144–4149.
optimization of new control algorithms. In addition, since it
[13] P. M. dos Santos Dias de Moraes and A. J. Perin, “An electronic control
is fully parametrized, temperature dependance or stochastic unit for reducing contact bounce in electromagnetic contactors,” IEEE
behaviors of the parameters could be easily added to the model Trans. Ind. Electron., vol. 55, no. 2, pp. 861–870, 2008.
to improve its performance or to carry out sensitivity analyses. [14] K. Austin and J. Wagner, “Development of an electromagnet excited
mass-pendulum system modeling and control laboratory experiment-
ACKNOWLEDGMENT theory and test,” in 2013 European Control Conf. IEEE, 2013, pp.
256–261.
This work was partially supported by projects [15] S. K. Chung, C. R. Koch, and A. F. Lynch, “Flatness-based feedback
IPT-2011-1158-920000 of subprogram INNPACTO and control of an automotive solenoid valve,” IEEE Trans. Control Syst.
RTC-2014-1847-6 of subprogram RETOS-COLABORACIÓN Technol., vol. 15, no. 2, pp. 394–401, 2007.
from Ministerio de Economı́a y Competitividad / European [16] Y. Kawakami, M. Takashima, M. Hasegawa, Y. Watanabe, and K. Sawa,
Union. “An evaluation method of the contact erosion based on the dc mode test
of electromagnetic contactor,” in Proc. 50th IEEE Holm Conf. Electrical
Contacts and 22nd Int. Conf. Electrical Contacts. IEEE, 2004, pp. 83–
R EFERENCES 89.
[1] P. Barkan, “A study of the contact bounce phenomenon,” IEEE Trans. [17] H. Liang, W. Wang, and G. Zhai, “Thermal analysis of sealed electro-
Power App. Syst., no. 2, pp. 231–240, 1967. magnetic relays using 3-d finite element method,” in Proc. 53rd IEEE
Holm Conf. Electrical contacts. IEEE, 2007, pp. 262–268.
[2] J. W. McBride, “Electrical contact bounce in medium-duty contacts,”
IEEE Trans. Compon. Hybrids, Manuf. Technol., vol. 12, no. 1, pp. [18] J. S. Pulskamp, R. M. Proie, and R. G. Polcawich, “Nano-and micro-
82–90, 1989. electromechanical switch dynamics,” J. Micromech. Microeng., vol. 24,
[3] T. S. Davies, H. Nouri, and F. W. Britton, “Towards the control of no. 1, p. 015011, 2013.
contact bounce,” IEEE Trans. Compon. Packag. Manuf. Technol. A,
vol. 19, no. 3, pp. 353–359, 1996.

View publication stats