Natural Heat-Sinking Control Method for
High-Speed Actuation of the SMA
Chee Siong Loh; Hiroshi Yokoi & Tamio Arai
Dept. of Precision Engineering, School of Engineering, University of Tokyo
7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-8656 Japan
saratuga@gmail.com
Abstract: This paper describes two methodologies for increasing the actuation speed of the shape memory alloy
(SMA) actuator in ambient environment. The first method involves the implementation of a simple, light-weight
heat sink, which consists only of a combination of an outer metal tube with the silicone grease, but able to cool the
heated alloy effectively. The second method describes a high current pulse actuation that actuates the alloy fastly
using pulses in the milliseconds order. We hypothesize that a fast actuation of the SMA results in small increase
in temperature, due to energy transformation from heat energy to the kinetic energy in the SMA. This new
heating method revolutionizes the actuation of the alloy for a significantly faster response.
Keywords: Shape memory alloy (SMA), simple light- weight heat sink, high current pulse actuation
1. Introduction
Shape memory alloys (SMAs) are materials that exhibit
the characteristics of both the shape memory effect and
the super-elasticity [Auricchio, F., 1995]. The shape
memory effect involves a temperature change during
phase change. Referring to Fig. 1, when SMA is at low
temperature, it mostly remains in the form of crystalline
structure of martensite (b), which displays an elastic
nature and easily deforms (c). When heated, the
crystalline structure transforms to the austenite structure
(a), which is less elastic thus strain induced to the SMA at
the lower temperature martensite phase can be recovered
in the austenite phase. The super-elasticity is an effect
where phase change from martensite to austenite occurs
without the change in temperature but rather in the
increase of load and vice versa ((a)-(c)). As unloading
takes place, martensite phase is restored, which can be
observed in its plasticity behavior [Bhattacharyya et. al,
2001].
temperature
Austenite
Deformed
Martensite
Twinned Martensite
load
The shape memory effect is the main concern of this
research. The SMA produces high recovery forces of
approximately 50[kgfmm-2] which gives a high output
force to weight ratio when using the alloy. The SMA wire
does not consume much space or weight when used in
the development of SMA based actuators.
However, the SMA also exhibits several disadvantages,
such as large hysteresis and slow in actuation speed, of
approximately 0.2[Hz] [Dono, S., Saito, A. & Kuwata, T.,
2003]. The slowness of the actuation speed, which is
known as the major hindrance to its wide application
prospects, is due to the slowness of heat dissipation from
the alloy to the environment. On top of that, when the
alloy is applied in an actuator system, inappropriate
heating for actuating the SMA also further contributes to
the slowness in the heat dissipation process. These factors
will then cause the alloy to remain in its less elastic
austenite structure, eventually leading to the difficulty in
extension and contraction of the alloy.
In recent years, there have been researches into increasing
the actuation speed of the SMA from the heat-sinking
aspect and actuation aspect of the alloy. [Russell, R.A. &
Gorbet, R.B., 1994] proposed the usage of a mobile
rotating heat sink, consists of a metal bar attached to a
shaft which rotates in the direction of the contraction of
the SMA wire on a horizontal plane, which then cools off
the heated SMA wire upon coming in contact with it. The
heat sink was reported successful in tracking a 0.16[Hz]
rectangular wave, but is bulky and heavy in its structure.
[Furuya, Y., & Shimada, H., 1990] described a crab-like
robot actuated using SMA springs in water immersion for
cooling purposes. The crab moves at a speed of 7.5[cms-1]
but the power used increases by 10 fold, compared to
Fig. 1. Thermo-mechanical property of the SMA
International Journal of Advanced Robotic Systems, Vol. 3, No. 4 (2006)
ISSN 1729-8806, pp. 303-312
303
International Journal of Advanced Robotic Systems, Vol. 3, No. 4 (2006)
when operated in ambient environment which consumes
a power of 1[kW].
[Kuribayashi, K., 1991] replaces conventional current
limiting method with direct measurement of the
temperature of SMA using Cu-Constantan thermocouple.
Forced air convection is incorporated using an electrical
fan, achieving a response of 0.4[Hz ]. The electrical fan
makes the actuator bulky and not widely applicable.
[Featherstone, R., & Teh, Y.H., 2004] proposed the usage
of resistance as a form of temperature measurement and a
maximum safe heating current was designed to prevent
overheating. This method was incorporated into Grant’s
two-stage relay controller which showed an improvement
of double in its actuation speed. However, the change in
resistance of the SMA is small, approximately 10[%]
which therefore makes measurement difficult and
inaccurate.
[Seldon, B., Cho, K.J. & Asada, H., 2004] showed how the
SMA can be heated segmentally using Peltier devices. In
their later work [Seldon, B., Cho, K.J. & Asada, H., 2005],
using the same experimental setup, they proposed
Hysteresis Loop Control (HCL) method, specifying
intermediate temperatures to be ‘pulled back’ to after
extreme temperatures of cooling and heating have been
reached. This methodology is claimed to have shortened
latency times significantly and saved power. The
actuation speed achieved by four segment coordination
was 0.1[Hz], and Peltier devices are large and bulky,
resulting in very low engineering prospects.
There are other control methods proposed in
conventional research efforts [Eren, Y. et.al, 2002] and
[Elahinia, M.H. & Ashrafiuon, H., 2002] which involve
heating of the alloy in a gradual and slow manner which
is inefficient and results in heating up the alloy
excessively. By increasing actuation speed of the alloy,
and maintaining the actuator’s light weight and
simplicity in nature (Fig. 2), a much wider application can
be realized and simultaneously many low force actuation
problems can be overcome, which are the objective and
motivation behind this research.
In later parts of this paper, we also describe initial efforts
in developing a SMA-based actuator for the application
as artificial muscles in prosthetics, implementing the
methodologies in (i) and (ii) above.
2.1. Improvement of Heat-Sinking
Natural cooling for the SMA is known to be slow and
ineffective, therefore heat-sinking is of utmost importance.
And however, due to the fact that complex heat-sinking
mechanisms reduce the engineering prospects of the SMA
actuator as a whole, we propose the application of a
simple, light-weight heat sink to facilitate heat-sinking in
ambient environment, as most engineering demands are
for ambient environment actuated systems.
2.1.1 Proposed Heat Sink
Our proposed heat sink consists of a combination of an
outer metal tube together with the silicone grease shown
in Fig. 3. The silicone grease is smeared onto the surface
of the SMA of diameter φ0.3[mm] making sure that no
area is left uncovered. In this condition, the silicone
grease coated SMA is then inserted into the outer metal
tube of diameterφ0.8[mm].
(a)
simple, light
design
bulky, heavy
design
i.) development of a heat-sinking mechanism for the SMA
in ambient temperature for maintaining simplicity for
wide application prospects of the alloy.
ii.) development of a high current pulse actuation method
for the SMA actuator for optimum heating of the alloy.
(b)
Research
target
Outer metal
tube (φ0.8)
[Featherstone, 2004]
-ER position control[Gorbet, 1995]
-mobile heatsink[Selden, 2004]
-Peltier devices-
[Kuribayashi, 1998]
-Electrical fan[Y.Furuya, 1985]
-Water cooling-
SMA (φ0.3)
HEAT-SINKING
Silicone grease
ACTUATION
Low
response
High
response
Fig. 2. Research targets simplicity in design and high
response of the SMA actuator
304
2. Methodology
This research deals with the slow actuation speed
problem from the following perspectives:
(C)
Fig. 3. (a) Silicone grease (b) outer metal tube and (c) cross
sectional diagram of the proposed heat sink with the
SMA wire
Chee Siong Loh; Hiroshi Yokoi & Tamio Arai / Natural Heat-Sinking Control Method for High Speed Actuation of the Shape Memory Alloy Actuator
The concept of our proposed heat sink is based on the
Fourier’s heat transfer equation [Masahiro, S., 1995] given
below:
⎛ ∂T ⎞
q = −λ ⎜
⎟
⎝ ∂n ⎠
⎛ ∂T ⎞
Q = −λAt ⎜
⎟
⎝ ∂n ⎠
(1)
(2)
ρcΔx
where q, T, and λ, are the rate of heat transferred per unit
area [Wm-2], temperature [K] and thermal conductivity
[Wm-1·K-1] respectively. For simplicity purposes, we
consider only the horizontal distance measured from the
SMA for a one dimensional analysis. The total rate of heat
transferred for the whole object can be obtained by
considering the total surface area of the object involved
which is given in (2). Next, we consider an infinitesimal
heat transfer in one dimension as shown in Fig. 4. The
difference between heat flowing into and out from two
vertical planes of distance Δx can be expressed in the
form of (3).
q
x−
q
Δx
2
x+
Δx
2
Δx
Fig. 4. One dimension infinitesimal heat transfer
q
x−
Δx
2
−q
x+
Since heat is a conserved quantity, the difference between
heat in and heat out from an infinitesimal block of heat
flow can be attributed to the rise in temperature of the
medium encapsulated by Δx. By considering the density
ρ and heat capacity c of the medium, its increase in
temperature is shown in the following equation.
2
(6)
when Δt￫0, we obtain the following differential equations
for one dimensional heat transfer.
ρc
∂T
∂ ⎛ ∂T ⎞
= ⎜ λx
⎟
∂t ∂x ⎝ ∂x ⎠
∂T
∂ 2T
ρc = λ 2
∂t
∂x
(7)
Considering the simple fact that thermal conductivity λSI
(for silicone) > λAIR (for air) and surface area ASMA (for
SMA) < ATUBE (for outer metal tube) (Fig. 5), we can easily
predict from (7) that the heat from the SMA spreads to a
wider surface area of the outer metal tube through the
high thermal conducting silicone grease, providing a
faster dissipation of heat from the SMA wire to the
environment. The thermal conductivities for the SMA,
silicone grease, outer metal tube and air are provided in
Table 1. SMA without heat sink has only air as its
medium of contact, which is slow in transferring of heat
from the heated SMA to the environment.
Δx
2
⎛ ∂T ⎞
⎛ ∂T ⎞
= −⎜ λ x
+ ⎜ λx
⎟
⎟
⎝ ∂x ⎠ x − Δx ⎝ ∂x ⎠ x + Δx
⎞
ΔT ⎛
= ⎜⎜ q Δx − q Δx ⎟⎟
x+
Δt ⎝ x − 2
2 ⎠
R
(3)
Tube
Q r’
Q
Q
r
2
SMA
However, expanding the right terms using Taylor’s
expansion for one dimension gives us the equation in (4).
(a) SMA with heat sink
⎛ ∂T ⎞
− ⎜ λx
⎟
⎝ ∂x ⎠ x ± Δx
2
⎧⎛ ∂T ⎞
⎫
∂ ⎛ ∂T ⎞ Δx
= −⎨⎜ λ x
+ 0(Δx 2 )⎬
⎟ ± ⎜ λx
⎟
⎩⎝ ∂x ⎠ x ∂x ⎝ ∂x ⎠ x 2
⎭
(4)
SMA
Thermal
conductivity
Applying (4) into (3), the equation can rewritten in the
following form.
q
Δx
x−
2
(b) SMA without heat sink
Fig. 5. Cross-sectional diagrams and each case’s way of
heat dissipation
−q
Δx
z+
2
=
∂ ⎛ ∂T ⎞
⎜ λx
⎟ Δx
∂x ⎝ ∂x ⎠ x
(5)
20.9
Silicone
2.0
Metal
315.0
Air
0.0386
Table 1. Thermal conductivities for related materials
2.1.2 Verification Experiments: FEM Simulation
To analyze the heat transfer property of the proposed
heat sink, we adopted the finite element method (FEM) to
solve the heat transfer equation (7) on a mesh consisting
305
International Journal of Advanced Robotic Systems, Vol. 3, No. 4 (2006)
of 100 Χ 100 nodes. A visual-based computer simulator
was constructed as seen in Fig. 6. Upper figures in (a) and
(b) represent the meshes for SMA without and SMA with
the heat sink respectively. The blue circles in (b) represent
the boundaries of the SMA wire, silicone grease and the
outer tube respectively.
By taking into account the thermal conductivities of the
SMA wire, silicone grease, air and outer metal tube, a
simulation of the rate of heat dissipation was carried out.
Initial temperature of SMA was set at 26[°C] and was
then increased to 66[°C]. In real world, SMA achieving
this temperature and above is assumed to have
undergone the austenite transformation, recovering a
large part of its strain. The model was then let to ‘cool
down’. Using this simulation, the rate of dissipation of
heat for the SMA with and without heat sink could be
observed and compared (Fig. 6).
In the case of the SMA without heat sink (Fig. 6(a)), we
can observe the heat builds up around the wire and
prevents further diffusion of heat out from the alloy. Due
to the low thermal conductivity of air, the relatively fast
heat diffusion from the SMA is slowed down by
surrounding air, resulting in heat accumulation effect
near the wire which can be observed in the diagram Fig.
6(b) below.
Conversely, when the heat sink is used, heat from the
surface of the SMA is spread to a larger surface area of
the outer metal tube via the high thermal conducting
silicone grease. As a result, heat does not accumulate near
the wire but transferred externally to a larger surface area
to be disposed to the environment, validating the
effectiveness of the proposed heat sink.
after an interval
(a)
without
heat sink
after an interval
(b)
with
heat sink
Fig. 6. Heat transfer model based on FEM for SMA
without heat sink and SMA with heat sink.
Simulation results show that within 4[s], outer metal tube
with silicone grease brought about a 7[°C] reduction in
temperature compared to natural cooling (Fig. 7). As soon
as heating was ‘switched off’ the cooling curve for when
both silicone grease and outer metal tube were in used,
shows a steep slope, and that cooling was taking place at
a rate of 5[°Cs-1], faster than that without heat sink. The
reduction in temperature of 7[°C] (when both the outer
metal tube with the silicone grease were used) after only
306
4[s] suggests that large amount of heat from the heated
SMA is being channeled out to the environment through
the larger surface area of the actuator.
Fig. 7. Simulation results of the rate of dissipation of heat
2.1.3 Verification Experiments : Real World Actuation
To further investigate the effectiveness of the proposed
heat sink, real experiments were conducted using the
SMA (manufactured by Nilaco) as an actuator to haul up
a weight of 3[kg] periodically (Fig. 8) for both conditions
with and without usage of the proposed heat sink. A
faster descend of the weight during cooling is expected to
be seen when heat sink is in used.
A SMA wire of 0.3[mm] in diameter and of length 60[cm]
was used for the experiments. A position sensor (PSD
sensor, GP2D12) was placed below the weight for
detection of the contraction of the SMA upon actuation.
On the alloy was attached a thermocouple (TCKT0022)
for temperature monitoring (to avoid overheating). Due
to difficulty in detecting temperature of SMA accurately
in the outer tube, temperature comparison could not be
shown in this section. The alloy was passed through a
current transducer (LTS25-NP) for contact-less currrent
detection. The sensory feedback to the PC was done using
a 12 bit CONTEC PC card and the heating of the SMA
was controlled using the H8/3664 micro-controller. The
sensors and control circuits are connected as shown in Fig.
9. A diode (V39A64) was inserted in the top portion of
the circuit for regulating one way flow of current.
Experiments were conducted in room temperature of
27[°C]-28[°C] with minimum wind condition. PWM
(Pulse Width Modulation) signals were induced to the
SMA to haul up the 3[kg] weight periodically.
Comparative experiments were done , where rectangular
wave of 0.2[Hz], 2.0[A] of duty ratio 0.4 was applied to
the SMA in both cases, with and without heat sink.
As we compare the results obtained from the experiments
shown in Fig. 10, in the case of the SMA without heat
sink, the alloy reached its maximum strain after
approximately 1[s] from start. Natural cooling, without
any heat sink seems to be ineffective and slow in
channeling heat out from the alloy.
Chee Siong Loh; Hiroshi Yokoi & Tamio Arai / Natural Heat-Sinking Control Method for High Speed Actuation of the Shape Memory Alloy Actuator
did not rise as fast as compared to its counterpart, and
when voltage source was cut off, it extended at a faster
rate than the SMA without heat sink, with only air as its
cooling medium. This can be thought of as, a large
amount of heat was being dissipated even during the
heating process was taking place.
Fig. 8. SMA wire used to haul up a 3[kg] weight
Fig. 9. Schematic diagram of the circuitry design for the
entire system settings
This phenomenon could be observed in Fig. 10(a) (a
graph which displays the position of the weight against
time) which shows that SMA wire without heat sink had
no sufficient time in recovering a visible portion of its
strain before the subsequent heating. In the case of SMA
without heat sink, even after its current supply was cut,
the weight remained at its maximum height and started
descending only after 2[s] the current was switched off.
As a result the weight remained at the position where the
alloy achieved its maximum strain, after about 5[s] from
start. As time progressed, the oscillation converged to the
position of its maximum strain. Conversely, observing the
results in Fig. 10(b), when the proposed heat sink was
incorporated, the SMA wire cooled off at the point when
current was turned off. The wire recovered a visible part
of its strain, which could be observed as the response of
continuous peaks and valleys. The oscillation width
maintained a value of 7.5[mm] even after 25[s] of
actuation, compared to only 3[mm] of width achieved by
natural cooling. With the usage of heat sink, the weight
Fig. 10. Comparative experiments for cases with and
without heat sink when 0.2[Hz], 2.0[A] of duty ratio 0.4
rectangular signal was applied to the SMA
The concept of the proposed heat sink is simple and yet
effective in cooling off the heated SMA as mentioned in
previous section. The silicone grease smeared onto the
SMA wire insulates the wire from the outer tube and
simultaneously acts as a high thermal conducting
material for the heat from the SMA to the outer metal
tube. The thermal conductivity of both silicone and air are
λSI=2.0 and λAIR=0.0386 which shows that the transfer of
heat from the SMA is done at a rate equivalent to the ratio
of
the
thermal
conductivities
of
both
the
materials,
λ SI
= 51.8 . It is therefore made sure that the
λ AIR
intermediate silicone grease is in contact with the tube
and the SMA wire. The outer metal tube then, provides a
larger surface area for the dissipation of heat from the
SMA wire to the environment (Fig. 11). Let us denote the
surface area of SMA and outer metal tube as ASMA and
307
International Journal of Advanced Robotic Systems, Vol. 3, No. 4 (2006)
ATUBE respectively, the diameter of SMA and outer metal
tube as ΦSMA and ΦTUBE respectively, the dissipation of
heat when with and without the heat sink as QWITH_HS and
QWITHOUT_HS respectively. From (2), we can rewrite the
relationship of both heat dissipations from the SMA to
the environment as the following:
QWITH _ HS
QWITHOUT _ HS
=
ATUBE
ASMA
QWITHOUT _ HS
=
1. Heat is reduced when SMA is actuated
2. Rate of heat reduction increases as the actuation speed
of SMA increases.
(8)
By substituting the surface area A=πΦL, we can express
the heat transfer in terms of the diameters as follows:
QWITH _ HS
SMA, the existence of two forms of energies makes it
possible for the interchanging of these energies. Hence,
we propose two hypotheses for describing this
characteristic of the SMA, which are as follow.
Φ TUBE
Φ SMA
(9)
which states that the ratio of the amount of heat being
dissipated in both conditions can be expressed as the ratio
of their effective diameters respectively. Therefore, we
can forecast the amount of heat dissipated for different
diameters of SMAs and outer metal tubes using (9) as
shown above.
To verify these hypotheses, a series of experiments were
carried out to compare the temperature increase and rate
of temperature increase for the strained SMA and nonstrained SMA. The experiment settings for both
conditions are shown in Fig. 8, except that for nonstrained SMA, the weight 3[kg] was replaced with a
weight of 1[g], light enough as not to strain the alloy.
Experimental results are presented as temperature curves
against time in Fig. 12.
L
Φ
Fig. 11. Outer metal tube
(a) Strained SMA
2.2 High Current Pulse Actuation Method
As mentioned earlier, conventional actuation methods
have a high tendency to heat the SMA wire gradually,
hence overheating the SMA wire causing difficulty in
cooling in between actuations. We believe that in order to
further improve the actuation speed of the SMA, a
different approach in heating or actuation method is
necessary. Instead of indirect heating using heat
conducting materials (eg. Peltier devices) or gradual
heating, high current in the form of short pulses can be
used as an effective replacement.
(b) Non-strained SMA
2.2.1 Hypothesis of Heat Energy Absorption
The concept of the high current pulse actuation method
arrives from the hypothesis of the heat energy absorption
during SMA actuation, based on the following equation
(10):
E Electrical = E Kinetic + E Heat
(10)
Compared to ordinary metals which do not exhibit shape
recovery characteristics, no kinetic energy is involved in
the energy transformation from electrical energy to heat
energy. Therefore, we believe that when dealing with the
308
Power:26.25 [W]
Duration: 2 [s]
Power:12.5 [W]
Duration:5 [s]
Power:7.5 [W]
Duration: 5 [s]
Fig. 12. Comparing temperature increase and rate of
temperature increase when heating strained and non
strained SMA
Firstly, we notice that, the strained SMA when used in
actuation shows a lower increase in temperature and rate
of increase in temperature compared to when SMA not in
Chee Siong Loh; Hiroshi Yokoi & Tamio Arai / Natural Heat-Sinking Control Method for High Speed Actuation of the Shape Memory Alloy Actuator
actuation, when both settings were actuated using three
different power inputs of 7.5[W] for a duration 5[s],
12.5[W] for a duration of 5[s] and 26.25[W] for a duration
of 2[s].
These results suggest to us that when SMA is actuated, a
part of the heat energy is absorbed for the transformation
into kinetic energy in the SMA, which is consistent with
the first hypothesis above.
Next, we compared the rate of temperature increase for
the strained SMA when three input powers were induced
into the alloy. When a high power of 26.25[W] was
induced, the rate of temperature increase which should
be double that of when inducing 12.5[W], showed a lesser
value. However, an almost double in ratio could be
observed for the gradient of the temperature curves when
12.5[W] and 7.5[W] were induced. As for controlexperiment, all three gradients of the temperature curves
obtained from the non-strained SMA when induced the
three power inputs show consistency with the ratio of the
induced power inputs. These results suggest to us that
the higher the power is induced (the faster the SMA is
actuated), the faster heat reduction takes place, hence a
much slower rate of temperature increase when the SMA
is actuated.
Using this concept, we propose the pulse actuation
method, which applies high current (for fast actuation of
the SMA) using short ON time pulses (for fast cooling of
the SMA as input is switched off quickly).
2.2.2 Verification of Proposed Method
To investigate the effectiveness of the proposed high
current pulse heating have over gradual actuation
methods on the response of the alloy, we conducted
experiments using settings shown in Fig. 13 for a more
accurate data acquisition purpose. The SMA wire of
length 85[cm] was used as an actuator on a horizontal
plane, pulling a bias spring in a periodic manner.
The circuitry system in Fig. 9 was also integrated into this
actuator. The position sensor was used to detect the
contraction of the SMA towards the sensor therefore the
displacement graphs are seen as decreasing distance
against time. The thermocouple and current transducer
are both for the detection of the temperature and the
current flowing in the alloy respectively. A comparative
experiment of actuation cycle between conventional
gradual actuation method and our proposed high current
actuation method was carried out. For gradual actuation,
a low current of 0.75[A] of ON time 3[s] in an interval of
3[s] was induced in a periodic manner for a contraction of
12.5[mm]. On the other hand, in the proposed method, a
current of 3[A] of ON time 30[ms] in an interval of 3[s]
was induced in a periodic manner, achieving a same
contraction of 12.5[mm].
Comparing the results obtained for the comparative
experiments as shown in Fig. 14, for gradual heating,
each actuation induced an amount of 23.6[J] to the SMA
whereas the proposed method induced an amount of
mere 3.8[J] to the actuation system. As temperature
reached 55[°C], the gradual heating method only
recorded a mere 4 actuation cycles, whereas the pulse
heating method recorded 9 actuation cycles (Table 2).
We notice that actuating the SMA with high current, the
rise time can be decreased and if the input current is
switched off quickly, the SMA can be cooled quickly
before the next heating. We believe that fast actuation of
the SMA results in low rate of temperature increase in the
alloy.
Fig. 13. SMA wire used to pull a bias spring periodically
(a)
Fig. 14 (a) Temperature, displacement and current graphs
of SMA when heated gradually using 0.75[A], of duration
3[s] in every 3[s] periodically
309
International Journal of Advanced Robotic Systems, Vol. 3, No. 4 (2006)
The duty ratio at this initial stage was regulated in a step
function manner, but as for future work, an adaptive
control in regulating duty ratio can be constructed for
improvement of this method. For example, initial pulses
can be longer in duration for faster reaching of its stable
oscillating state, only then shorten for temperature and
oscillation stability maintenance.
Referring to the results obtained in Fig. 16 for 1.33 [Hz]
actuation, a significant improvement in stability of
actuation when compare the case of with and without
temperature control. Fig. 16(a) shows that the SMA
oscillates initially with a width of 1.72[mm] within a
temperature range of 45[°C]-54[°C]. The centre of
oscillation width increases in the direction of increasing
strain. As the temperature increases above 55[°C], a
decrease in strain can be observed, which finally
converges to a horizontal displacement of 12.8[mm].
(b)
Fig. 14 (Continued) (b) Temperature, displace- ment and
current graphs of SMA when heated with high voltage of
3[A] of duration 30[ms] in every 3[s] periodi- cally
SMA
temperature [°C]
Duty ratio [%]
Heating
Gradual
Pulse
Power
[W]
7.9
126
Energy
[J]
Actuation Cycles
When T°C ≈ 55°C
23.6
3.8
4
9
Table 2. Results of verification experiments
2.2.3 Temperature Control for Fast SMA Actuation
One of the advantages of pulse actuation method is in the
ability to control energy supply, hence the temperature
rise of the system by regulating the duty ratio of the
signals generated.
In the next experiments, by
implementing temperature control (Fig. 15), we show that
when the SMA oscillation width reaches a maximum or
constant value, the temperature of the alloy can be
controlled within a certain value which further increase
in temperature can be avoided.
Duty ratios of the input pulses were regulated in the
manner shown in the Table 3 and 4 for actuation of
1.33[Hz] and 2.0[Hz] respectively.
Fig. 15. Regulation of duty ratio
310
SMA
temperature [°C]
Duty ratio [%]
T <45
1.33
45<=T<47
47<=T<50
1.20
1.17
T =>50
1.13
Table 3. Temperature control by regulation of duty ratio
for 1.33[Hz] actuation
SMA
temperature [°C]
Duty ratio [%]
SMA
temperature [°C]
Duty ratio [%]
T <47
47<=T<49
49<=T<51
1.0
0.95
0.9
51<=T<53
0.85
T =>53
0.8
Table 4. Temperature control by regulation of duty ratio
for 2.0[Hz] actuation
On the contrary, implementing temperature control (Fig.
16(b)) enables the SMA temperature to be regulated
within the range where a stable oscillation of 1.72[mm]
can be maintained without further heating, A similar
result can be observed in Fig. 17 where a stable oscillation
of width 1.70[mm] was achieved slightly below the
temperature of 54[°C] for 2.0[Hz] actuation. High current
pulse actuation can be used for high speed actuation of
the SMA, coupled with temperature control by duty ratio
regulation the temperature of the SMA can be kept within
the range where stability of oscillation can be maintained
without further heating.
Chee Siong Loh; Hiroshi Yokoi & Tamio Arai / Natural Heat-Sinking Control Method for High Speed Actuation of the Shape Memory Alloy Actuator
(a) Without temperature control
strain), the metal outer tube provides an appropriate
housing to the SMA. One end of the wire is fixed using a
stopper while the other open end of the wire, which we
call the ‘force point’ is attached to the object needed to be
actuated. The end of the outer metal tube before ‘force
point’, we call the ‘pivot point’. This mechanism is
explained in Fig. 18.
Here, the SMA incorporated with the outer metal tube is
used in the actuation of a robotic finger, a project
currently in the research state in our laboratory. When
the SMA strains upon heating by applied current, the
stopper and the pivot point stop the contraction of the
wire, allowing just the force point to act as the actuation
point. The distance between the actuation point and the
pivot point is changeable depending on the nature of the
object to be actuated.
Force
Point
F
Pivot
point
T
1.72[mm]
Stopper
(b) With temperature control
Fig. 16. 1.33[Hz] actuation of SMA actuator
1.70[mm]
Fig. 17. 2.0[Hz] SMA actuation with temperature control
3.
Application
Incorporating the SMA with the metal outer tube enables
the increase in the application field of the SMA wire. In
the field of robotics where relatively long wires are used
to produce large movements (SMA has a 4%-5% recovery
3
1
T
T
T
2
SMA
Metal tube
Fig. 18. Application for the SMA wire when incorporated
with the metal outer tube in a robotic finger
Based on the pivot-force point concept explained above, a
more applicable SMA based actuator has been developed
(Fig. 19). A diameter of Φ 0.3[mm], length of 1.0[m] SMA
wire is inserted into the outer tube (Φ 0.8[mm]) and both
ends of the wire are each fixed to an electrode. An
opening of 25[mm] at the centre is made for the
movement of another electrode attached to the wire at its
centre. The electrodes are the points where current is
passed through. Between the end electrodes and the outer
metal tube is fixed a plastic stopper which acts as an
insulator, and also to stop the ends of the SMA wire from
contracting into the outer tube, thus allowing only the
middle part of the wire (centre electrode) to be
contracted, pulling the PE line in the direction of its
contraction. Object to be actuated is attached to the PE
line.
In order to actuate a degree of freedom (DOF), two
actuators are used for the flexion and extension
movements. When the finger is in flexion (Fig. 20), the
SMA wire in parts AB and EF are in their original lengths
while BC and DE are in extension. As the switches for AB
311
International Journal of Advanced Robotic Systems, Vol. 3, No. 4 (2006)
and EF circuits are turned off while switches for BC and
DE circuits are turned on, BC and DE contract, pulling the
wire in parts AB and EF into extension. A torque in the
clockwise direction occurs and the finger is extended as
shown. A reverse process can be expected for the flexion
of the finger when the orders for the switches are altered.
electrode
stainless outer tube(Ф0.8-1.6)
stainless pipe (Ф0.7-1.0)
PEEK Tube (Ф0.35-0.45)
plastic stopper
heat shrinking tube
SMA wire (Ф0.3)
nylon band
length when contract 500
extension appx.20 stainless pipe (Ф1.6-2.5)
PE line for stretching the SMA (Ф0.29)
full stroke 20
Fig. 19. Schematic drawing of the SMA based actuator
OFF
TORQUE
A
Power source
stainless
outer tube with
inserted SMA
B
ON
Electrode
C
Joint of finger
PE line for
stretching
SMA
D
ON
E
OFF
PE line for
stretching
SMA
F
Fig. 20. The mechanism of the actuator for actuation of
the robotic finger, clockwise torque (top) and
counterclockwise torque (bottom)
4. Conclusion
We described two methodologies for increasing the
actuation speed of the SMA. Firstly, we introduced a
simple heat sink consisting of the silicone grease and
outer metal tube. We verified the effectiveness of heatsink using FEM analysis and real world experiments.
Then we introduced a high current pulse actuation
method and coupled with temperature control, a
response of 2.0[Hz] of 1.70[mm] oscillation width has
been achieved. Lastly we described the initial efforts in
developing a new SMA based actuator and carried out
preliminary trials for its application in prosthetics.
5. Future Work
As for the extension of our work, we would like to
investigate the effectiveness of our two methodologies
312
when combined. For that purpose, an adaptive control
method to regulate the duty ratio of the pulse control to
the SMA incorporated with heat sink is necessary.
Mobilization will also be the research attention in our
future work in the application of the SMA.
6. References
Auricchio, F. (1995). Shape Memory Alloys: applications,
micromechanics, macromodelling and numerical
simulations. Ph.D Dissertation of University of
California at Berkeley, pp.4-6, Department of Civil
Engineering.
Bhattacharyya et. al (2001). Shape Memory Alloys,
Available
from
http://web.cs.ualberta.ca/~database/MEM
S/sma_mems/sma.html, Accessed: 2005-02-07.
Dono, S., Saito, A. & Kuwata, T. (2003). Knit Structure
SMA Actuator for Wearable Artificial Muscle
Systems. Matsushita Electronics Technical Report.
Elahinia, M.H. & Ashrafiuon, H. (2002). Non-linear
Control of a Shape Memory Alloy Actuated
Manipulator. ASME Journal of Vibration and Acoustics,
124:566-575, October 2002.
Eren, Y. et.al (2002). B-Spline Based Adaptive Control of
Shape Memory Alloy Actuated Robotic Systems.
Proceedings of IMECE ’02, 33425, November 2002.
Featherstone, R., & Teh, Y.H. (2004). Improving the Speed
of the Shape Memory Alloy Actuators by Faster
Electrical Heating. International Symposium of
Experimental Robotics (ISER’04), 128, Singapore, 18-21
June 2004.
Furuya, Y., & Shimada, H.(1990). Shape Memory
Actuators for Robotic Applications, In: Engineering
Aspects of Shape Memory Alloys, T.W Duerig et.al,
pp.
338-355,
ISBN
0-750-61009-3,
London:
Butterworth Heinemann, 1990.
Kuribayashi, K.(1991). Improvement of the Response of
an SMA Actuator Using a Temperature Sensor. The
International Journal of Robotics Research, vol. 10, pp.
13-20, February 1991.
Masahiro, S. (1995). Heat Transfer for Advanced Mechanical
Engineering, pp. 31-33, ISBN: 4130628267, University
of Tokyo Press, 1995.
Russell, R.A. & Gorbet, R.B. (1995). Improving the
Response of SMA Actuators. Proc. of the 1995 IEEE
ICRA, vol. 3, pp. 2299-2304, May 1995.
Seldon, B., Cho, K.J. & Asada, H. (2004). Segmented
Binary Control of Shape Memory Alloy Actuator
System Using the Peltier Effect. Proc. of the 2004
IEEE ICRA, pp. 4931-4936, 2004.
Seldon, B., Cho, K.J. & Asada, H. (2005). Multi Segment
State Coordination for Reducing Latency Time of
Shape Memory Alloy Actuator System. Proc. of the
2005 IEEE ICRA, pp. 1362-1367, 2005.