Mechatronics 63 (2019) 102280

Contents lists available at ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

A threshold gyroscope based on a bistable mechanism☆

Hong Van Tran a, Tien-Hoang Ngo a, Pei-Lun Chang a, I-Ting Chi a, Ngoc Dang Khoa Tran b,
Dung-An Wang a,b,∗
a
Graduate Institute of Precision Engineering, National Chung Hsing University, 250 Kuo Kuang Rd., Taichung 40227, Taiwan, ROC
b
Faculty of Mechanical Engineering, Industrial University of Ho Chi Minh City, 12 Nguyen Van Bao, Ward 4, Go Vap District, Ho Chi Minh City, Vietnam

a r t i c l e i n f o a b s t r a c t

Keywords: This paper describes a threshold gyroscope based on a bistable mechanism (BM). The device is capable of sensing
Bistable one distinct signal when angular rate threshold is exceeded along one axis. This design minimizes the integrating
Threshold scheme required by the analog gyroscope to obtain the threshold angular rate input. The automatic detection
Gyroscope
of the threshold value is achieved by the snap-through behavior of the BM and a wireless sensing scheme. A
model is developed for efficient design of the device. Feasibility and effectiveness of the device is validated by
experiments.

1. Introduction olds can be estimated by integrating signals of analog gyroscopes, the

high computational cost may be undesirable in applications that require
Automatic detection of angular velocity beyond certain threshold real-time output. A threshold gyroscope that can detect preset threshold
may see a number of applications, such as detection of falls in the elderly angular acceleration may reduce the computing and calibration effort.
[1,2] and gait stabilization of humanoid robots [3]. Wu [4] used three Precise states of positions provided by switches, latches, compliant
cameras and motion analyses to detect fall prior to impact by thresh- multistable mechanisms can be employed for detection of threshold val-
olding the horizontal and vertical velocity profiles of the trunk of hu- ues of angular velocity. With the advantage of no backlash, open-loop
man subjects. In the attempt to deploy hit protectors to cushion the fall, control, zero power consumption to remain in stable equilibrium states,
accelerometers and gyroscopes can be employed to obtain the velocity and reduced noise sensitivity [8], compliant multistable mechanisms
profiles. Bourke and Lyons [1] used a gyroscope to detect falls at impact. are suitable for design of threshold gyroscopes. Compliant BMs have
A threshold-based algorithm was needed to discriminate between falls been employed to detect acceleration thresholds. Todd et al. [9] em-
and normal activities in their approach. Huynh et al. [2] developed a fall ployed a BM to design a threshold accelerometer. Zhao et al. [10] pre-
detection algorithm based on data collected from 3-axis accelerometers sented a threshold accelerometer based on a tristable mechanism, where
and gyroscopes in a wearable sensor system. Critical sensor thresholds magnetic forces are utilized to switch the equilibrium positions of their
were adopted in their algorithm. Baltes et al. [3] used rate feedback from tristable mechanism. Tran et al. [11] proposed a threshold accelerome-
two gyroscopes to stabilize the walking gait of a humanoid robot. Their ter based on a tristable mechanism. Their design is capable of detecting
motion compensation algorithm is active when the gyroscope reading is two acceleration thresholds. Frangi et al. [12] devised a shock sensor
larger/smaller than the maximum/minimum velocity threshold. based on a bistable cosine-curved beam. Their device can sense acceler-
In order to obtain the threshold signals from analog measurement ations in the sense that it has two preferential directions. Nelson et al.
of the traditional gyroscopes, additional computing resource is needed. [13] combined a bistable beam and a triboelectric generator for sensing
Chen et al. [5] reported that constant calibration in output gain and of threshold acceleration. By increasing the compressive axial force of
input excitation might be required in the micro gyroscopes used in a the bistable beam their device can sense various acceleration thresholds.
wearable sensor system. Hong and Park [6] pointed out that due to the In this investigation, a threshold gyroscope with a compliant BM
numeric integration process of angular velocity (rate) output for the an- embedded in an oscillator is developed. The drive-mode oscillator main-
gle estimation purposes, gyroscopes may suffer from the accumulated tains a linear momentum and the sense-mode BM may switch between
angular error over time. A high precision control and tuning scheme its two stable equilibrium positions due to the Coriolis force generated
was used by Hu and Gallacher [7] to reduce the angle errors of their by the drive vibration and an angular rate input. A semi-analytical model
vibratory gyroscopes. Although accurate angular acceleration thresh- for design of the device is described. Finite element analyses (FEA) are

☆
This paper was recommended for publication by Associate Editor Chun-Yi Su.
∗
Corresponding author at: Graduate Institute of Precision Engineering, National Chung Hsing University, 250 Kuo Kuang Rd., Taichung 40227, Taiwan, ROC
E-mail address: [email protected] (D.-A. Wang).

https://doi.org/10.1016/j.mechatronics.2019.102280
Received 14 January 2019; Received in revised form 12 August 2019; Accepted 21 September 2019
0957-4158/© 2019 Elsevier Ltd. All rights reserved.
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Ωz
Substrate Solenoid actuator, SA 3
Shuttle mass Sense
direction
Solenoid actuator, SA1 y

V-beam Anchor

V-beam
spring

Drive
direction
Decoupling frame
x
Solenoid actuator, SA 2

Fig. 1. A schematic of a threshold gyroscope.

carried out to examine the effectiveness of the model. A readout scheme two horizontal segments (segments 1 and 3) and one slanted segment
for passive sensing of the positon change of the BM is presented. The (segment 2). The initial positon of the shuttle mass can be taken as the
performance of the device is demonstrated by experiments. first equilibrium position of the BM. The combination of the initial de-
flected shape of the support beams with a double beam configuration
achieves bistability of the BM. When a force f is applied to the shuttle
2. Design mass in the +𝑦 direction, a displacement 𝛿 is produced. As f is increased
and reaches a critical value, the shuttle mass snaps into its second equi-
2.1. Operational principle librium position as seen in Fig. 2(b). Fig. 2(c) is an illustration of a typ-
ical force-displacement curve of the BM. The force–displacement plot is
Fig. 1 is a schematic of a threshold gyroscope. A Cartesian coor- based on a displacement control scheme, where the force is controlled
dinate system is also shown in the figure. The drive direction and the with certain constant displacement increments. The first and second
sense direction are along the x and y direction, respectively. The angular equilibrium position of the BM are marked by S1 and S2 , respectively. U
speed Ωz with respect to the z direction is the rate input. The threshold is the unstable equilibrium position of the BM. fmax is the critical force
gyroscope consists of a BM, a decoupling frame, two V-beam springs, magnitude for the BM to switch from S1 to S2 during the forward motion
and three solenoid actuators. The BM, comprised of a shuttle mass and of the BM. In reality, when the shuttle mass of the BM reaches the unsta-
four support beams, is embedded in the frame, which is suspended by ble equilibrium position U during the forward motion, the BM snaps into
two V-beam springs. Two solenoid actuators, SA1 and SA2 , generate a its second equilibrium position S2 without traversing through the nega-
drive-mode oscillation. A solenoid actuator, SA3 , is employed to reset tive portion of the 𝑓 − 𝛿 curve. |fmin | is the critical force magnitude to
the position of the BM. The V-beam springs are compliant in the drive reset the BM back to S1 . The BM with two stable equilibrium positons is
direction and have high stiffness in the sense direction. Other types of intended to detect the threshold angular rate signal in the environment.
springs, such as fixed-fixed flexure, straight folded beam, crab-leg spring When the angular rate is higher than the threshold value, the induced
and serpentine spring [14], can also provide the functional suspension Coriolis force has a magnitude greater than fmax and therefore the BM
of the BM. Zhou and Dowd [15] found that tilted folded beam (V-beam) jumps to S2 . For repetitive use of the device, the solenoid actuator SA3
can be less stiff than straight folded beam. Saggere et al. [16] reported is actuated to push the shuttle mass of the BM back to S1 .
that the V-beam design provides a large displacement with a linear re-
sponse in the desired motion direction, and maintains about one order
of magnitude higher stiffness in the undesired direction of motion. The 2.2. Modeling
decoupling frame is implemented to mechanically decouple the drive
and sense direction oscillations. A lumped model is developed to predict the dynamical behaviors of
Fig. 2(a) shows the initial configuration of the BM. The end of the the device. Fig. 3(a) is a schematic of the lumped model. The decoupling
support beam connected to the frame can be represented by a clamped frame is employed to minimize forces in the sense-mode due to drive-
end boundary condition. As shown in Fig. 2(b), the support beam has mode actuator imperfections. The drive-mode oscillator with the frame
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Clamped end Shuttle mass Support beam y

Sense
Substrate direction Decoupling frame

Frame mass, m f
δ
y
Segment 1 Anchor
Segment 2 kx Drive
x f Segment 3 direction
Shuttle mass, ms
x
FSA
(a) (b)
cx
f(y) cy

f max

f( yi-1)
y (a)
f( yi )
Ly
f( yi+1) x
Δ
F
Force, f

S1 U S2
0 w2
Lx
yi-1 yi yi+1

w1 l2
θθ
l1

fmin

(b)
Displacement, δ
(c)
L2
Fig. 2. The first (a) and second (b) equilibrium position of a BM. (c) A typical W1 Δi
W2
force-displacement curve of a BM. W3
φ

mass mf and the shuttle mass ms is connected to a substrate via a linear

spring of stiffness kx . Mechanical damping is represented by a dashpot
L1 L3
with a damping coefficient cx . A force due to the solenoid actuators FSA
applied to the proof mass makes it translate in the x direction. FSA has
sinusoidal form with a magnitude of F0 and a frequency of 𝜔SA . The (c)
sense-mode oscillator is formed by the shuttle mass ms , a damper and a
Fig. 3. (a) A lumped model of the device. (b) A schematic of the V-beam spring.
nonlinear spring. Assuming that the device is very stiff in the z direction,
(c) A quarter model of the BM.
the equations of motion of the device for a constant angular rate input
Ω̇ 𝑧 = 0 are

( ) 𝑑2𝑥 ( ) ( ) 𝑑𝑦 model (BCM) [17] is used to find the stiffness kx of the V-beam spring.
𝑑𝑥
𝑚𝑓 + 𝑚𝑠 + 𝑐𝑥 + 𝑘𝑥 𝑥 = 𝐹0 sin 𝜔𝑆𝐴 𝑡 + 2 𝑚𝑓 + 𝑚𝑠 Ω𝑧 The BCM can give a relation between the force, F, and displacement,
𝑑 𝑡2 𝑑𝑡 𝑑𝑡
( ) Δ, of the quarter model of the V-beam spring as illustrated in Fig. 3(b).
+ 𝑚𝑓 + 𝑚𝑠 Ω2𝑧 𝑥 (1) When the displacement is small, F may be represented as
𝑘𝑥
𝐹 = Δ (3)
𝑑2𝑦 𝑑𝑦 𝑑𝑥 4
𝑚𝑠 + 𝑐𝑦 + 𝑓 (𝑦) = −2𝑚𝑠 Ω𝑧 + 𝑚𝑠 Ω2𝑧 𝑦 (2)
𝑑 𝑡2 𝑑𝑡 𝑑𝑡
The equations for determination of the stiffness kx of the V-beam spring
where cy is the damping coefficient of the dashpot in the sense direction. based on a reduced-order BCM can be found in Appendix A.
f(y) is the nonlinear spring force of the BM. Since the force-displacement An analytical model based on a chained beam constraint model
curve of the BM is highly nonlinear (see Fig. 2(c)), the value of f(y) can (CBCM) [18] is developed to obtain the force-displacement curve f(y)
be interpolated from the force-displacement curve at the value of y. In of the BM. The CBCM is based on the BCM where each flexible beam
the interpolation process, a set of (y, f(y)) data are sampled at the force- is divided into a few elements and each element is modeled by the
displacement curve as illustrated in Fig. 2(c). A unknown force f(yi ) at BCM. A quarter model of the BM is considered in the development of
a specific query displacement yi is calculated using linear interpolation the analytical model due to geometry symmetry and the loading con-
of two neighboring displacements 𝑦𝑖−1 and 𝑦𝑖+1 , where 𝑦𝑖−1 < 𝑦𝑖 < 𝑦𝑖+1 . dition. Fig. 3(c) shows a quarter model of the BM. The support beam
For angular rates at much lower frequencies than the driving fre- has three segments. The length and width of segment i (from 1 to 3) of
quency 𝜔SA of FSA (t), the terms Ω2𝑧 can be neglected. Fig. 3(b) is a quarter the support beam are Li and Wi , respectively. The inclination angle of
model of the V-beam spring, where l1 and w1 are the length and width segment 2 is 𝜙. The geometry parameters of the beam are marked in
of the left half of the V-beam, respectively, and l2 and w2 are the length the figure. Each segment of the support beam is modeled by the BCM
and width of the right half of the V-beam, respectively. 𝜃 is the half of developed in Appendix A. The element number of segment 1, 2, and
the angle between the two segments of the V-beam. A beam constraint 3 used in CBCM is 1, 3 and 1, respectively. Given a displacement Δi
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

View A Fig. 4. Mesh of the device (a), a quarter of the

V-beam spring(b), and the BM (c).

Fixed end Decoupling frame

Fixed end

Fixed end

y
A

x
(a) (b)

Fixed end Fixed end

(b)
Fixed end Fixed end

Displacement

(c)

(indicated by an arrow in the figure) and using the force/moment equi- In order to obtain the stiffness of the V-beam spring, kx , a displace-
librium equations, kinematic equations for the geometry constraints and ment in the x direction is applied at the decoupling frame as seen in
the load-deflection relations of the segments, the analytical expressions Fig. 4(a). Fixed boundary conditions are applied to the anchors of the
for obtaining the force-displacement curve of the BM can be found and V-beam spring to restrain the displacement degrees of freedom (see
solved. See Appendix A for the equilibrium equations, kinematic equa- Fig. 4(a)). The reaction force at the fixed ends is computed and divided
tions and the load-deflection relations based on the BCM. The analytical by the displacement to obtain kx . Fig. 4(c) shows a model to attain the
expressions also hold for the BM after the initial elastic phase. nonlinear force-displacement relation of the BM. A displacement is ap-
In order to verify the solution of kx for the V-beam spring and plied in the y direction to the top end of the BM. Displacement degrees
f(y) (the nonlinear force-displacement curve) for the BM based on the of freedom of the nodes at the fixed ends are constrained.
reduced-order BCM and CBCM, respectively, two dimensional finite el- 4-node plane strain elements of type CPE4R are used in the finite
ement analyses were carried out. An engineering thermoplastic, poly- element model. CPE4R is a 4-node bilinear, reduced integration with
oxymethylene (POM), is assumed to be the material of the device in hourglass control element. Fig. 4(a) shows meshes of the decoupling
the analyses. The Young’s modulus E and the Poisson’s ratio 𝜈 are frame and the V-beam springs. Fig. 4(b) shows the meshes of a quarter
taken as 2.6 GPa and 0.39, respectively. The density of the material is of the V-beam spring. Fig. 4(c) shows the meshes of the BM. There are 4
1.39 g/cm3 . The force-displacement relations for the V-beam spring and elements through the beam depth of the V-beam spring and the beams
the BM are computed by the static analysis of the commercial finite of the BM. With reduced integration the number of elements through
element program ABAQUS, where inertia effects and time-dependent the beam depth plays a critical role. For bending problems, four ele-
material effects are neglected. During the static analysis, automatic dis- ments through the depth provide acceptable results [19]. The V-beam
placement increment scheme is adopted to reach stable and accurate in Fig. 4(b) and one support beam of the BM have 2912 and 1440 ele-
solutions. The large-displacement formulation of ABAQUS is used to ac- ments, respectively. The decoupling frame and the shuttle mass have rel-
count for large-displacement effects and geometrically nonlinear behav- atively coarse meshes because they behave nearly as rigid bodies when
ior. Newton’s method is used by ABAQUS to solve nonlinear equilibrium the beams of the device are deformed. The finite element model for the
equations. device has 18,256 elements.
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

10.00 mm

12.00 mm
254.68 mm

95.33 mm

54.80 mm
60.00 mm 33.92 mm

8.00 mm
2.00 mm

8.00 mm
5.00 mm

124.96 mm
179.68 mm

(a)
3.00 mm
14.00 mm
20.00 mm

45.00 mm mm
13.20 mm

0.80
0m
80.0

0
4.5
0.80 mm

40.0

10.00 mm 10.00 mm
3.00 mm

0
0m

40.0
m

(b) (c)

Fig. 5. Dimensions of the device (a), the slanted beam of the BM (b), and the V-beam spring (c).

2.3. Analysis device is 5.5 mm. Fig. 6 shows the force-displacement curve of the V-
beam spring based on the reduced-order BCM and the finite element
An initial design of the device is obtained using an iterative trial and analyses of the full model shown in Fig. 4(a). The results based on the
error design process. Fig. 5(a) show the values of the geometry param- solution of the reduced-order BCM are in good agreement with the finite
eters of the device. The detailed dimensions of the BM and the V-beam element analyses. The kx values (slope of the force-displacement curve)
spring are shown in Fig. 5(b) and (c), respectively. The thickness of the are estimated as 118.4 and 108.9 N/mm based on the reduced-order
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

1.6 2
FEA (full model shown in Fig. 4(a)) FEA
1.4 Reduced-order BCM CBCM
1.5 Experiment
1.2 Experiment

Force (N)
1
Force (N)

0.8
0.5
0.6

0.4 0
0.2
-0.5
0
0 2 4 6 8 10 12 0 1 2 3 4 5 6 7
Displacement (mm) Displacement (mm)
(a)
Fig. 6. Force-displacement curve of the V-beam spring.
2
FEA
CBCM
1.5 Experiment
BCM and the finite element analyses, respectively. When the displace-
ment is larger than 4 mm, the reduced-order BCM results deviate from

Force (N)
the finite element analyses due to the nonlinear, large deformation of the 1
V-beam. During operation of the device, the displacement of the V-beam
spring is restricted to 4 mm to avoid the nonlinear stiffness observed in
Fig. 6. Fig. 7(a) shows the force-displacement curves of the BM based 0.5
on the CBCM model and the finite element analyses. The results pre-
dicted by the CBCM model and the finite element analyses are in good
0
agreement. The values of fmax and fmin based on the CBCM model are
1.11 and −0.49 N, respectively, and the fmax and fmin obtained from the
finite element analyses are, 1.09 and -0.48 N, respectively. The excellent -0.5
agreement between the model and finite element analyses for kx , fmax
0 1 2 3 4 5 6 7
and fmin proves the effectiveness and accuracy of the developed models.
The reduced order BCM and the CBCM can be employed to at- Displacement (mm)
tain acceptable force-displacement relation of the V-beam spring and (b)
the BM. Li and Chen [20] proposed a rational function with a cu-
bic polynomial numerator and a quadratic polynomial denominator to Fig. 7. Force-displacement curves based on the CBCM model, the finite element
analyses and experiments for forward motion (a) and backward motion (b).
describe kinetostatic behaviors of compliant mechanisms accurately.
Force-displacement relation of the V-beam and the BM can be explicitly
represented by their approach. For modeling compliant bistable mech-
anisms, six closed-form equations are required to describe the relation- with complicated geometry, CBCM can be a robust approach to solve for
ships between the rational function parameters and mechanism’s design the kinetostatic behaviors of beam-type compliant mechanisms.
parameters, which are achieved using a multi-variable nonlinear regres- The material of the device, POM, is a polymer, which has a high
sion [20]. Without resorting to the rational function and regression anal- damping ratio. The vibrating beam method has been used to measure
ysis, the CBCM method can be employed for various types of compliant damping ratios of engineering materials where structural damping is
mechanisms for modeling and design. The CBCM has been extended for crucial for dynamic simulations of mechanical structures [25–28]. Cata-
analyzing large spatial deflections of flexible, curved beams [21]. nia and Sorrentino [29] adopted the vibrating beam method to measure
Qiu et al. [22] adopted a mode superposition method to obtain the damping ratios of several engineering materials. A cantilever beam
a force-displacement curve of a bistable cosine-curved beam. Their with a small flexural deformation is considered in the vibrating beam
buckling-beam based analysis was shown to be accurate and to pro- method. The behavior of the first mode of vibration of a cantilever beam
vide insight into the behavior of a single beam. Brake et al. [23] applied can be approximated to the behavior of a one-degree-freedom mass-
a large deformation theory to obtain the force-displacement curve of a spring-damper system, which governing equation can be written as
non-straight bistable beam. In their model, large deflection equations
for the beam are established with the beam’s bending behavior mod- 𝑥̈ + 2𝜍 𝜔1 𝑥̇ + 𝜔21 𝑥 = 0 (4)
eled by a moment-rotation angle relation, and the equations are solved where 𝜔1 and 𝜍 are the resonant frequency of the first mode and damp-
by a boundary value solver. Provided a reasonably close initial guess, ing ratio, respectively, and x is the tip deflection of the cantilever beam.
their numerical method allows for the boundary conditions to be exactly 𝜔1 and 𝜍 can be expressed as
satisfied and an acceptable force-displacement relation can be achieved. √
Andò et al. [24] applied the theory of the pseudorigid body model [8] to 𝜔1 = 𝑘∕𝑚 (5)
calculate the force-displacement curve of a bistable beam. Accurate esti-
mation of spring constants of flexural pivots and beams are critical to the ( )
𝜍 = 𝑐∕ 2𝑚𝜔1 (6)
accurate solution of the force-displacement curve. Frangi et al. [12] and
Nelson et al. [13] also used the mode superposition method to produce where k and m are the stiffness and the mass of the one-degree-freedom
force-displacement curves of bistable beams. For compliant mechanisms mass-spring-damper system. c is the damping coefficient. The solution
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Table 1 4
Values of the lumped mass, damping coefficient and spring Model
constant of the device. 3.5 Experiment

RMS Amplitude of x (mm)

Parameter Value
3
Frame mass mf (g) 32.5
Shuttle mass ms (g) 23.2 2.5
Damping coefficient cx (N sec/m) 0.0148
Damping coefficient cy (N sec/m) 0.0842 2
Spring constant of V-beam spring kx (N/m) 0.109
f(y) force-displacement curve of the BM See Fig. 7(a) 1.5
Slope of f(y) at 𝑦 = 0, ky (N/mm) 2.122
1

to Eq. (4) with the following initial conditions of 𝑥(0) = 𝑥0 and𝑥̇ (0) = 0 is
0.5
given as 0
( ( ))
𝑥0 √ −𝜍 0 5 10 15 20 25 30 35 40
−𝜍 𝜔1 𝑡
𝑥 (𝑡 ) = √ 𝑒 cos 𝜔1 1 − 𝜍 𝑡 + tan
2 −1
√ (7)
1 − 𝜍2 1 − 𝜍2 Frequency (Hz)
If the tip of the cantilever beam is driven to an initial displacement x0 (a)
and then left free to oscillate, the parameters 𝜔1 and 𝜍 can be determined
by fitting Eq. (7) to the experimental data measured with a vibrating 2.5
cantilever beam. Then, the damping coefficient c is calculated by Eq. (6). Up (model)
Down (model)
A typical value of the damping ration 𝜍 for POM materials can be
Up (experiment)

RMS Amplitude of y (mm)

taken as 0.01 for a cantilever beam based on the work of Catania and 2
Down (experiment)
Sorrentino [29]. The damping characteristics of materials are influenced
by the operating conditions, the past history and the geometry of the
1.5
damping mechanism [30]. It has been found that the structural damp-
ing is approximately proportional to the square of vibration amplitude
[31]. During the design stage of the threshold gyroscope, the damping 1
ratios of the drive mode and the sense mode, 𝜍 x and 𝜍 y , respectively, are
assumed to have a value of 0.003 and 0.006, respectively, for analyzing
the dynamic response of the device. The damping coefficient cx and cy 0.5
can be expressed as
√
𝑐𝑥 = 2𝜍𝑥 𝑘𝑥 (𝑚𝑓 + 𝑚𝑠 ) 0
√ 0 5 10 15 20 25 30 35 40
𝑐𝑦 = 2𝜍𝑦 𝑘𝑦 𝑚𝑠 (8)
Frequency (Hz)
where ky is the slope of the force-displacement curve f(y) at 𝑦 = 0.
(b)
Table 1 lists the values of the lumped masses, damping coefficients and
spring constant of the device. Fig. 8. Frequency response of the V-beam spring (a) and the BM (b).
The drive and sense mode resonant frequencies should be separated
in certain amount in order to achieve a stable response and to remove
the sensitivity to fabrication imperfections and fluctuations in operating
conditions [32]. In order to obtain the frequency response of the drive down sweep frequency response of the BM based on Eq. (10) is shown in
mode, the drive mode of Eq. (1) is excited by the motion of a shaker and Fig. 8(b) with 𝜔shaker ranging from 1 to 40 Hz and a harmonic vibration
the terms including Ωz and Ω2𝑧 of Eq. (1) are neglected. The equation to amplitude X of 0.07 mm. The increment of the up and down frequency
obtain the frequency of the drive mode can be expressed as sweep for the sense mode is 0.1 Hz. Sudden jump phenomena can be
( ) 𝑑2𝑥 𝑑𝑥 ( ) ( ) seen near 29.7 and 22.5 Hz during up sweep and down sweep, respec-
𝑚𝑓 + 𝑚𝑠 + 𝑐𝑥 + 𝑘𝑥 𝑥 = 𝑚𝑓 + 𝑚𝑠 𝑋𝜔2shaker sin 𝜔shaker 𝑡 (9)
𝑑 𝑡2 𝑑𝑡 tively. As shown in Fig. 8(b), the BM exhibits a spring softening nonlin-
earity. The jump-up and jump-down frequencies of the sense mode are
where 𝜔shaker and X are the vibration frequency and amplitude of the
much higher than the drive mode resonance frequency, which is a char-
shaker, respectively. Similar to the drive mode oscillator, the equation
acteristics desirable for stable operation. Acar and Shkel [32] reported
for the sense mode oscillator excited by the motion of a shaker can be
that operating close to sense mode resonance peak makes the system
written as
very sensitive to variations in system parameters that may cause s shift
𝑑2𝑦 𝑑𝑦 ( )
𝑚𝑠 + 𝑐𝑦 + 𝑓 (𝑦) = 𝑚𝑠 𝑋𝜔2shaker sin 𝜔shaker 𝑡 (10) in the resonant frequencies or damping. Matching the drive mode and
𝑑𝑡 2 𝑑𝑡 sense mode resonant frequencies requires that the operating frequency
Fig. 8(a) shows the frequency response of the V-beam spring based has to be controlled with extreme precision. It is difficult to control the
on the model of Eq. (9) with 𝜔shaker ranging from 1 to 40 Hz and a har- drive-mode and sense-mode frequencies with the precision required for
monic vibration amplitude X of 0.1 mm. The resonant frequency of the mode-matched devices given the structural and environmental effects
V-beam spring is nearly 7 Hz based on the model. Frequency response of that result in large variations in the resonant frequencies [32]. There-
the sense mode is obtained by both increasing frequency (up sweep) and fore, it is a common practice by setting the sense-mode frequency away
decreasing frequency (down sweep) to characterize nonlinear behavior from the drive-mode frequency.
of the BM. Upon changing to each frequency, a dwell time is allowed for Model predictions of the response of the device subjected to various
the BM to attain a steady state response. Frequency sweep with an incre- angular rate inputs are shown in Figs. 9 and 10. The driving frequency
ment of 0.4 Hz for the drive mode and sense mode is used. Up sweep and 𝜔SA and the amplitude F0 of FSA (t) are selected as 14 Hz and 1.450 N,
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

300 300
270
260 260
250
200 200

Ω z (rpm)
Ω z (rpm)

150 150

100 100

50 50

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Time (s) Time (s)
(a) (d)
25 25
20 20
15 15

Displacement, x (mm)
Displacement, x (mm)

10 10
5 5
0 0
-5 -5
-10 -10
-15 -15
-20 -20
-25 -25
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Time (s) Time (s)
(b) (e)
12 12
10 10
S2 S2
Displacement, y (mm)

Displacement, y (mm)

8 8
6.9 6.9
6 6
4 4
2 2
0 0
-2 -2
-4 -4
-6
0 10 20 30 40 50 60 70 80 90 100 -6 0 10 20 30 40 50 60 70 80 90 100
Time (s) Time (s)
(c) (f )

Fig. 9. Time history of the angular rate input 250 rpm (a), the displacement of the drive mode (b) and displacement of the sense mode (c). Time history of the
angular rate input 270 rpm (d), (e) and (f) are the displacement history of the drive mode and the sense mode, respectively, for the angular rate input of 270 rpm.

respectively. The drive mode is actuated at time = 0 s. The value of F0 is displacement fluctuates around the first stable equilibrium position of
selected to generate a vibration amplitude of the drive mode of 4.4 mm. the BM, an intrawell vibration mode (see Fig. 9(c)). At the angular rate
The angular rate input is switched on at time = 50 s when the drive mode input of 270 rpm, the sense mode displacement reciprocates between
has settled into its steady state. The time history of the angular rate in- the first and second stable equilibrium positions of the BM, exhibiting
put Ωz of 250 and 270 rpm are shown in Fig. 9(a) and (d), respectively. an interwell vibration mode (see Fig. 9(f)). While the BM oscillates in
Fig. 9(b) and (c) are the time histories of the drive mode displacement the interwell mode (oscillating between the first and second stable equi-
and the sense mode displacement, respectively, when the angular rate librium states) and the drive mode is turned off, the BM may settle into
input is 250 rpm. At the angular rate input of 250 rpm, the sense mode one of the two stable equilibrium positions. We can design a passive
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280
Position of shuttle mass of BM (mm)

S2 = 6.9
z

y Solenoid actuator, SA 3

S1 = 0

259 260 Gyroscope

Spacer
0 5 100 150 200 250 300 350 400 450 500

Ω z (rpm)
Substrate
Fig. 10. Position of the BM of the sense mode at various angular rate input.

fixture to grip the shuttle mass once the shuttle mass enters the sec-
ond equilibrium position, so that the BM remains in the second stable
configuration after the transition.
Fig. 10 shows the traveling range of the BM of the sense mode at the Solenoid actuator, SA 1
angular rate inputs from 5 to 500 rpm. The BM has an interwell vibration Solenoid actuator, SA 2
when the angular rate input is greater than or equal to 260 rpm. 260 rpm
can be taken as the threshold angular rate of the gyroscope. The results Fig. 11. An exploded view of a fabricated prototype.
shown in Figs. 9 and 10 give the evidence that the jumping phenomenon
of the BM is convenient for sensing the threshold of the angular rate
input. The design is based on a simple two degree-of-freedom dynamic
model, where the values of the structural damping are assumed. The
model is intended for the initial design of the device.
Translation stage 2cm
3. Fabrication and testing

3.1. Fabrication

Prototypes of the device were fabricated by a numerical controlled

milling machine in order to demonstrate the feasibility of the threshold
gyroscope. POM is selected as the material of the device due to its ease
of machining, relatively high mechanical strength and wear resistance. Force gauge
Fig. 11 is an exploded view of the assembly of the prototype. It consists x
of the device, two spacers, a substrate plate, and three solenoid actu- Probe
ators. The three layers are fastened tightly together by screws through y z
the holes milled in the three layers. The drive mode motion is actuated
by the solenoid actuators, SA1 and SA2 (MD-266-2, Maruha Electric Co., (a)
Ltd., Japan). The solenoid actuator, SA3 , serves as a reset mechanism to
move the BM back to its first stable equilibrium position after the event Translation stage
of snap through of the BM. The dimensions of the device are indicated Probe
in Fig. 5. A groove along the symmetry axis of the BM is utilized for
alignment of a force gage with the BM during testing.

3.2. Testing
y
Fig. 12(a) and (b) are the photos of the experimental setup to acquire Force gauge BM
the force-displacement curves of the drive mode and the sense mode, z x 2cm
respectively, of a fabricated device. As seen in the figure, a prototype
was fixed on an optical table. A force gage (DS2-5 N, Dgzhiqu Precision (b)
Co., Ltd., China) was held in a two-axis translation stage. The probe
Fig. 12. Experimental setup for measurement of force-displacement curves of
axis of the force gage was aligned with the symmetric axis of the drive
the (a) drive mode; (b) sense mode of a fabricated prototype.
mode and the sense mode of the device as seen in Fig. 12(a) and (b),
respectively, during testing. The displacement of the probe was read
from a Vernier micrometer of the two-axis translation stage.
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Fig. 14 is a schematic of the experimental setup for measurement

of the threshold angular rate of the device. The angular rate input was
Shaker
provided by a centrifuge (TC-16, Torng Yih Machinery Works Co., Tai-
wan). The centrifuge was driven by an inverter motor which can rotate
up to 3000 revolutions per minute (rpm). A variable frequency drive
Decoupling (VFD007M43B, Delta Electronics, Inc., Taiwan) was used to control the
frame variable speed of the inverter motor. A prototype was fixed at the cen-
Point A Solenoid, ter of a rotary disk of the centrifuge. The solenoid actuators, SA1 and
SA 2 SA2 , provided the drive mode motion. A microcontroller board, Arduino
Solenoid,
SA 1 Substrate MEGA 2560, was used to control the solenoids. Since the solenoids re-
quire higher current than the microcontroller can provide, a power cir-
Point B Controller cuit with a ULN2803AP transistor was built to drive the solenoids. The
position of the BM of the device was detected by an infrared reflective
Laser sensor, IR sensor 1 (TCRT5000, Vishay Intertechnology Inc., US) fixed
LK-H020
Laser on the rotary disk. The rotational speed of the rotary disk was mea-
LK-H150 sured by an infrared reflective sensor, IR sensor 2, which was affixed
at the edge of the rotary disk and senses light reflected from a shiny
target fastened on the cylinder wall of the centrifuge. A Wifi board with
a microchip ESP8266 (WeMos D1 R2 V2.1.0 WiFi, WEMOS Electronics,
China) transmitted the sensor output signal wirelessly to a computer.
Fig. 13. Experimental setup to obtain frequency response of the sense mode of
A Li-Ion 10.8-volt battery was utilized to supply power for the sensors,
the device.
the power circuit, the microcontroller board and the Wifi board. Fig. 15
(a-b) are photos illustrated the position sensing scheme for the BM us-
ing IR sensor 1. Fig. 15 (c) and (d) shows the rotational speed sensing
scheme for the rotary disk using IR sensor 2.
Experiments were carried out to obtain the frequency responses
of the drive mode and the sense mode of the threshold gyroscope. 4. Results and discussion
Fig. 13 shows an experimental setup to obtain the frequency response of
the sense mode. The base excitation was provided by an electrodynamic Fig. 6 shows the experimental force-displacement curve of the drive
shaker (APS 113, APS Dynamics, Germany). Two laser displacement sen- mode of the prototype. The experiments were repeated five times at
sors LK-H150 and LK-H020 (Keyence, Japan) measured the vibrations each displacement, and the averaged force value was taken. The force in-
of the point A of the BM and the point B of the shaker, respectively, creases almost linearly as the displacement increases based on the exper-
for the sense mode test. LK-H150 and LK-H020 were used to measure iments. The experimental results agree with those based on the reduced-
the vibrations of the decoupling frame and the shaker, respectively, for order BCM. If a straight line extended from the elastic region is plotted
the drive mode test. The vibration amplitude of the shaker was kept in Fig. 6, the solutions based on the FEA and the reduced order BCM
at 0.1 mm during the frequency sweep tests of the drive mode and the are slightly higher and lower than the straight line in the displacement
sense mode. ranging from 4 to 12 mm. The experiments should be closer to the FEA

Solenoid actuator, SA 1 Fig. 14. Schematic of an experimental setup

O1 Power I1 O1 Arduino
for testing the threshold of the angular rate
Solenoid actuator, SA 3 circuit MEGA 2560
O3 I2 O2 input.
O2 I1 I2 RX/TX
IR sensor 2
UART

Solenoid actuator, SA 2

Wifi board
ESP8266

IR sensor 1

Shiny target
RS485

Centrifuge Computer
Variable
Rotary disk frequency drive
VFD007M43B

Motor
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Solenoid, IR sensor 1 Solenoid, IR sensor 1

SA 3 SA 3

Solenoid, Solenoid, Solenoid, Solenoid,

SA 1 SA 2 SA 1 SA 2
Position 1 Position 2

(a) (b)

Arduino Wifi board Arduino

MEGA 2560 ESP8266 MEGA 2560

Power
Battery circuit Battery

Power circuit Wifi board

ESP8266
Shiny target IR Sensor 2 IR Sensor 2 Shiny target

(c) (d)
Fig. 15. (a-b) Position sensing scheme for the BM. (c-d) Rotational speed sensing scheme for the rotary disk.

results than the reduced order model. This discordance can be attributed be seen near 29.5 and 22.5 Hz during up sweep and down sweep, respec-
to the machining and measurement errors. Fig. 7(a) and (b) show the tively. The jump up frequency and jump down frequency based on the
experimental force-displacement curves of the BM during forward mo- model are 29.7 and 22.5 Hz, respectively. The jump-up frequency and
tion and backward motion, respectively. The experiments were repeated the jump-down frequency of the sense mode based on the experiments
five times at each displacement. The averaged value of the experiments are in good agreement with the model predictions. The experimental
was shown in the figure. The experimental force-displacement curves of vibration amplitudes near the jump down frequency of the down sweep
the BM during the forward motion (Fig. 7(a)) and the backward motion are slightly lower than those based on the model. Note that the damp-
(Fig. 7(b)) agree with those predicted by the CBCM analytical model and ing ratio of the model is assumed to be constant. Structural damping
the finite element analyses. As seen in Fig. 7(a), the experimental value generally increases with response amplitude [33]. Near the jump down
of fmax is slightly higher than that based on the analytical model and the frequency, large vibration amplitudes were observed, and the damping
finite element analyses. The discrepancy between the experiments and of the device might increase to cause slightly lower vibration amplitudes
the model may be attributed to the machining error. than those predicted by the model.
Fig. 8(a) and (b) show the frequency responses of the drive mode Fig. 16(a) shows the time history of the angular rate input of the
and the sense mode based on the experiments. The vibration amplitudes threshold gyroscope during the experiments for a time span of 180 s. The
of the shaker were 0.1 and 0.07 mm to obtain the frequency response time history of the displacement of the drive mode and the sense mode
of the drove mode and sense mode, respectively. The increment of the are shown in Fig. 16(b) and (c), respectively. As shown in Fig. 16(b),
frequency sweep was 0.1 Hz. As shown in Fig. 8(a), the experimental the vibration amplitude of the drive mode, 4.4 mm, was kept the same
resonant frequency of the drive mode, 7 Hz, is nearly the same as that as that in the simulation (see Fig. 9(b) and (e)). The vibration frequency
based on the model. As seen in Fig. 8(b), sudden jump phenomena can of the drive mode was 14 Hz. The angular rate input was increased from
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

300 247

250 246

245
Ω z (rpm) 200

Ω z (rpm)
244
150
243
100
242
50 241
0 240
0 20 40 60 80 100 120 140 160 180 161 161.5 162 162.5 163 163.5 164
Time (s) Time (s)
(a) (d)
15

10 10
Displacement x (mm)

Displacement x (mm)
5 5

0 0

-5 -5

-10 -10

-15 -15
0 20 40 60 80 100 120 140 160 180 161 161.5 162 162.5 163 163.5 164
Time (s) Time (s)
(b) (e)

S2
S2
Equilibrium position
Equilibrium position

S1 S1
0 20 40 60 80 100 120 140 160 180 161 161.5 162 162.5 163 163.5 164
Time (s) Time (s)
(c) (f )

Fig. 16. Time history of (a) angular rate input; (b) displacement of the drive mode; (c) displacement of the sense mode; (d-f) are the enlarged view of (a-c),
respectively, near the threshold angular rate, 246 rpm.

0 to 252 rpm (see Fig. 16(a). At the beginning, the displacement of the rate computed by the model, 260 rpm, the experimental threshold value
sense mode oscillated around its first stable equilibrium position, S1 . As is 5% lower. The prediction of the threshold value based on the model
the angular rate input was increased, the displacement of the sense mode can be considered acceptable given the small percentage of error. It is
became larger. When the displacement of the sense mode reaches its sec- assumed that structural damping is significant in the device considered
ond stable equilibrium position, S2 , the corresponding angular rate in- in this investigation. Damping ratio may depend on the frequency and
put is the threshold value of the gyroscope. Fig. 16(d–f) are the enlarged amplitude of response. However, the damping ratios of the drive mode
views of Fig. 16(a–c), respectively, near the time when the snap-through and the sense mode, 𝜍 x and 𝜍 y , respectively, are assumed to be constants
of the BM occurred. As seen in Fig. 16(f), near the time of 162 s, the sense with values of 0.003 and 0.006, respectively. The simulated result may
mode reached the position S2 , corresponding to an angular rate input Ωz be over-damped and exhibited a slightly higher value than the measured
of 246 rpm (see Fig. 16(d)). Compared to the threshold value of angular threshold angular input rate. Slight variations in structure geometry,
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Measured contour
Designed contour

Designed contour

Measured contour

Fig. 17. Surface contours of the fabricated device and the designed device.

boundary conditions and material properties can cause asymmetric vi- to obtain the transient response. Fig. 18(a) and (b) show the unit-step
bration modes [34]. It can be seen from Fig. 17 that the contour of the input of the angular rate Ωz of 260 rpm and the transient response of the
fabricated device deviates slightly from the ideal device geometry. An device to the step input, respectively. The rise time to the angular rate
image dimension measurement system (VMM-D, Yakon Accuracy Ltd., input was estimated as 1.35 s. For a lower damping ratio 𝜍 y of 0.001, the
Taiwan) with a measurement precision of 3 micrometer was utilized rise time is nearly 0.24 s. Materials with lower damping characteristics
to measure the dimensions of the fabricated prototype. Some vibration can be considered in the design of the threshold gyroscope in order to
modes are sensitive to the small device asymmetry, which might be con- achieve quick response to the input [35].
sidered to be a possible reason that the testing result deviates from the We explored the relation between the critical force fmax of the BM
simulated one, which is obtained under ideal device geometry. and the threshold angular rate input of the device. Fig. 19 shows the
Ideally, the output of the gyroscope should follow sudden changes in threshold angular rate input as a function of the critical force magni-
the input instantaneously. However, transient phenomena exist before tude fmax based on the model. The value of fmax was adjusted while
the steady state is reached. The transient response of the device to a unit- keeping the first and second equilibrium position of the BM, S1 and
step angular rate input was investigated by the rise time required for its S2 , respectively, as S1 = 4.7 mm and S2 = 6.9 mm. The driving fre-
response to rise from 0% to 100% of the threshold value. The model quency 𝜔SA and the amplitude F0 of FSA (t) are selected as 14 Hz and
based on Eqs. (1) and (2) with the damping ratios 𝜍𝑥 = 0.003 and 𝜍𝑦 = 1.450 N, respectively, for the drive mode of Eq. (1). The case of fmax =
0.006 for the drive mode and the sense mode, respectively, were used 1.09 N of the device is marked by a triangle in the figure. The thresh-
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

300
380
260
360

Threshold angular input rate (rpm)

250
340

200 320
Ω z (rpm)

300
150 280
260
100
240

50 220
200
0 180
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 1 1.02 1.04 1.06 1.08 1.1 1.12 1.14 1.16 1.18 1.2
Time (s) Critical force (N)
(a)
Fig. 19. The threshold angular rate input as a function of the critical force ma
fmax based on the model.
12

10
300
8
6.9
Displacement, y (mm)

245 246 245 245 246 245

6 250

4
200
Ω z (rpm)

2
150
0
100
-2 100
66
-4 1.35
50 30
-6
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0
Time (s) 0 10 20 30 40 50 60 70 80 90 100 110
(b) Time (s)
(a)
Fig. 18. (a) Unit-step input of the angular rate signal of 260 rpm. (c) Transient
response of the sense mode to the step input.

S2
old angular rate input increases nearly linearly as the critical force
Equilibrium position

increases.
The response of the BM under a cyclic history was examined by in-
creasing the angular rate input Ωz up to a value exceeding the thresh-
old and then decreasing under it. Fig. 20(a) and (b) is the time his-
tory of the angular rate input, and the corresponding displacement his-
tory of the sense mode. As Ωz was set at a value below the threshold
value, 246 rpm, the sense model had an intrawell oscillation about the
first equilibrium position S1 of the BM. When Ωz was increased to the
threshold angular rate, 246 rpm, the sense mode exhibited an interwell
oscillation between its first equilibrium position S1 and second equi-
librium position S2 . While Ωz was lowered from the threshold value, S1
the sense mode returned to the intrawell oscillations about position S1 . 0 10 20 30 40 50 60 70 80 90 100 110
As Ωz was raised higher up to 246 rpm, the sense mode oscillated be- Time (s)
tween position S1 and position S2 until Ωz was lowered to 245 rpm. It
appears that the threshold angular rate is not dependent on the loading (b)
path.
Fig. 20. Time history of the angular rate input (a), and displacement of the
The present design of the threshold gyroscope is in a macroscopic sense mode (b).
scale. Miniaturization of the device and integration with signal pro-
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

cessing, telemetric communication and power management circuits can [ ][ ][ ]

𝑤21 𝑝1 𝛿𝑦1 𝛼1 6 1
− 10 𝛿𝑦1
pave the way for its mass production by microfabrication technologies. 𝛿𝑥1 = − 5
The current design of the device is restricted to sense a single threshold 12𝑙12 2 1
− 10 2
𝛼1
15
value of angular rate input in one axis. A sensor array can be used to [ ][ ]
[ ] − 1 1
𝛿𝑦1
detect multiple threshold values. A multiaxial threshold gyroscope can − 𝑝 1 𝛿𝑦 1 𝛼1 700 1400 (A5)
be constructed using multiple uniaxial gyroscopes [36].
1 11
− 6300 𝛼1
1400

5. Conclusions [ ][ ][ ]
𝑤22 𝑝2 𝛿𝑦2 𝛼2 6 1
− 10 𝛿𝑦2
𝛿𝑥2 = − 5
A threshold gyroscope based on a BM is proposed and characterized. 12𝑙22 2 1
− 10 2
𝛼2
15
A design method based on a semi-analytical model is developed for ef- [ ][ ]
[ ] − 1 1
𝛿𝑦2
ficient design of the device. The designed threshold value is very close − 𝑝 2 𝛿𝑦 2 𝛼2 700 1400 (A6)
to the experimental result. A wireless signal transmission scheme is de-
1 11
− 6300 𝛼2
1400
vised for sensing of the angular rate threshold. This investigation proves
where w1 and w2 are the beam width of the left half and the right half
the design concept of the threshold gyroscope.
of the V-beam, respectively. Equilibrium at the joint of the left half and
Declaration of Competing Interest the right half of the V-beam requires
𝑝1 𝐸 𝐼1 [ ( ) ( ) ] 𝐸 𝐼2
− cos 𝛼1 − 𝜃1 + 𝜃2 𝑝2 − sin 𝛼1 − 𝜃1 + 𝜃2 𝑓2 =0 (A7)
None. 𝑙12 𝑙22
𝑓1 𝐸 𝐼1 [ ( ) ( ) ] 𝐸 𝐼2
Acknowledgement − sin 𝛼1 − 𝜃1 + 𝜃2 𝑝2 + cos 𝛼1 − 𝜃1 + 𝜃2 𝑓2 =0 (A8)
𝑙12 𝑙22
The computing facilities provided by the National Center for High- [ ]
𝑚1 𝐸 𝐼1 𝑚2 𝐸 𝐼2 𝑓2 𝐸 𝐼2 ( ) 𝑝 𝐸𝐼
Performance Computing (NCHC) are greatly appreciated. The authors − + 1 + 𝛿𝑥2 − 2 2 𝛿𝑦2 = 0 (A9)
𝑙1 𝑙2 𝑙2 𝑙2
are also thankful for the financial support from the Ministry of Science
and Technology, R.O.C., under Grant no. MOST 105-2221-E-005-060. The geometric constraint of the V-beam can be expressed as
[ ] [ ( ) ( )][ ]
𝐿𝑥 cos 𝜃1 − sin 𝜃1 𝑙1 𝛿𝑥1 + 𝑙1
Appendix A. Equations to determine the stiffness of the V-beam = ( ) ( )
𝐿𝑦 − Δ sin 𝜃1 cos 𝜃1 𝑙1 𝛿𝑥1
spring based on the BCM
[ ( ) ( )][ ]
cos 𝛼1 + 𝜃1 − sin 𝛼1 + 𝜃1 𝑙2 𝛿𝑥2 + 𝑙2
Fig. A1(a) shows the original and deformed shape of the V-beam. We + ( ) ( ) (A10)
sin 𝛼1 + 𝜃1 cos 𝛼1 + 𝜃1 𝑙2 𝛿𝑥2
divide the V-beam into two elements as illustrated in Fig. A1(b) and (c).
The 𝑥 − 𝑦 coordinate system represents the global coordinate frame. The
𝛼1 + 𝛼2 = 0 (A11)
𝑋1 − 𝑌1 and 𝑋2 − 𝑌2 coordinate system are the local coordinate frame
of the left half and right half of the V-beam, respectively. As seen in where Lx , Ly and Δ are illustrated in Fig. 3(b). Note that 𝐿𝑥 = 𝑙1 cos(𝜃1 ) +
Fig. A1(b), P1 , F1 and M1 are the axial force, transverse force and mo- 𝑙2 cos(𝜃2 ), and 𝐿𝑦 = 𝑙1 sin(𝜃1 ) + 𝑙2 sin(𝜃2 ). Using Eqs. (A3–A11) and pre-
ment, respectively, applied at the end of the left half of the V-beam. ΔX1 , scribing a displacement, Δ, of the V-beam, the loads p1 , f1 , m1 , p2 ,
ΔY1 and 𝛼 1 are the corresponding tip deflections. P2 , F2 , M2 , ΔX2 , ΔY2 f2 , and m2 , and the deflections 𝛿 x1 , 𝛿 y1 , 𝛼 1 , 𝛿 x2 , 𝛿 y2 and 𝛼 2 can be
and 𝛼 2 are defined in a similar manner for the right half of the V-beam solved numerically. The external force F applied to the V-beam, which
(see Fig. A1( c)). The rotation of the local coordinate frame of the left is equilibrated by the loads at the fixed end of the V-beam, can be
half and the right half of the V-beam with respect to the global coordi- written as
nate frame are denoted by 𝜃 1 and 𝜃 2 , respectively. Note that 𝜃1 = 𝜃 and [ ( ) ] 𝐸 𝐼1
𝜃2 = 𝜋 − 𝜃, where 𝜃 is the half of the angle between the two segments of 𝐹 = 𝑓1 cos 𝜃1 + 𝑝1 sin(𝜃1 ) (A12)
𝑙12
the V-beam in the undeformed configuration. The loads and deflections
are normalized as Therefore, a value obtained by dividing F by Δ within the linear range
of the F-Δ curve can be taken as the stiffness of the quarter model of the
𝑀1 𝑙1 𝐹1 𝑙12 𝑃1 𝑙12 Δ𝑌1 Δ𝑋1
𝑚1 = , 𝑓1 = , 𝑝1 = , 𝛿𝑦 1 = , 𝛿𝑥 1 = (A1) V-beam spring.
𝐸 𝐼1 𝐸 𝐼1 𝑙1 𝑙1 𝑙1 A reduced-order model can be built in order to provide rapid design
𝐹2 𝑙22 𝑃2 𝑙22 prediction. The BCM for the V-beam can be simplified by truncating
𝑀2 𝑙2 Δ𝑌2 Δ𝑋2
𝑚2 = , 𝑓2 = , 𝑝2 = , 𝛿𝑦 2 = , 𝛿𝑥 2 = (A2) second-order terms of 𝑝21 and 𝑝22 in Eq. (A3) and (A4), respectively, as
𝐸 𝐼2 𝐸 𝐼2 𝑙2 𝑙2 𝑙2 [ ]
[ ] [ ][ ] 6 1 [ ]
where I1 and I2 are the second moment of inertia of left half and right 𝑓1 12 −6 𝛿𝑦1 − 10 𝛿𝑦1
= + 𝑝1 5 (A13)
half of the V-beam, respectively, and E is the Young’s modulus of the 𝑚1 −6 4 𝛼1 − 1 2
𝛼1
10 15
V-beam material. Based on the BCM, the load-deflection relations can
[ ] [ ][ ] [ ][ ]
be written as [17] 𝑓2 12 −6 𝛿𝑦2
6 1
− 10 𝛿𝑦2
[ ] [ ][ ] [ ] = + 𝑝2 51 (A14)
6 1 [ ] 𝑚2 −6 4 𝛼2 − 10 2
𝛼2
𝑓1 12 −6 𝛿𝑦1 − 10 𝛿𝑦1 15
= + 𝑝1 51
𝑚1 −6 4 𝛼1 − 10 2
𝛼1 Considering the case of the V-beam with 𝜃1 = 𝜃, 𝜃2 = 𝜋 − 𝜃, 𝑤1 = 𝑤2 =
15
[ ][ ]
1 1 𝑤, and𝐼1 = 𝐼2 = 𝐼, and substituting Eqs. (A13)-(A14) and (A5)-(A6) into
− 700 𝛿𝑦1
+ 𝑝21 1400 (A3) Eqs. (A7–A11), we can express a new set of six equations with six inde-
1
− 11 𝛼1
1400 6300 pendent variables, p1 , p2 , 𝛿 y1 , 𝛼 1 , 𝛿 y2 and 𝛼 2 , as

[ ] [ ][ ] [ ][ ] [ ( )]
𝑓2 12 −6 𝛿𝑦2
6 1
− 10 𝛿𝑦2 𝑝1 − cos(−2𝜃+ 𝛼1 )𝑝2 − sin(−2𝜃+ 𝛼1 ) 12𝛿𝑦2 − 6𝛼2 + 𝑝2 1.2𝛿𝑦2 − 0.1𝛼2
= + 𝑝2 51 − =0
𝑚2 −6 4 𝛼2 − 10 2
𝛼2 𝑙22 𝑙22
15
[ ][ ] (A15)
1 1
− 700 𝛿𝑦2
+ 𝑝22 1400 (A4)
1
− 11 𝛼2
1400 6300
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

Ly Fig. A1. (a) Original and deformed configura-

tion of the V-beam represented by solid lines and
Δ dashed lines, respectively. Free body diagram of
(b) the left half; (c) the right half of the quarter
model of the V-beam.

Lx

l2

l1

Δ ΔY 2
(a) M2
ΔX 2 P2
α2
F2
y

Y1

θ1
X1

θ
θ
X2
Y2
F1

M1 ΔX 1
θ2

α1 M’1
ΔY 1
P1

(b) (c)
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

[ ( )]
sin(−2𝜃 + 𝛼1 )𝑝2 + cos(−2𝜃 + 𝛼1 ) 12𝛿𝑦2 − 6𝛼2 + 𝑝2 1.2𝛿2 − 0.1𝛼2 0.4

𝑙22 FEA
( ) 0.35 BCM
12𝛿𝑦1 − 6𝛼1 + 𝑝1 1.2𝛿𝑦1 − 0.1𝛼1 BCM (reduced order)
+ =0 (A16)
𝑙12 0.3

( ) ( )
2 2
−6𝛿𝑦1 + 4𝛼1 + 𝑝1 −0.1𝛿𝑦1 + 𝛼
15 1
6𝛿𝑦2 − 4𝛼2 − 𝑝2 −0.1𝛿𝑦2 + 𝛼
15 2
0.25
+

Force (N)
𝑙1 𝑙2
[ 0.2
[ ( )] ( )
1 11 1
− 12𝛿𝑦2 − 6𝛼2 + 𝑝2 1.2𝛿𝑦2 − 0.1𝛼2 1 + 𝛼2 𝑝 2 𝛼 − 𝛿
𝑙2 6300 2 1400 𝑦2 0.15
( ) ( )
1 1 1 1
− 𝛿𝑦 2 𝑝 2 𝛼 − 𝛿 − 𝛼2 𝛼 − 𝛿
1400 2 700 𝑦2 15 2 20 𝑦2 0.1
]
( ) 𝛿𝑦2 𝑝2
1 𝑤 𝑝2
2
1 3
+ 𝛿𝑦 2 𝛼2 − 𝛿𝑦2 + + =0 (A17) 0.05
20 5 12 𝑙2 𝑙2
2

[ 0
( ( ) ( ) 0 2 4 6 8 10 12
11 1 1 1
𝐿𝑥 − cos(𝜃) 𝑙1 𝛼1 𝑝1 𝛼 − 𝛿 − 𝛿𝑦 1 𝑝 1 𝛼 − 𝛿
6300 1 1400 𝑦1 1400 1 700 𝑦1 Displacement (mm)
) ]
( ) ( ) Fig. A2. F-Δ curve of the V-beam based on the BCM, reduced-order BCM and
1 𝑝1 𝑤
2
1 1 1 3
− 𝛼1 𝛼1 − 𝛿𝑦1 + 𝛿𝑦1 𝛼1 − 𝛿𝑦1 + + 𝑙1 finite element analyses.
15 20 20 5 12 𝑙2
1
[
( ( )
11 1
+ sin(𝜃)𝑙1 𝛿𝑦1 + cos(𝛼1 − 𝜃) 𝑙2 𝛼2 𝑝2 𝛼2 − 𝛿𝑦2
6300 1400 References
( ) ( )
1 1 1 1
− 𝛿𝑦 2 𝑝 2 𝛼1 − 𝛿𝑦2 − 𝛼1 𝛼2 − 𝛿𝑦2 [1] Bourke AK, Lyons GM. A threshold-based fall-detection algorithm using a bi-axial
1400 700 15 20 gyroscope sensor. Med Eng Phys 2008;30:84–90.
) ]
( ) [2] Huynh QT, Nguyen U, Irazabal LB, Ghassemian N, Tran BQ. Optimization of an ac-
1 𝑝2 𝑤
2
1 3
+ 𝛿𝑦 2 𝛼 − 𝛿 + + 𝑙2 − sin(𝛼1 − 𝜃)𝑙2 𝛿𝑦2 = 0 celerometer and gyroscope-based fall detection algorithm. J Sens 2015;2015 Article
20 2 5 𝑦2 12 𝑙2 ID 452078. doi:10.1155/2015/452078.
2
[3] Baltes J, McGrath S, Anderson J. Active balancing using gyroscopes for a small hu-
(A18) manoid robot. In: Proceedings of the 2nd International Conference on Autonomous
Robots and Agents December 13-15; 2004. p. 470–5.
[4] Wu G. Distinguishing fall activities from normal activities by velocity characteristics.
[
( ( ) ( ) J Biomech 2000;33:1497–500.
11 1 1 1
𝐿𝑦 − Δ− sin(𝜃) 𝑙1 𝛼1 𝑝1 𝛼1 − 𝛿𝑦1 − 𝛿𝑦1 𝑝1 𝛼1 − 𝛿𝑦1 [5] Chen Y, Aktakka EK, Woo J-K, Oldham KR. Modeling and calibration of a capac-
6300 1400 1400 700 itive threshold sensor for in situ calibration of MEMS gyroscope. In: Proceedings
) ] of 2016 IEEE international conference on advanced intelligent mechatronics (AIM)
( ) ( )
1 𝑤 𝑝1
2 July 12–15; 2016. p. 999–1004.
1 1 1 3
− 𝛼1 𝛼1 − 𝛿𝑦1 + 𝛿𝑦1 𝛼1 − 𝛿𝑦1 + + 1 + 𝑙1 [6] Hong SK, Park S. Minimal-drift heading measurement using a MEMS gyro for indoor
15 20 20 5 12 𝑙1 2
mobile robots. Sensors 2008;8:7287–99.
[
( ( ) [7] Hu Z, Gallacher BJ. Precision mode tuning towards a low angle drift MEMS rate
11 1 integrating gyroscope. Mechatronics 2018;56:306–17.
− cos(𝜃)𝑙1 𝛿𝑦1 + sin(𝛼1 − 𝜃) 𝑙2 𝛼2 𝑝2 𝛼2 − 𝛿𝑦2
6300 1400 [8] Howell LL. Compliant mechanisms. New York: John Wiley & Sons; 2002.
( ) ( ) [9] Todd B, Phillips M, Schultz SM, Hawkins AR, Jensen BD. RFID threshold accelerom-
1 1 1 1 eter. IEEE autotestcon 8-11 September 2008 Salt Lake City, UT; 2008.
− 𝛿𝑦 2 𝑝 2 𝛼2 − 𝛿𝑦2 − 𝛼2 𝛼2 − 𝛿𝑦2
1400 700 15 20 [10] Zhao J, Gao R, Yang Y, Huang Y, Hu PA. Bidirectional acceleration switch incor-
) ]
( ) 𝑤 2𝑝 ( )
porating magnetic fields based tristable mechanism. IEEE/ASME Trans Mechatron
1 3 1 2 2013;18:113–20.
+ 𝛿𝑦 2 𝛼 − 𝛿 + + 𝑙2 + cos 𝛼1 − 𝜃 𝑙2 𝛿𝑦2 = 0
20 2 5 𝑦2 12 𝑙2 [11] Tran HV, Ngo TH, Tran NDK, Dang TN, Dao T-P, Wang D-A. A threshold accelerom-
2 eter based on a tristable mechanism. Mechatronics 2018;53:39–55.
(A19) [12] Frangi A, Masi BD, Confalonieri F, Zerbini S. Threshold shock sensor based on a
bistable mechanism: Design, modeling, and measurements. J Microelectromech Syst
2015;24:2019–26.
𝛼1 + 𝛼2 = 0 (A20) [13] Nelson D, Ibrahim A, Towfighian S. Dynamics of a threshold shock sensor: Combin-
ing bi-stability and triboelectricity. Sens Actuators A 2019;285:666–75.
With a prescribed Δ, the forces f1 and p1 can be solved. By Eq. (A12) the [14] Chuang C-T, Chen R. Design, fabrication, and characterization of out-of-plane
external force F can be computed, then the stiffness of the V-beam spring W-form microsprings for vertical comb electrodes capacitive sensor. J Mi-
cro/Nanolithography, MEMS, MOEMS 2009;8:033021.
is taken as the initial slope of the F-Δ curve. [15] Zhou G, Dowd P. Tilted folded-beam suspension for extending the stable travel range
Fig. A2 shows the F-Δ curves of a quarter model of the V-beam spring of comb-drive actuators. J Micromech Microeng 2003;13:178–83.
with l1 = 40 mm, l2 = 80 mm, 𝑤1 = 𝑤2 =0.8 mm, 𝜃 = 20o , E = 2.6 GPa [16] Saggere L, Kota S, Crary SB. A new design for suspension systems of linear microac-
tiators. In: ASME proceedings of the international mechanical engineering congress
and a thickness of 5.5 mm based on the BCM, reduced-order BCM and fi- and exhibition, DSC 55-2; 1994. p. 671–5.
nite element analyses. The curve based on the reduced-order BCM agree [17] Chen G, Ma F. Kinetostatic modeling of fully compliant bistable mechanisms using
with the BCM and the finite element analyses. For the V-beam consid- Timoshenko beam constraint model. ASME J Mech Des 2015;137:022301.
[18] Ma F, Chen G. Modeling large planar deflections of flexible beams in com-
ered in this investigation, its force-displacement relation can be pre-
pliant mechanisms using chained beam-constraint-model. ASME J Mech Robot
dicted by the reduced-order BCM. The initial slope of the F-Δ curves 2016;8:021018.
based on the reduced-order BCM, BCM and finite element analyses is [19] Smith M. ABAQUS/Standard user’s manual, version 6.9. Providence, RI: Simulia;
2009.
0.02961 N/mm, 0.02963 N/mm and 0.02940 N/mm, respectively. The
[20] Li G, Chen G. A function for characterizing complete kinetostatic behaviors of com-
initial slope predicted by the reduced-order BCM is nearly identical to pliant bistable mechanisms. Mech Sci 2014;5:67–78.
that predicted by the BCM, and is only 0.7% higher than that com- [21] Chen G, Bai R. Modeling large spatial deflections of slender bisymmetric beams in
puted by finite element analyses. Therefore, the stiffness of the V-beam compliant mechanisms using chained spatial-beam-constraint-model. ASME J Mech
Robot 2016;8:041011.
spring can be taken as 0.1184 N/mm, which is four times the value of [22] Qiu J, Lang JH, Slocum AHFellow, IEEE. A curved-beam bistable mechanism. J Mi-
0.02961 N/mm, the stiffness of the quarter model of the V-beam spring. croelectromech Syst 2004;13:137–46.
H.V. Tran, T.-H. Ngo and P.-L. Chang et al. Mechatronics 63 (2019) 102280

[23] Brake MR, Baker MS, Moore NW, Crowson DA, Mitchell JA, Houston JE. Modeling Pei-Lun Chang received the M.S. degree in the Graduate Insti-
and measurement of a bistable beam in a microelectromechanical system. J Micro- tute of Precision Engineering, National Chung Hsing Univer-
electromech Syst 2010;19:1503–14. sity, Taiwan, ROC, in 2013. He is currently working towards
[24] Andò B, Baglio S, L’Episcopo G, Trigona C. Investigation on mechanically bistable the Ph.D. degree in the Graduate Institute of Precision Engi-
MEMS devices for energy harvesting from vibrations. J Microelectromech Syst neering, National Chung Hsing University, Taiwan, ROC. His
2012;21:779–90. research interests are actuators, and compliant mechanisms.
[25] Xu Y, Nashif A. Measurement, analysis and modeling of the dynamic properties of
materials. Sound Vib 1996;30:20–3.
[26] Wojtowicki JL, Jaouen L, Panneton R. New approach for the measurement of damp-
ing properties of materials using the Oberst beam. Rev Sci Instrum 2004;75:2569–74.
[27] Crespo JE, Juliá E, Parres F, Segura J, Gadea JM, Nadal AV. Investigation of damping
properties using products coming from ELT (end-of-life-tires). Ann Oradea Univ. Fasc
Manag Technol Eng 2010;9:21–4.
[28] Pérez-Peña A, García-Granada AA, Menacho J, Molins JJ, Reyes G. A methodology I-Ting Chi received the M.S. degree in Graduate Institute of
for damping measurement of engineering materials: application to a structure under Precision Engineering from the National Chung Hsing Univer-
bending and torsion loading. J Vib Control 2014;22:2471–81. sity, Taiwan, ROC, in 2017. He is currently working towards
[29] Catania G, Sorrentino S. Experimental evaluation of the damping properties of beams the Ph.D. degree in the Graduate Institute of Precision Engi-
and thin-walled structures made of polymeric materials. In: Proceedings of the 27th neering, National Chung Hsing University, Taiwan, ROC. His
conference and exposition on structural dynamics (imac-XXVII) February 9-12, Or- research interests are microelectromechanical systems, and
lando, Florida USA; 2009. compliant mechanisms.
[30] Tse FS, Morse IE, Hinkle RT. Mechanical vibrations: theory and applications. Boston,
United States: Allyn and Bacon, Inc.; 1979.
[31] Muravskii GB. On frequency independent damping. J Sound Vib 2004;274:653–68.
[32] Acar C, Shkel A. MEMS vibratory gyroscopes: structural approaches to improve ro-
bustness. New York, USA: Springer Scuebce+Business Media, LLC; 2009.
[33] Ting JM, Crawley EF. Characterization of damping of materials and structures from
nanostrain levels to one thousand microstrain. AIAA J 1992;30:1856–63. Ngoc Dang Khoa Tran received the Ph.D. degree in the Grad-
[34] Hwang RS, Fox CHJ, McWilliam S. The in-plane vibration of thin rings within-plane uate Institute of Precision Engineering, National Chung Hsing
profile variations —part I: general background and theoretical formulation. J Sound University, Taiwan, ROC, in 2018. He is currently a Lecturer
and Vib 1999;220:497–516. in the Faculty of Mechanical Engineering, Industrial Univer-
[35] Thomson WT. Theory of vibration with applications. Prentice-Hall International, sity of Ho Chi Minh City, Vietnam. His research interests are
Inc.; 1988. p. 92. sensors and actuators, and compliant mechanisms.
[36] InvenSense. IDG-300 datasheet, Sunnyvale, California: InvenSense, Inc; 2008.
United States of America.

Hong Van Tran received the M.S. degree in electrical engi-

neering from the Ho Chi Minh City University of Transport,
Vietnam, in 2011. He is currently working towards the Ph.D.
degree in the Graduate Institute of Precision Engineering, Na-
tional Chung Hsing University, Taiwan, ROC. His research in-
terests are sensors and actuators, and compliant mechanisms. Dung-An Wang received the Ph.D. degree in mechanical en-
gineering from the University of Michigan at Ann Arbor, in
2004. He is currently a Professor in the Graduate Institute
of Precision Engineering, National Chung Hsing University,
Taiwan, ROC. His research interests include sensors and ac-
tuators, energy harvesters, ultrasonic welding and compliant
mechanisms.

Tien Hoang Ngo received the M.S. degree in the Graduate In-
stitute of Precision Engineering, National Chung Hsing Univer-
sity, Taiwan, ROC, in 2017. He is currently working towards
the Ph.D. degree in the Graduate Institute of Precision Engi-
neering, National Chung Hsing University, Taiwan, ROC. His
research interests are energy harvesters, and compliant mech-
anisms.