biocybernetics and biomedical engineering 40 (2020) 1359–1368

Available online at [Link]

ScienceDirect

journal homepage: [Link]/locate/bbe

Original Research Article

Simulation analysis for optimal design of

pneumatic bellow actuators for soft-robotic glove

Ning Guo a,b, Zhidao Sun c, Xiaojun Wang a,d, Eric Hiu Kwong Yeung b,
Michael Kai Tsun To a,d, Xiaodong Li a,b, Yong Hu a,d,*
a
Department of Orthopaedics and Traumatology, The University of Hong Kong, Hong Kong, China
b
Department of Physiotherapy, The University of Hong Kong - Shenzhen Hospital, Shenzhen, China
c
Bridge Engineering Design and Research Institute, China Railway Engineering Design and Consulting Group Co., Ltd,
Beijing, China
d
Department of Orthopaedics and Traumatology, The University of Hong Kong - Shenzhen Hospital, Shenzhen, China

article info abstract

Article history: Soft actuators that have a bellows structure are favorable candidates for robots designed to
Received 31 May 2020 interact with humans. However, a weak point in the actuator can occur as a result of
Received in revised form deformation from the driving pressure. In this study, a simulation analysis of a soft bellows
31 July 2020 actuator composed of ethylene-vinyl acetate copolymer molding was conducted. The
Accepted 5 August 2020 mechanical characteristics along different latitudes of the bellows in the soft actuator were
Available online 28 August 2020 evaluated using finite element modeling and analysis. Functional performance was studied
during both compression and inflation using two driving methods (constant pumping rate-
Keywords: driven and constant displacement-driven). To validate the simulation, experimental tests
Soft robot were performed on a version of the soft bellows actuator that was constructed according to
Soft pneumatic actuator the same specifications as the model version; simulation and experimental displacements
Bellows joint in relation to air pressure were compared. The results showed points near the trough were
Finite element modeling more likely to experience the largest stress during inflation and may suffer critical structural
damage. During compression, points near the crest were more easily damaged. Stress
variation showed good symmetry at points of interest on either side of the trough, during
both inflation and compression. These findings provide a basis for precise control of and
design improvements to soft bellows actuators for human-friendly usage.
© 2020 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish
Academy of Sciences. Published by Elsevier B.V. All rights reserved.

* Corresponding author at: Department of Orthopaedics and Traumatology, The University of Hong Kong, Hong Kong, China
E-mail address: yhud@[Link] (Y. Hu).
Abbreviations: AD, additive manufacturing; CD, constant deformation; CPR, constant pumping rate; CSR, commercial silicon rubber;
EVA, ethylene-vinyl acetate; ICC, intraclass correlation coefficient; POI, point of interest; TPU, thermoplastic polyurethanes.
[Link]
0208-5216/© 2020 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences. Published by Elsevier
B.V. All rights reserved.
1360 biocybernetics and biomedical engineering 40 (2020) 1359–1368

extension and the rod structure on the lower half creating

1. Introduction
bending stress that limits extension [14–16]. However, when
considering soft robotic glove applications, the half bellow–
Robots are widely used for industrial production, and also for half rod bending actuator cannot be properly customized to
convenience in daily life wherever human power can be the hand, as each finger comprises both deformable joints and
replaced, especially in the field of rehabilitation. Hand function non-deformable bones. Half bellow–half rod structure could
rehabilitation therapy is used in patients with neurological result in control instability and user discomfort [17]. Further-
disorders such as stroke [1], cerebral palsy [2], spinal cord more, the complex articulation of the thumb requires
injury [3], and Parkinson disease [4]. In the long term, consideration. Thus, it is better to combine the non-deform-
rehabilitation training can cost a large amount of time, energy, able rod part as bone, and a deformable soft actuator part for
and human resources [5]. Rehabilitation robots, such as articulation as separate components [18] when designing a
rehabilitation gloves for home use, are designed as a solution rehabilitation glove (Fig. 1). As such, the joint consists of an
to these problems; however, the primary design concern is entirely symmetrical bellows structure capable of expansion
safety, since the robot device physically interacts with a in every direction. Furthermore, the direction in which the
person. glove bends is driven by the natural motion trajectory of the
Soft robots have the potential to reduce stress concentra- hand. To mimic bending and unbending of the fingers, it is
tion and reproduce a variety of complex motions [6], making important to first quantify the mechanism of action of the
them perfect candidates for interactions with humans. Of the bellows actuator in response to air pressure.
variety of actuators that have been used in soft robots (e.g. Metal bellows are utilized to relieve stresses induced by
driven thermally [7], electrically [8], by shafts [9], or by other thermal expansion in a piping system. These types of bellows
means [10]), air-activated pneumatic soft actuators are the have been widely analyzed [19–21]. By contrast, soft bellows
most common. Pneumatic soft actuators have been used as actuators are used to provide motion different to that of metal
grippers and muscles [11] and are classified into fiber- bellows during compression and inflation processes. While the
reinforced or uniform bellows-type actuators [12]. The stress difference in soft bellows actuators with various layers
bellows-type actuator is the easiest to fabricate [12,13]; have been simulated under compression [22], the specific
however, its infinite degrees of freedom make it difficult to stress variation at different points of a soft bellows actuator
control. Precise control is very important in hand function has not been studied. Furthermore, only a few studies have
rehabilitation training to allow patients to reproduce different examined the complete bellows structure of soft actuators
complex motions and to grasp various objects. Thus, assess- with respect to shape optimization, deformation, and fabrica-
ment of the air pressure-deformation relationship is very tion method (Table 1).
important for evaluating the working mechanism of bellows- In the present study, the objective was to achieve precise
type soft actuators and for providing a basis for achieving control using a pressure and action curve and to determine
precise control of a robotic hand (designed with bellows-type optimal soft bellows actuator design during compression and
soft actuators) in rehabilitation. Furthermore, improvements inflation and by modeling the relationship between deforma-
in regions prone to structural weakness are necessary to tion displacement and air pressure of the bellows and the
prevent fatigue and breakage with use over long periods. stress in the material during the driving process using two
Most bellows-type actuators have a half bellow–half rod driving methods: a constant pumping rate (CPR) and constant
structure, with the bellow on the upper half providing deformation (CD). Furthermore, we aimed to examine the

Fig. 1 – Bellow joint actuator bending and unbending mimicking a finger.

 biocybernetics and biomedical engineering 40 (2020) 1359–1368 1361

Table 1 – Bellow element application in actuator.

Actuator/Sensor Fabrication Size(radius) Drive
PolyJet elastomer [23] Additive Manufacturing (AD) 30mm Pneumatic
Gly-KCl bellow fluidic channel [24] 3D printed + spin coated 12mm –
Polydimethylsiloxane [25] 3D printed 0.9mm Pneumatic
Multi-material robot [26] 3D printed + AD 25 mm Hydraulic
Shape memory alloy [27] AD 0.25 mm Hydraulic

productive manufacturing. Satisfactory elongation ratio could

guarantee sufficient joint motion. Thus, EVA plastic was used
as the material for the design of the bellow actuator in this
study. Different proportions of vinyl acetate to ethylene could
produce materials with various plastic and elastic properties
for different industrial purposes. In this study, the vinyl
acetate weight percent was 18%, with the remainder being
ethylene. The mechanical properties of the sample are shown
Fig. 2 – The structure and parameter of the bellow actuator in Table 2.
simulated in this experiment. To evenly distribute the applied air pressure at both ends of
the actuators, the surfaces of the bellows actuator were
separately meshed. The circular surfaces at each end were
meshed using a hex-dominated element shape and a medial
axis algorithm, while the lateral surfaces were meshed using
weak regions of the structure during the bellow deformation. an advancing front algorithm to better match the peaks and
In the air pumping procedure, knowledge of the weaknesses in troughs of the bellows structures.
the structure allowed the selection of a suitable pump, as well
as optimization of the design of the actuator structure, and 2.2. Defining points of interest
improvement of the actuator mold.
To examine the weak points of the bellows structure during air
pumping, we simplified the model and only studied the
2. Materials and methods stresses at points with representative information. Fig. 3
shows the points of interest (POIs) for a single bellows. Each
The compression and inflation experiments were simulated POI represents a latitude line. Based on the symmetry of a
with a bellows actuator model with V-shaped bellows. The circle, the stress at any point along the same line of latitude
pressure induced by the difference between standard atmo- will be the same during deformation. Thus, we simplified the
spheric pressure and the air pressure within the bellows was analysis by selecting one POI from each latitude. Each bellows
applied to the external surface of the bellow model, which had 10 latitude lines defined, and thus, had 10 POIs (A–J). The
consisted of 14 individual bellows and two circular ends. The intraclass correlation coefficient (ICC) was utilized to assess
models were designed in Solidworks (Dassault Systèmes Corp, the reliability of the stress value at corresponding POIs (i.e., the
France), and simulations were performed using Abaqus/CAE same latitude denoted by the same letter) from each of the 14
2016 (Dassault Systèmes Corp, France). bellows. Reliability was analyzed using statistical software
SPSS (IBM Corp, New York, U.S.).
2.1. Finite element model
2.3. Compression and inflation simulation based on CPR
The parameters of the bellows actuator used in this experi-
ment are described in Fig. 2. Because Solidworks has better The CPR condition simulated the response of the bellows
performance in model design than Abaqus, the bellows actuator when a constant air flow rate was pumped in and out
actuator was designed in Solidworks and imported into of the bellows. Although the air flow rate was constant, the air
Abaqus. pressure inside the bellows did not change linearly with time.
Thermoplastic polyurethanes (TPU), commercial silicon According to Eq. (1), for bellows deformation, the effective
rubber (CSR) and ethylene-vinyl acetate (EVA) are common pressure Peff is calculated as:
materials used to produce soft actuators. Under the same
structure, EVA bellow actuator could be fabricated with the
Peff ¼ Pp Ps (1)
highest precision, thinnest wall and best elongation ratio with

Table 2 – EVA mechanics properties in this study.

Material Young's modulus Poison's ratio Yield stress Plastic strain
EVA 80 N/ mm 2 0.4 10 MPa 6 MPa
1362 biocybernetics and biomedical engineering 40 (2020) 1359–1368

25 mm/0.4 s. Stress from same points of interests were

discussed. ICC analysis was performed on the ten averaged
POIs from different latitude of the bellow.

2.5. Theoretical analysis of relation between stress and

deformation of POI

As shown in Eq. (2), when the deformation (Dl) is constant, the

stress is only determined by Young's Modulus (E) no matter
how the velocity changes during deformation. Within elastic
Fig. 3 – POIs location takes one bellow for example. deformation range, the stress-deformation relation should be
linear and the scope of the stress-deformation would be E :

stress ¼ E  Dl (2)
representing the difference between the air inflation pressure
Pp and the bellows structure's resistance pressure Ps.
We imported pressure curve data that had been recorded
stress Dl
from the experimental results into the model. By controlling ¼E ¼ E (3)
Dt Dt
the effective pressure and the deformation time constant
between analytical and experimental data, the displacement Divided by Dt at both sides of Eq. (2), we could see from
could be compared. Boundary conditions were not set. Eq. (3) that the velocity () could influence the velocity of stress
During the compression process, atmospheric pressure was change, but it would not influence the order of stress
applied normal to the outer surface of the bellows actuators, magnitude. Therefore, it means that we could use a random
directed inward. For inflation, pressure was applied in the velocity to discuss structural weak points by the maximum
opposite direction. The loading pressure was set according to stress order during the deformation of different POI in both
the experimental data of the sensor to determine the driving conditions.
displacement difference of the simulated and experimental As acceleration is related to mass, Eq. (4) shows the relation
conditions under the same air-pressure driving conditions. between stress, mass (m), area (S) and acceleration (a):
In the preliminary CPR compression and CPR inflation
experiments, the bellows actuator began to exhibit surface
stress ¼ m  a=S (4)
contact between adjacent bellows after it was compressed too
hard. The surface contact first occurred at the distal portion of
the bellows actuator, and then the middle of the bellows In the present study, the mass of bellow is 0.2 g and the
actuator. As soon as contact began, the deformation process speed of the finger joint movement is quite slow. Thus, the
would translate from linear elastic deformation into nonlinear acceleration influence on the bellow damage is slight and we
plastic deformation. The stress-displacement curve became would not discuss influence from acceleration in this study.
nonlinear and irregular. The surface contact would make the
deformation too complicated to be discussed. Therefore, in the 2.6. Experimental data acquisition
present study we only assessed the process before surface
contact occurred. Based on our preliminary experiments, the To achieve high precision and sufficient elongation ratio as for
time periods of the bellows actuator compression and inflation the soft bellow actuator, blow molding was used to fabricate
steps were set as 0.4 s and 0.5 s, respectively, while the limit the EVA actuator.
values of the compression displacement and inflation dis- Then, the practical actuator was set in a control system to
placement were set as 25 mm and 13 mm, respectively. obtain real-time air pressure inside the bellow and the
deformation displacement during driving. The compression
2.4. Compression and inflation based on CD and inflation time of the control system is 0.4 s and 0.5 s
respectively. The air pressure was recorded during the whole
The CD condition was used to simulate the response of the bellow motion and was input to drive the bellow in the
bellows actuator when the displacement caused by compres- simulation. The real deformation displacement was to
sion or inflation varied evenly. In this process, the boundary compare with the simulation.
conditions were only set at both ends of the bellows actuator to The control system of the actuator is shown in Fig. 4. The air
simulate constant displacement. The reference step time of driving system worked the same as that of the CPR process in
each compression and inflation period lasted 0.4 s and 0.5 s, the modeling simulation. The pressure range of the pump was
respectively, to remain consistent with the CPR condition. from 60 kPa to 160 kPa. The bellows was pumped to a 25 mm
Different velocity of CD driving condition were simulated to compression displacement and a 13 mm inflation displace-
test different velocities influence during same deformation or ment during pumping periods of 0.4 s and 0.5 s, respectively.
displacement in both compression and inflation states. The Air pressure at the inlet of the bellows was measured during
velocities were simulated at 0.5, 1, 2 times the value of the process. The process was video-recorded using markers on
reference speed in the study. For inflation the reference speed both ends of the bellows to extract the compression and
is 13 mm/0.5 s and for compression the reference speed is inflation displacements. The final bellows displacement
 biocybernetics and biomedical engineering 40 (2020) 1359–1368 1363

Table 3 – p value from ICC analysis of POI from different

bellows.
Driving method CPR CD
Compression 0.971 0.994
Inflation 0.982 0.983

Because we would like to know the deformation difference

between practical CPR and idealized CD, the simulation time
Fig. 4 – Evaluation system monitor and control the soft and maximum deformation for CD and CPR were set the same,
bellow actuator. the converge of displacement at the end of the curves was
witnessed.

3.2. Statistical validation of points of interest definition

during compression and inflation was obtained by measuring rules
the change in pixel-values extracted from the video using
MATLAB 2019 (MathWorks Inc., Massachusetts, USA). In both driving conditions and for both deformation states, the
von Mises stress variation at points along the same latitude
exhibited 100% intraclass correlation, proving the structural
3. Results and discussion symmetry of the bellows actuator. The ICC of the maximum
von Mises stress values at corresponding POIs of the 14 bellows
The finite element model of the soft bellows actuator was during compression and inflation for both driving conditions
constructed, and the behavior of the actuator during com- are shown in Table 3. These data demonstrate repeatability of
pression and inflation was analyzed. The stress and displace- the stress characteristics.
ment performance in relation to air pressure were also
calculated in the CPR and CD process, and weak points in 3.3. Stress variation during CPR and CD compression and
the structure were determined. inflation

3.1. Displacement performance comparison between CPR The ICC results of different deformation velocities in simula-
and CD tion results indicated no difference with velocity variation
(p ¼ 1). The results indicate no significant influence of
The displacement-time relationships of both the CPR-and CD- different velocities on the weak points variation and sequence.
driving conditions are shown in Fig. 5. Shown in Fig. 5a, the Thus, in the following results, we will only discuss the
compression deformation for both CPR and CD conditions deformation with the reference velocity (0.4 s for 25 mm
were linear, which corresponded with the analytical explana- compression and 0.5 s for 13 mm inflation) to remain consis-
tion in Method 2.5. tent with the CPR condition.
From Fig. 5b, we could see the CD-driving reached 10 mm Fig. 6 shows the stress rate in relation to the displacement
elongation with constant speed and during CPR-driving of the POIs during the CPR compression process. Although the
inflation, the displacement was concave. This was probably POIs were evenly distributed along the bellows, the von Mises
because in CPR condition the bellow first went through elastic stresses were not evenly distributed during compression and
elongation, quickly reached the boundary and started plastic inflation processes. Besides, there were differences in the
deformation. And according to the results from preliminary stress velocity rate in different segment of the deformation
simulation, the maximum elongation of the bellow is 10 mm. shown in Figs. 7 and 8.

Fig. 5 – Displacement comparison between CPR and CD driving during inflation and compression.
1364 biocybernetics and biomedical engineering 40 (2020) 1359–1368

Fig. 6 – Stress variation of the POI under CPR and CD driving method in compression and inflation process.

Fig. 7 – The averaged maximum stress and stress velocity variation of POIs under CPR and CD inflation.
 biocybernetics and biomedical engineering 40 (2020) 1359–1368 1365

Fig. 8 – The averaged maximum stress and stress velocity variation of POIs under CPR and CD compression.

For the compression state in CPR/CD condition, the stress symmetrical POI on either side of the trough during both
changed linearly. While, the inflation curve of the first inflation and compression conditions.
segment of the curve was approximately linear (i.e. showed To have a better understanding of the maximum stress, the
an approximately constant stress rate), and then, in the velocity of stress variation should also be considered. Figs. 7–8
second segment, demonstrated an increasing stress rate. The b,d. showed the velocity variation of stress from each POI.
velocity change in inflation corresponded to the displacement According to Fig. 5, the deformation velocity during the elastic
curve (Fig. 5b) indicating the bellow deformation translated variation in CD-inflation, CD-compression and CPR compres-
from elastic translate to plastic. sion should be constant. For both driving methods (Fig. 7),
during the inflation process, the velocity of mise increment of
3.4. Mean maximum stress during CPR and CD most POI had remained constant until the plastic deformation
compression and inflation in the last part of the process. When plastic deformation
occurred, the velocity of mise increment enlarged. During the
The maximum stress value of the mean POIs (A–J) of the 14 compression process of CPR and CD driving method, the speed
bellows are shown in Fig. 7–8 a, c. These data indicated that of stress increment of D, E, F, H was witnessed slightly rise
inflation and compression from both driving methods had through most of the compression and the remaining POI
similar mean maximum von Mises trends. The average von decreased.
Mises stress of the E latitude showed the greatest difference By combining both maximum stress and the stress velocity
from those of the other bellows latitudes. variation, during compression, latitude F and G both had the
For the CPR driving method, during the inflation process,
the A latitude was the point of largest stress, followed by
points B, E, F, and G. During compression, point G showed the
greatest stress concentration, followed by E, F, and B.
For the CD driving method, the stress concentration was
similar to that of CPR, except that the stress of the B latitude
decreased during inflation, increased during compression, and
had closer von Mises stress values to those at points E, F, and G.
These data suggest that different methods of pumping may
induce differences in stress concentration.
During inflation, point A was most likely to exhibit the
largest stress, and may suffer critical structural damaged.
However, point G was most likely to be damaged during Fig. 9 – Origin state, compression state and inflation state of
compression. The stress variation showed good symmetry at the soft actuator.
1366 biocybernetics and biomedical engineering 40 (2020) 1359–1368

risk of suffering structural damage. Because G has the

maximum stress and stress at latitude F has the most
significant growing velocity. During inflation, A latitude was
the weak point that may priority had structural damage.

3.5. Experimental validation with a fabricated version of

the soft bellows actuator

An experimental version of the soft bellows actuator was

fabricated with copolymer blow molding techniques using the
same materials and size parameters as those used in the
simulation model. The original, compression, and inflation
states of the fabricated model are shown in Fig. 9.
Fig. 11 – air pressure sensor results in CPR inflation and
Linear regression between analytical and experimental
compression process.
displacement was calculated to test the dependent relation.
For compression state: r = 0.984, p = 0.00 and inflation state:
r = 0.933, p = 0.00 indicating significant prediction rate that
analytical result has on experimental result. During compres- the surface more quickly, thus the deformation was less
sion, the air pressure inside the bellows increased from 0 kPa delayed for the latter segment of the curve.
to 31.41 kPa when the bellows was compressed 26.35 mm from Inflation curve could be divided into two segments to
its original state during 0.4 s of compression. In the inflation discuss. At the beginning of inflation, the simulated deforma-
process, the air pressure increased 53.70 kPa from its original tion was quicker than the experimental deformation (Fig. 10b).
state, and the bellows expanded 12.98 mm. The same air This result shared the same explanation principle as Fig.10a. It
pressure variation curves were also used to drive the finite might be caused by the time latency for the air pressure input
element method model in both the CPR compression and to be evenly distributed to the entire bellow. As the space near
inflation conditions. the outlet was priority filled and with higher density, the inside
The change in displacement relative to the air pressure bellow remains lower air pressure, which caused latency
inside the bellows actuator of the experimental and simulated deformation for practical. For the latter part of the curve, the
CPR models are shown in Fig. 10. Although not perfectly analytical deformation became slower and reached the
matched, the simulated model and experimental model had a deformation limit before the practical. The reasonable
similar displacement after compression when the inside air explanation was that there was a slight material property
pressure reached 31.41 kPa. Compared with the analytical difference when fabricating the real bellow actuator, which
curve, there was a delay in the deformation of the experimen- was unavoidable in fabrication techniques. When the linear
tal curve. This was likely because, during compression, as the elastic deformation translated to non-linear plastic deforma-
air near the outlet was preferentially pumped out, the reduced tion, the difference between the model and real bellow became
air pressure inside could not immediately become evenly obvious. Finally, the inflation displacement of the simulated
distributed to the bellows surface. However, with compression CPR inflation was 12.98 mm, while the experimental model
of the bellows actuator, the space inside the bellows reached its limit when the inflation deformation was
decreased, allowing the reduction in air pressure to reach 10.32 mm. The air pressure on the outlet of the bellows is

Fig. 10 – Normalized Analytical-Experiment curve of compression and inflation in CPR.

a. compression process; b. inflation process.
 biocybernetics and biomedical engineering 40 (2020) 1359–1368 1367

shown in Fig. 11 and represents the effective pressure applied

for bellows deformation. Acknowledgments

3.6. Outlook and limitation This work was partially supported by Shenzhen/Hong Kong
Innovation Circle Programme (Type D
The simulation method in discussion of structural weakness SGDX2019081623201196) and High Level-Hospital Program,
and pressure-deformation relation in this study could be Health Commission of Guangdong Province, China
applied to other bellow shaped soft material, such as silicon (HKUSZH201902033). We thank Coren Walters-Stewart, PhD,
rubber, thermoplastic polyurethanes and other elastic soft and Edanz Group ([Link]
polymer that could present linear elastic deformation. com/ac) for editing a draft of this manuscript.
The present study has some limitations. First, the abnormal
conditions of the bellow actuator were not under concerned as
references
there were amount of different performances of abnormal
states, such as bending and local deflated. These conditions
will be discussed in our future work. Second, in the compres-
[1] Weiwei Zhao CJ, Wang Daozhen, Shen Huifen. Clinical
sion state, we only discussed the deformation before the
observation of effects of ultrashort wave therapy combined
surface contact and non-linear plastic deformation. Third, due with acupuncture and rehabilitation training in the
to the little mass of the bellow and low velocity in actuator treatment of patients with dysphagia after stroke. Journal
motion, we did not consider the influence of acceleration of Neurorestoratology 2019;7:136–42.
change on the structural weakness of the bellow actuator. [2] Reid LB, Rose SE, Boyd RN. Rehabilitation and
neuroplasticity in children with unilateral cerebral palsy.
Nat Rev Neurol 2015;11:390–400.
4. Conclusions [3] Wenbin Ding SZ, Dajiang Wu, Zhang Yanpeng, Ye Hualong.
Hand function reconstruction in patients with chronic
incomplete lower cervical spinal cord injury by nerve
We performed a simulation analysis and experimental segment insert grafting: a preliminary clinical report.
validation of a pneumatic bellows joint actuator design Journal of Neurorestoratology 2019;7:129–35.
composed of ethylene-vinyl acetate copolymer. The actuator [4] Kadkhodaie M, Sharifnezhad A, Ebadi S, Marzban S, Habibi
was simulated during compression and inflation in both CPR- SA, Ghaffari A, et al. Effect of eccentric-based rehabilitation
on hand tremor intensity in Parkinson disease. Neurol Sci
and CD-driven conditions. The air pressure-deformation
2020;41:637–43.
relation was analyzed and well consistent with the experi-
[5] Yue Z, Zhang X, Wang J. Hand rehabilitation robotics on
mental data using an actuator of the same size and material. poststroke motor recovery. Behav Neurol 2017;2017.
During inflation, points near the trough were more likely to 3908135-3908135.
exhibit the largest stresses and may suffer critical failure. [6] Wallin TJ, Pikul J, Shepherd RF. 3D printing of soft robotic
During compression, points near the crest were more likely to systems. Nat Rev Mater 2018;3:84–100.
be damaged. Stress variation was symmetric for POIs on either [7] Tang X, Li K, Liu Y, Zhou D, Zhao J. A soft crawling robot
driven by single twisted and coiled actuator. Sens Actuators
side and equidistant from the trough during both inflation and
A 2019;291:80–6.
compression. In the air pumping procedure, knowledge of the [8] Zhou F, Zhang M, Cao X, Zhang Z, Chen X, Xiao Y, et al.
weaknesses in the structure allowed the selection of a suitable Fabrication and modeling of dielectric elastomer soft
pump, as well as optimization of the design of the actuator actuator with 3D printed thermoplastic frame. Sens
structure, and improvement of the actuator mold. These Actuators A 2019;292:112–20.
findings provide a basis for developing precise control of soft [9] Tan N, Gu X, Ren H. Design, characterization and
applications of a novel soft actuator driven by flexible
bellows actuators in human-friendly soft robotics.
shafts. Mech Mach Theory 2018;122:197–218.
[10] P. Boyraz, G. Runge, and A. Raatz, An overview of novel
Conflict of interest actuators for soft robotics, Multidisciplinary Digital
Publishing Institute, pp. 48.
[11] Hall KL, Phillips CA, Reynolds DB, Mohler SR, Rogers DB,
The authors declare that they have no known competing Neidhard-Doll AT. Haptic control of a pneumatic muscle
financial interests or personal relationships that could have actuator to provide resistance for simulated isokinetic
exercise; Part II: control development and testing. Comput
appeared to influence the work reported in this paper.
Methods Biomech Biomed Eng 2015;18:1–14.
[12] Wang Z, Hirai S. Chamber dimension optimization of a
CRediT authorship contribution statement bellow-type soft actuator for food material handling. 2018
IEEE International Conference on Soft Robotics (RoboSoft).
2018. pp. 382–7.
Ning Guo: Visualization, Methodology, Software, Writing - [13] Marchese AD, Katzschmann RK, Rus D. A recipe for soft
original draft. Zhidao Sun: Methodology, Software, Data fluidic elastomer robots. Soft Robot 2015;2:7–25.
[14] Wang T, Ge L, Gu G. Programmable design of soft pneu-net
curation. Xiaojun Wang: Validation, Data curation. Eric Hiu
actuators with oblique chambers can generate coupled
Kwong Yeung: Investigation, Resources. Michael Kai Tsun To: bending and twisting motions. Sens Actuators A
Investigation, Resources. Xiaodong Li: Validation, Data cura- 2018;271:131–8.
tion. Yong Hu: Conceptualization, Writing - review & editing, [15] Salem MEM, Wang Q, Wen R, Xiang M. Design and
Supervision. characterization of soft pneumatic actuator for universal
1368 biocybernetics and biomedical engineering 40 (2020) 1359–1368

robot gripper. 2018 International Conference on Control and [22] Zhou L, Li L, Zhao C. Performance analysis of the multi U-
Robots (ICCR). 2018. pp. 6–10. shaped bellows based on CAE. Mater Sci Forum 2014;800-
[16] De Payrebrune K, O'Reilly O. On the development of rod-based 801:390–3.
models for pneumatically actuated soft robot arms: a Five- [23] Dämmer G, Gablenz S, Hildebrandt A, Major Z. PolyJet-
parameter constitutive relation. Int J Solids Struct 2017;120. printed bellows actuators: design, structural optimization,
[17] Moviglia MTMBGustavo A, Couto Damián, Piccone Samanta. and experimental investigation. Frontiers in Robotics and
Local immunomodulation and muscle progenitor cells AI 2019;6:34.
induce recovery in atrophied muscles in spinal cord injury [24] Soomro AM, Khalid MAU, Shah I, Kim Sw, Kim YS, Choi KH.
patients. Journal of Neurorestoratology 2018;6:136–45. Highly stable soft strain sensor based on gly-KCl filled
[18] Zhou JS, Chen XJ, Chang Y, Lu JT, Leung CCY, Chen YH, et al. sinusoidal fluidic channel for wearable and water-proof
A soft-robotic approach to anthropomorphic robotic hand robotic applications. Smart Mater Struct 2020;29025011.
dexterity. Ieee Access 2019;7:101483–95. [25] Jung W, Kang Y, Han S, Hwang Y. Biocompatible micro, soft
[19] Pachpande D, Tembhare G, Wagle M. Stress analysis of bellow actuator rapidly manufactured using 3D-printed
thick wall bellows using finite element method. soluble mold. J Micromech Microeng 2019;29125005.
International Journal of Engineering Research and V4 2015. [26] MacCurdy R, Katzschmann R, Kim Y, Rus D. Printable
[20] Gawande SH, Pagar ND, Wagh VB, Keste AA. Numerical hydraulics: a method for fabricating robots by 3D co-
investigations on characteristics of stresses in U-shaped printing solids and liquids; 2016.
metal expansion bellows. International Journal of Metals [27] Spindler C, Juhre D. Development of a shape memory alloy
2015;2015:9. actuator using generative manufacturing. The
[21] Robyr S. FEM modelling of a bellows and a bellows-based International Journal of Advanced Manufacturing
micromanipulator. Helsinki University of Technology; 1999. Technology 2018;97:4157–66.