Improving the gravity-rotation-excited vibration energy

harvesting in offset configurations

Yilong Wang1,2,*, Yang Zhao1, Yishen Tian1, Dengqing Cao1, Zhengbao Yang2,3,*

1 School of Astronautics, Harbin Institute of Technology, Harbin 150001, China

2 Department of Mechanical Engineering, City University of Hong Kong, Hong Kong
3 Shenzhen Research Institute of City University of Hong Kong, Shenzhen 518057, China

*Correspondence to: Yilong Wang ([email protected]); Zhengbao Yang ([email protected]).

Abstract

The self-powered sensing technology based on gravity-rotation-excited vibration energy harvesting (GRE-
VEH) enables information acquisition directly from rotatory machines online. However, the centrifugal
effect due to the distance between the mass centroid of the energy harvester and the rotating axis (i.e., offset
distance) seriously limits its broad engineering applications. To address this issue, we propose methods of
using a tensile force provided by pre-deformed springs to counter the constant centrifugal force around the
target rotation speed for improving the performance of the energy harvester affected by offset
configurations. The GRE-VEH system with a linear spring is first studied by experiments and simulations,
where a distributed-parameter model is developed and experimentally validated. The results well validate
the proposed mechanism, where the performance of the harvester prototype significantly increases (by up
to 1225%). A parametric study is also performed theoretically. The results indicate that the best performance
improvement can be obtained by tuning the parameters related to the tensile length, the stiffness, and the
damping of the spring. To enable the proposed mechanism in large centrifugal force cases, a pair of negative
springs is introduced to the system for constructing a quasi-zero stiffness (QZS) structure together with the
linear spring. The new GRE-VEH system is theoretically studied by an updated model, where the system’s
main characteristics are revealed. The simulation results well indicate the feasibility of the method using
the QZS mechanism. This work paves a new way of solving the offset distance problem for the application
of the GRE-VEH and also provides new insight into the application of the QZS mechanism.
Keywords: Energy Harvesting; Rotation; Quasi-Zero Stiffness; Piezoelectric; Nonlinear

Manuscript Page 1 of 35
Highlights

1. The RMS voltage response of the energy harvester is increased by up to 1225%;

2. We present a new quasi-zero stiffness mechanism for energy harvesters;
3. We derive a stable solution to the adverse effect of the offset distance;
4. The nonlinear dynamics of the proposed systems are analyzed comprehensively.

1 Introduction

Rotary machines are widely used in industrial applications, from large machines (such as steam, gas and
wind turbines, and turbine engines) to medium- and small-sized devices (such as wheels, transmission
shafts, and gears). Their rotors typically experience complex dynamic loads resulting from mass imbalance
[1], misalignment [2], shaft bending [3], thermal deformation [4], cracks [5], rubbing [4, 6, 7], flow [8],
etc., and thus are highly vulnerable to failures during work [9]. Even if a slight fault is caused, the
subsequent chain reaction may rapidly lead to a catastrophic failure of the machine. Therefore, if the fault
is detected online, the danger can be dealt with in time. However, the incumbent fault detection techniques
mainly serve offline for providing information that is used to initiate maintenance actions, rather than real-
time decision-making [10]. This is largely due to the physical inaccessibility between the rotor and the
stator, where information exchange can be only achieved by some special means, such as slip rings and
wireless communication. These means can be of some use for large rotary machines, but fail to fit into the
development trend of modern devices towards high performance, high efficiency, miniaturization, and
lightweight. As a result, the online acquisition of information from the rotor by the special means is severely
limited. Although the battery-based wireless communication technology fits well with the pursuit of small
size and lightweight, it fails to be applied for long-term use due to battery issues [11]. Currently, the main
technical difficulty is to measure the operating status of rotor components directly, in long term, and online.
To achieve the aforementioned goals, gravity-rotation-excited vibration energy harvesting (GRE-VEH)
that is a process to harness mechanical energy in rotational [12-16] or swing [17-21] environments and
convert it into electrical energy was proposed to integrate with wireless sensors to construct a monitoring
system architecture for online directly detecting the status of rotor components in long-term [10]. This kind
of method utilizes the periodically changed component of the gravitational force to excite the energy
harvester (EH) installed on the rotor and achieves power generation from the harvester’s oscillations via
electromechanical coupling processes of piezoelectric [22-26], triboelectric [27-29], and electromagnetic

Manuscript Page 2 of 35
effects [19, 30], etc. In recent years, GRE-VEH has been increasingly developed for different application
contexts and its derived self-powered wireless electronics have shown high feasibility. Roundy and Tola
[12] presented an EH in which the ball is oscillated by gravity in rotations and then periodically impacts
piezoelectric cantilever beams for power generation. Their prototype was capable of directly powering an
RF transmission in a vehicle wheel every 60 seconds or less over a speed range of 10 to 155 kph. Zhao et
al. [22] put forward a novel maximal utilization strategy for piezoelectric-electromagnetic-triboelectric
rotational energy harvesting to achieve the self-powered rotor condition monitoring system. Their prototype
could operate effectively in a wide speed range (0–1000 rpm) and charge a 100 μF capacitor to 5 V within
11 seconds at 500 rpm. Mei et al. [31] proposed to develop self-powered sensing for monitoring the wind
turbine’s blade and studied the piezoelectric cantilever beam mounted on the wind turbine’s blade
respectively in forward and inverse configurations. Their prototype was tested to charge a 10 μF capacitor
to 10 V within 5.2 seconds at 120 rpm. We also previously studied the GRE-VEH and achieved a direct
online monitoring method for jet engine rotors in long-term use [10]. Our method showed a high-efficient
wireless communication (every 3 seconds at 1000 rpm or more frequent at higher speeds) between the
wireless sensor in the rotor and the data receiver on the ground. In addition to the representative studies
above, there are extensive results [32] well showing promising industrial applications for Internet of Things
(IoTs) based on the GRE-VEH but some problems still need further investigations before practical
application, such as those about centrifugal effect, reliability, and stability.
In our previous works [10, 33], we demonstrated the adverse effect of the offset configuration and its
induced centrifugal force on the PEH. Generally, to avoid this effect, three effective methods were proposed.
The most effective method is to set the mass centroid of the EH exactly on the rotating center [14, 34] and
thus the effect of the centrifugal force is reduced the most. If the harvester is symmetric about the rotating
center, it is better [10]. However, this method cannot be applied in situations where the rotating center is
spatially unavailable (e.g., vehicle tires). As a result, the second method that makes the cantilever
piezoelectric beam mounted in the forward or inverse radial direction was proposed. Since the centrifugal
force increases along with the rotation speed, thus based on the second method not only the adverse effect
on the harvester’s vibration is alleviated largely but also its working bandwidth can be broadened. Gu et al.
[35], Hsu et al. [36], Mei et al. [37-40], Zhao et al. [23, 41], Fang et al. [42-45], and Zou et al. [46] carried
out a series of investigations on the second method and comprehensively revealed the dynamical
characteristics of their proposed systems. Their methods as solutions to the centrifugal effect and its derived
problems showed high potential but fail to well fit into some high-efficient EHs with the doubly clamped
Manuscript Page 3 of 35
boundary [47, 48]. To overcome this, we proposed the third method that uses the magnetic repulsion to
counter the constant centrifugal force and makes the adverse effect of the offset distance alleviated [33].
Although the result showed the mechanism was feasible, we found its engineering application still faces
some challenges, especially in large centrifugal force cases, such as instability, poor adjustability, low
scalability, and potential of bending-torsional coupling and high complexity. This is mainly because the
repulsive force tends to cause the vibrator off the torsional equilibrium position. It is not only an unstable
force for the EH’s bending vibration but also increases the risk of the bending-torsional coupling vibration
problem [49]. Considering that high-efficient EHs are highly necessary for realizing a high-power
generation that is one of the most important obstacles for achieving self-powered IoTs, it is imperative to
seek for a new solution to the aforementioned issue.
In this paper, we propose to replace the repulsive force with a tensile force to meet the requirement of
a stable force for countering the constant centrifugal force due to the offset distance. The tensile force is
first achieved by a pre-extended linear spring. Based on the EH system with a linear spring, theoretical and
experimental studies are carried out to verify the proposed solution and reveal the effect of key parameters
on the EH’s electrical performance, where a distributed-parameter model of the system is developed and
validated by experimental data. Then, the previous and the current methods are comprehensively discussed
and their common issue is also indicated. To address the issue, we introduce the quasi-zero stiffness (QZS)
mechanism. A theoretical model of the EH system with a QZS structure based on the previously validated
model is developed to verify the feasibility of the method and also study the dynamical characteristic of the
system.

2 Method of using a linear spring

To verify the proposed solution of using a tensile force to the offset distance issue, the first step is to design
a proper configuration of the system and its working principle, which is presented in subsection 2.1. Then,
a reliable theoretical model of the system also needs to be developed and experimentally validated, which
are respectively presented in subsections 2.1 and 2.2. On this basis, the effects of the key parameters on the
performance of the solution can be studied to provide insight into the system design, which is carried out
in subsection 2.3.

2.1 System design and working principle

In this study, the energy harvester adopts the construction of the high-efficiency compressive-mode
Manuscript Page 4 of 35
piezoelectric energy harvester (HC-PEH) in Refs. [49, 50] but with a shape difference in mass blocks (see
Figure 1). It is installed with an offset distance (r) parallelly on the surface of a rotating disk (angular
velocity: ) whose radial direction is perpendicular to the support beams of the harvester and also passes
through its symmetry center [10, 33]. The working principle of the rotational HC-PEH system can be
elucidated in two steps: 1) the rotation induces an excitation with an acceleration amplitude of 1 g (9.81
m/s2) due to gravity; 2) the vibration of the HC-PEH causes periodic compressions of the PZT plate by the
bow-shaped beams due to the axial deformation of the support beams, consequently resulting in electric
output by the piezoelectric effect. According to our previous result [10], this rotational energy harvesting
method is very promising in achieving high performance if there exists only a small offset distance.
However, the method rapidly fails as the offset distance (r) increases, because the stiffening effect of the
constant centrifugal force on the harvester simultaneously becomes ever larger, increasingly suppressing
the vibration of the harvester [33].

Figure 1 Harvesting energy in rotating machinery using the high-efficiency compressive-mode piezoelectric energy
harvester (HC-PEH). When rotating, the harvester experiences a time-varying force due to gravity, which is along the direction
where the harvester is prone to bending deformations, and thus causes bending vibrations of the harvester. Then, by doing work
on the piezoelectric unit in the center, the harvester inverts the mechanical energy into electricity. High power output (max. to
78.87 mW), broad working bandwidth (22.5 Hz), and strong reliability (2100 rpm) make the proposed method a promising
sustainable power supply for sensors in rotational systems of low to medium rotation speeds, such as turbines, vehicle tires,
wheels of the high-speed train, and gears [10].
Manuscript Page 5 of 35
 In practical engineering, rotational applications that only provide spaces with offset distances are
common, such as blades of turbines and vehicle tires, as shown in Figure 1. It is of great meaning to make
the GRE-VEH method [10] better fit into the offset configuration scenario, referring to the previously
validated method of using a preloaded constant force to counter the constant part of the centrifugal force
[33]. Consequently, we propose to use a linear spring and its provided tensile force to achieve the restoring
of the harvester’s performance at and near the target rotation speed target (see Figure 2). In the initial state
(=0), as shown in Figure 2a, the linear spring that has a constant stiffness (k) and a constant damping
coefficient (c), is tense with another end fixed on the top of the harvester’s central structure. In this case,
the initial extension of the spring is lw0 and the initial vertical (along the Y-axis) deformation of the harvester
in the center is w0. Then, as  increases, the static equilibrium point of the harvester relative to the host
moves radially outwards due to the centrifugal effect and finally back to the place where the harvester has
no initial deformation when =target (see Figure 2b). At this time, the dynamical characteristic of the
harvester at the relative static equilibrium point is free from the stiffening effect of the constant centrifugal
force, making that the adverse effect on the harvester’s performance is alleviated the most when the

Figure 2 Schematic of the proposed method to alleviate the adverse effect of the offset distance. (a) When the rotating host
is stationary. Assuming that there are no gravity and rotations, the system is thus in static equilibrium where the linear spring is
simply stretched by the elastic recovery force of the HC-PEH. Once the host rotates, the static equilibrium point relative to the
host changes and is determined by the elastic recovery force of the harvester, the centrifugal force, and the linear spring's tensile
force. On this basis, once there is gravity, the harvester vibrates around the relative static equilibrium point. (b) When the host
rotates at the target rotation speed, corresponding to the relative static equilibrium point of w=0. w denotes the deformation of
the harvester along the Y-axis at x=L, i.e., w(L,t). In this case, the vibration of the harvester is affected the least once there are
gravity and rotations.
Manuscript Page 6 of 35
harvester vibrates. As a result, the harvester still keeps a good performance in the case of the offset
configuration theoretically according to Ref. [33]

2.2 Modeling of the rotational HC-PEH system with a linear spring

A good theoretical model with distributed parameters can provide physical insight and in-depth guidance
in understanding a dynamic system. Thus, to comprehensively investigate the proposed system, a
distributed-parameter analytical model according to the schematic in Figure 2 is developed in this section.
Considering the modeling approach proposed in Ref. [50] was verified to be reliable and that the HC-
PEH shown in Figure 2 is essentially the same as its counterpart in Ref. [50], this paper directly adopts the
results of Ref. [50] to develop the analytical model of the proposed system. Since the mass of the spring is
trivial to the HC-PEH, its effect is neglected in this paper. Together with the results on the system’s
mechanical energy in Ref. [50], the total potential and the kinetic energies of this system can be simplified
as
1
k  lt  w  L, t   ,
2
U  U HC-PEH  U ls  U HC-PEH  (1)
2
T  THC-PEH , (2)

where UHC-PEH and THC-PEH are respectively the total elastic potential energy and the kinetic energy of the
rotational HC-PEH system in Ref. [50] (see Appendix A). lt is the extension of the spring when the vibration
displacement (w) of the harvester at x=L along the Y-axis (pre-extension) is zero. Similarly, the variation of
the system’s external virtual work can be simplified as
1
 W   WHC-PEH  cw  L, t   w  L, t  , (3)
2
where WHC-PEH is the variation of the rotational HC-PEH system’s external virtual work in Ref. [50].
According to the Galerkin discretization [50], the y-axis motion of the support beam (see Figure 2b)
can be expanded as a summation of trial functions:

w  x, t    n  x  qn  t , (4)
n 1

where n (x) are linear vibration mode shapes of the support beam and qn (t) are generalized coordinates.

For vibration-based energy harvesters, since the fundamental mode generally contributes the most to the
vibration, their fundamental resonant frequencies are generally tuned to be around the excitation frequency
to exploit the vibration energy the most. In addition, it was experimentally validated in Refs. [49, 50] that

Manuscript Page 7 of 35
the fundamental mode is sufficient for obtaining an approximate response of the overall displacements in
the support beams of the HC-PEH. Thus, in this paper, only the first bending mode is considered.

By Lagrange’s equation, choosing q (i.e., q1 ) and V (the voltage generated by the PZT plate) as the

generalized coordinates, we can derive the dynamical equation by

d  La  La  W
   ,
dt  q  q  q (5)
d  La  La  W
  ,
dt  V  V  V

where La=T-U. Since there are no q and V components in U, nor q and V components in T, Eq. (5) can be

rewritten as

d  T  U  W
   ,
dt  q  q  q
(6)
d  T  U  W
   ,
dt  V  V  V

d  T  d  THC-PEH 
where     =M q q, (7)
dt  q  dt  q 

U UHC-PEH Uls
    K  2Mq  q  Kn q3  Vq  Fc  k2  L q  k  L lt , (8)
q q q

W 1
 Cq  Cnq2q  FG  KP q  c2  L q, (9)
q 2 0

U U HC-PEH U ls 1
   q 2  C pV  Q0 , (10)
V V V 2
  T    THC-PEH 
 =    0, (11)
t  V  t  V 
 W  WHC-PEH
=  Q. (12)
V V
The expressions of Mq, C, Cn, K, KP0 , Kn, FG, Fc, , Cp, and Q0 are given in Appendix B. Moreover, Q is

the electric charge output of the PZT plate, and the time rate change of Q is the electric current passing

through the resistive load, i.e. Q  V R . In addition, considering the stiffness softening of the connection

by the epoxy glue between the elastic beam and the bow-shaped beam, a coefficient () is introduced in the
cubic nonlinear term. Then, with Eqs. (6)-(12), one can obtain the governing equations of the system:

 
M q q   C  Cl  q  Cn q 2 q  K   2 M q  K P0  K l q   K n q3  Vq  Fc  FG  Fl , (13)
Manuscript Page 8 of 35
 1
 q 2  C pV  Q0  Q , (14)
2
1
where Cl  c 2  L  , K l  k  2  L  , Fl  k   L  lt . When the energy harvester connects to an
2
external resistance (R), deriving both sides of Eq. (14) with respect to time (t) gives
V
 qq  C pV   . (15)
R
Herein, we take the same assumption as in Ref. [49] that the support beam is a cantilever beam and
the proof mass at its free end is elastically constrained in the axial direction (x-axis), as shown in Figure
D1. In this case, the fundamental bending modal function of the system is[51, 52]

1   
  x  1  cos  x   . (16)
2  L 

Then, with Equations (13) and (15), the response of the system can be approximately simulated after the
values of each parameter are correctly taken.

2.3 Experimental verification and analysis

To verify the proposed method and the developed model, we carried out extensive experiments during
studying. First, to ensure that the parameters are taken more reliably, we fabricated a HC-PEH prototype
(see Appendix C and Table D1 for its details) based on the design shown in Figure 2 and tested the prototype
without the linear spring in self-designed rotational testing rigs (see Figure 3), respectively when r = 0, 5,
10 and 15 mm. Then, based on the obtained experimental data, values of each parameter required by the
theoretical models were determined via direct measurements and fitting calculations by setting k, c, and lt
zero, which are assumed to be fixed in this study unless otherwise stated, as listed in Table 1 (see
Nomenclature for descriptions of each parameter). For better comparisons, the RMS voltage-rotation speed
responses respectively from experiments and simulations are shown in Figure 3b. The results well reveal
that the electrical performance of the harvester deteriorates with the increase of the offset distance, i.e., r,
as validated in Ref. [33], and the simulation responses closely render the experimental data. Thus, it is
believed that the taken values in Table 1 are reliable.
Based on the test rigs and the system shown in Figure 2, we designed and installed a device with a
linear spring on the rotating host (see Figure 4), whose extension is in the same direction as the offset
distance. It is noteworthy that the linear spring in this design essentially works the same as the theoretical
counterpart, although it is realized in a slightly different way. Actually, the way of its implementation largely
depends on the characteristics of the practical application, including the space, offset distance, the
Manuscript Page 9 of 35
dynamical characteristic of the harvester, etc., and thus needs to be adapted case by case. Since the
experiment was mainly carried out to validate the feasibility of the proposed method that uses the tensile
force (see Figure 2) and the experimental setup (see Figure 4) indeed reacts to the essence, it is believed
that the result obtained thereafter is sufficiently reliable provided the experiment is correctly conducted.

Figure 3 Experiment setup and results. (a) Self-designed rotational testing rigs. The HC-PEH prototype was installed on a
support structure that was mounted on a disk and had a vibration direction vertical to the rotating direction of the disk. The disk
was installed on the rotating shaft that was supported by two bearings on the platform. Between the disk and the left bearing,
there was a slip ring installed on the shaft, which rotated with the shaft or disk and connected the PEH to the oscilloscope. (b)
Comparison of the results respectively from the experiment and the corresponding simulation on the RMS voltage-rotation speed
response. The simulation result well matches the experimental data, indicating that the developed model has good reliability.

Table 1 Parameters of the theoretical model used in the simulation

Description Value Description Value
Length of the support beam (L) 74 mm Waist length of the bow-shaped beam (LB) 16 mm
Thickness of the support beam (h) 0.25 mm Torsional stiffness of the bow-shaped beam (k) 10.13 Nm/rad
Width of the support beam (b) 4.5 mm Young’s modulus of the piezo plate (Ep) 63 GPa
Density of the support beam () 7850 kg/m3 Density of the piezo plate 7800 kg/m3
Young’s modulus of the support Size of the piezo plate 40×15×1 mm3
200 GPa
beam (E) Softening coefficient of the nonlinear stiffness 0.37
Equivalent proof mass (M) 26.925 gram Length of the PZT plate's compressive area (Hy) 36 mm
Mass of the PZT plate (Mp) 4.7 gram Coupling coefficient of the PZT plate (d31) -250×10-12 CN-1
Initial angle () 9 degree Coupling coefficient of the PZT plate (d33) 525×10-12 CN-1
Preloaded axial force (P0) 0 Electrical permittivity of the piezo plate (33) 1.1×10-8 FN-1
Offset distance (r) 0.5 mm External resistance (R) 105 k
Gravitational acceleration 9.81 m/s2 Damping coefficient of the bow-shaped beam (cS) 91.16 N/(m/s)
Deformation transmission
0.95 Damping coefficient of the support beam (cB) 0.014 N/(rad/s)
efficiency ()

In the experiment of verifying the proposed method, the offset distance (r), the pre-extension (lt) of
the spring when w=0, and the stiffness (k) of the linear spring were respectively chosen as variables. To
Manuscript Page 10 of 35
ensure reliable results, the experiments were extensively performed at each rotation speed. Moreover, based
on the experimental setup (see Figure 4), the corresponding simulations were also carried out (see Table
D2 for parameter values of the linear spring). Herein, for better comparisons, both the experiment and the
simulation results on the RMS voltage-rotation speed response with a resistor of 100 k are presented
together as respectively shown in Figures 5 to 7, where r is 5 to 15 mm.

Figure 4 Experimental setup for equivalently verifying the method of using a linear spring to provide the tensile force.
To ensure the rotating disk is as dynamically balanced as possible, a fitting counterweight was installed. As a result, the vibration
of the harvester was hardly affected by the vibration of the disk. Linear springs with stiffnesses of 16.95 and 37.75 N/m were
used, respectively.

Figure 5 RMS voltage vs rotation speed curves of the system when r is around 5 mm. (a) k=16.95 N/m. (b) k=37.75 N/m.
By comparing with the results of no-spring cases (i.e., the initial systems), it is found that the proposed method achieved a
performance improvement of up to 1225% when k=37.75 N/m and was verified to have good feasibility.

By analyzing the three figures, it is observed that 1) the simulation results well render the experimental
data, thus indicating that the developed model is valid; 2) the electrical performance of the harvester

Manuscript Page 11 of 35
Figure 6 RMS voltage vs rotation speed curves of the system when r is around 10 mm. (a) k=16.95 N/m. (b) k=37.75 N/m.
By comparing with the results of no-spring cases (i.e., the initial systems), it is found that the proposed method achieved a
performance improvement of up to 966% when k=37.75 N/m.

Figure 7 RMS voltage vs rotation speed curves when r is 15 mm. (a) k=16.95 N/m. (b) k=37.75 N/m. By comparing with the
results of no-spring cases (i.e., the initial systems), it is found that the proposed method achieved a performance improvement
of up to 1177% when k=16.95 N/m.

increases along the pre-extension (lt) and the stiffness (k), meaning that the proposed method of using the
linear spring and the tensile force indeed works well; 3) the rotation speed corresponding to the peak voltage
increases along with the pre-extension (lt) and the stiffness (k) but reduces with the increase of the offset
distance. This is because the constant centrifugal force is completely countered only when the rotation
speed reaches a certain value, i.e., the target rotation speed (target), at which the performance of the harvester
restores the best. To better understand this, we can learn from Eq. (13) that there is no effect of the constant
centrifugal force when Fl = -Fc, i.e.,

Manuscript Page 12 of 35
  L  1  
2 t2arg et r    A  x  dx   M  M P    L     k   L  lt， (17)
 0
 2  

where A is the cross-sectional area of the support beam. M and MP are annotated in Table 1. Eq. (17) well
reveals that an increase in both k and lt leads to an increase in target, while an increase in r has the opposite
effect. To sum up, the comparison of the results in Figures 5 to 7 reveals that the proposed method of using
the linear spring is feasible and the developed model of the system is indeed reliable and can be used for
further theoretical analysis of the system.

2.4 Parametric study

For the proposed method, there are only three parameters that can be used for tuning the performance of
the harvester, i.e., the pre-extension (lt), the stiffness (k), and the damping (c) of the spring. Although Figures
5 to 7 clearly show the effects of lt and k on the system’s RMS voltage-rotation speed response, whether the
increases of lt and k always contribute to the performance improvement still needs further investigation.
Additionally, it is also unknown how the damping (c) affects the system. To clarify them, we took lt, k, and
c as variables respectively, and meanwhile kept the non-variable as constant to perform a parametric study
based on the developed theoretical model. Note that, when the pre-extension (lt), the stiffness (k), or the
damping (c) of the spring was non-variable, they were respectively taken as 70 mm, 37.7545 N/m, and 1
Ns/m (refer to Table 1). Via extensive simulations, the results on the RMS voltage vs rotation speed curves
were obtained, as shown in Figure 8.
It is observed from Figure 8a and 8b that the improvement by increasing the pre-extension (lt) or the
stiffness (k) has a maximum, where the rotation speed corresponding to the peak voltage is around 870 rpm
instead of 1000 rpm (Figure 3d). There are two main causes for the result. First, once the harvester is
determined, the voltage-rotation speed curve of the system without effects of the offset distance (r=0) and
the linear spring (k=0) is also determined (see Figure 3d) and the best when working in rotational
environments. Herein, the peak voltage and corresponding rotation speed (jump-down point) in that
voltage-rotation speed curve are respectively denoted by VRMS|r=0 and JD|r=0. The introduction of the linear
spring to the system causes the equivalent linear stiffness of the system to increase, as shown in Eq. (13),
but has few effects on the cubic nonlinear stiffness that is mainly caused by the doubly-clamped boundary
and only determined by the axial stiffness of the support beam. This means that the ratio of the nonlinear
stiffness vs the linear stiffness decreases and thus leads to a weakened nonlinearity of the system,
meanwhile causing a weaker hardening resonance that includes a reduced peak voltage (denoted by VRRMS|r=0,
Manuscript Page 13 of 35
VRRMS|r=0 < VRMS|r=0) and corresponding rotation speed (denoted by RJD|r=0, RJD|r=0<JD|r=0). Accordingly, the
peak voltages shown by the green curve in Figure 8a and the yellow curve in Figure 8b are near 870 rpm
and also reduce significantly compared to that in Figure 3d. Second, no matter what the stiffness of the
linear spring is, the proposed method only works for alleviating the adverse effect of the offset distance,
which does not improve the voltage-rotation speed curve of the harvester without any adverse effects (i.e.,
r=0 and k=0). This means that the rotation speed (RJD|r=0) is unchanged provided the stiffness of the linear
spring (k) is determined and the resonance of the system only happens when the rotation speed of the host
() is smaller than RJD|r=0. As learned from Eq. (17), target increases along with lt and k. Once lt or k
increases to a value that makes target exceed RJD|r=0, although the adverse effect of the offset distance is
completely eliminated when target, the harvester fails to resonate at such a rotation speed since  is
larger than RJD|r=0. Accordingly, the performance of the system drops significantly.

Figure 8 Simulation results on the RMS voltage vs rotation speed curve when a single parameter changes. (a) when k is
the variable. (b) when lt is the variable. (b) when c is the variable. The results predict further performance improvement of the
system in a larger offset distance (referring to Figure 7b) by tuning parameters. However, for this improvement, there is an
optimal value for each of lt and k, while c is as small as possible.

Figure 8c shows the variation of the RMS voltage vs rotation speed curves when the damping (c) of
the linear spring increases from 0.25 to 1.75 Ns/m. It reveals that the adverse effect of increasing c on the
system, which reduces the peak voltage and the corresponding rotation speed, is nonlinearly enhanced.
When c is small, the effect is small too. However, when it increases to a certain value (e.g., 1 Ns/m), the
Manuscript Page 14 of 35
effect rapidly becomes larger as c keeps increasing.
In summary, the simulation results further show the potential of the proposed method and reveal that
the system can attain the best electrical performance by optimizing the pre-extension (lt) and the stiffness
(k), but also indicates the problems from the introduction of the linear spring, i.e., the added linear stiffness
and damping. Thus, when applying the proposed method, it is better to use low-damping springs or elastic
structures to provide the preloaded tensile force and also to reduce their stiffnesses as much as possible.

3 Discussion on the methods of using the magnet and the linear spring

In the reference [33], the method of using a preloaded force provided by magnetic repulsions to counter the
constant centrifugal force for the energy harvester experiencing the adverse effect of the offset distance was
proposed and verified to be effective for the first time (see Figure 9a and 9b). Its essential principle
conforms to logic but the example approach of adopting magnetic repulsion has flaws that make the method
only work well in small constant centrifugal force cases, such as low rotation speeds and small offset
distances.

Figure 9 The method of using magnetic repulsion to reduce the adverse effect of the offset distance. (a) Working schematic
in the initial state. (b) Working schematic in the working state. In this case, the harvester is subjected to a significant negative
stiffness torque (TNS), especially when the constant centrifugal force needed to counter is large. (c) Relationship among the
magnetic spacing (d0), the counter force provided by the magnetic repulsion, i.e., km/( d0+w)3, the initial deformation of the
harvester (w0), and the elastic force, i.e., kw+knw3, as d0 increases. Note that all symbols here keep the same as in Ref. [33]. (d)
Relationship among d0, km/( d0+w)3, w0, and kw+knw3. It is clear that w0 increases along with either d0 or km, meaning that the risk
of the harvester’s damages rises as the constant centrifugal force needed to counter increases.
Manuscript Page 15 of 35
 The first flaw is that the magnetic repulsion causes an instability problem for the system. This is
because the magnetic repulsion leads to a negative torsional stiffness along the direction of the support
beam. Provided the repulsive force is sufficiently large, the equilibrium point of the system’s torsional
vibration becomes unstable (see Figure 9b). Actually, for the HC-PEH, the positive torsional stiffness is
provided by the support beam, which is often small and is easily reduced to minus by the magnetic repulsion
in most cases. As a result, when the system has to face a large constant centrifugal force, the magnetic
repulsive force usually causes the torsion of the harvester and then the method of providing a preloaded
force fails to work according to its principle. Moreover, once the harvester vibrates in the bending mode,
the magnetic force even causes the bending-torsional coupling vibration. In this case, the performance of
the harvester may significantly reduce due to the torsion-induced jump-down phenomenon [49], and also
the harvester is more vulnerable to failures due to torsions.
In addition to the instability, the magnetic force also faces another problem when used in large constant
centrifugal force cases. The magnitude of the magnetic repulsion is inversely proportional to the cube of
the magnetic spacing (d0), but it cannot increase infinitely as the magnetic spacing reduces to zero but only
to a certain value depending on the total magnetic flux of the two magnets. As a result, to fit into the large
constant centrifugal force, it is the most effective to increase the magnetic flux of the magnet or/and reduce
the magnetic spacing. However, this strategy requires a high load-carrying capacity of the harvester,
otherwise the harvester in its initial state (see Figure 9a) is prone to plastic deformations or failures in large
constant centrifugal forces cases (see Figure 9c and 9d), and thus has very limited applicability. If taking
the strategy of increasing the magnetic flux and meanwhile decreasing the magnetic spacing, it poses a
huge challenge to finding the matching permanent magnets. Furthermore, the increase of the magnetic flux
stiffens the system, which may largely cause the performance reduction of the harvester according to Ref.
[33].
To solve the aforementioned issue, other measures are needed, such as installing a guide rail assembly
[34] and a stopper [14] to limit the initial deformation of the harvester (see Figure D2), but these in turn
cause the system further complicated and even new problems. In comparison, the method of using the linear
spring provides the tensile preloaded force that is free from instability and torsion problems. Moreover, it
is more flexible in practice than the method of using magnets in terms of parameter tuning and the load-
carrying capacity of the harvester since the relationship among the spring’s force, stiffness and extension is
linear. As shown in Figure 10, by setting lt smaller than the ultimate deformation (wu) of the harvester and
meanwhile increasing the stiffness (k) of the spring to a sufficiently large value, then the initial deformation
Manuscript Page 16 of 35
(w0) is never larger than wu, while the constant centrifugal force can be also countered at the target rotation
speed. However, this strategy also leads to a new problem, which is the significant increase of k stiffens the
system and suppresses the vibration of the harvester in large constant centrifugal force cases. It is
noteworthy that the method of using magnets faces the same problem if it is possible to employ the strategy
of increasing the magnetic flux or/and reducing the magnetic spacing. Accordingly, it is imperative to
propose a method to reduce the stiffness of the system no matter what method is adopted.

Figure 10 Strategy of preventing the deformation of the harvester from failures when adopting the linear spring. wu
denotes the ultimate of the harvester’s deformation before failure. ku and cu are respectively the spring’s stiffness and damping
in this case. Provided lt is smaller than wu, the linear spring prevents w from further increasing once w is larger than lt. As a result,
w0 and w never exceed wu and the harvester’s failure is prevented.

4 Method of using the quasi-zero stiffness mechanism

In passive vibration control methods, the quasi-zero stiffness (or high static and low-dynamic stiffness)
vibration isolator currently attracts extensive interest from academia and industry alike due to its high static
load-carrying capacity, excellent performance in low-frequency band, and impact protection ability [53,
54], etc. Its basic principle is to use a negative stiffness mechanism to reduce the original system's positive
stiffness and achieve a stable zero stiffness equilibrium point, near which the stiffness is quasi-zero [55]
and the elastic energy transmission is nearly blocked while the mass is still supported. To develop this
vibration isolation method, extensive investigations have been carried out to improve the dynamical
performance and the structural configuration of the vibration isolators, such as on realizing better force-
displacement characteristics by introducing various nonlinearities [56, 57], on realizing those nonlinearities
by various structural designs of positive or negative stiffness elements [58-60], and even on realizing novel
structures [61, 62]. Up to now, most of the studies on the quasi-zero stiffness mechanism have achieved
excellent results, and the relevant technologies are actually not limited to vibration isolation applications,
for example by Ref. [63]. In this study, when the energy harvester experiences a large constant centrifugal
force, the linear spring needs to provide the same magnitude of force for the harvester based on the proposed
Manuscript Page 17 of 35
method. To protect the harvester from experiencing a large preloaded tensile force and thus failing, the pre-
extension of the linear spring has to be set very small (lt <wu) and the stiffness very large (see Figure 10).
However, this strategy leads to a stiffening of the system and thus a significant vibration suppression of the
harvester (i.e., the high additional stiffness problem). Thus, to solve this problem, inspired by the principle
of the quasi-zero stiffness, we propose to introduce the negative stiffness mechanism to reduce the
additional stiffness of the linear spring.

4.1 Design of the quasi-zeros stiffness device and its working simulation

Figure 11a shows the schematic of the rotational HC-PEH system with a quasi-zero stiffness (QZS) device.
Compared to the previous system of using a linear spring, there is a negative stiffness device (NSD) added
for reducing the stiffness provided by the linear spring, which consists of two compressed springs with a
constant stiffness (kne) in this paper. H0 denotes the initial deformation of the HC-PEH when there are no
effects of gravity and no rotations.  is the angle between the spring of the NSD and the X-axis, and le is
the length of the NSD’s spring when is zero (see Figure 11b). Note that H0 is different from w0, because
the addition of the NSD may change w0 with lt unchanged. Only when the NSD and the linear spring share
the same initial position (see Figure 11c), H0 equals w0. In this case, the springs of NSD keep the original
length (l0) and the initial angle (𝜙0 ), meaning that the addition of the NSD only provides the negative
stiffness and does not contribute to the load-carrying capacity of the original system. Moreover, because
this study only aims to theoretically verify whether the negative stiffness mechanism solves the high
additional stiffness problem, herein we consider that the NSD consists of ideal linear springs which are
essentially the same as the linear spring used in the previous method. This means that the practical and
technical factors, such as friction locking [63, 64] and bending deformations of the compressed spring, are
neglected and the contribution of the NSD’s damping can be quantitively considered into c for simplicity.
Based on the schematic in Figure 11a and 11b as well as the modeling method in Section 2.2, the
theoretical model of the rotational HC-PEH system with a QZS device can be developed to study the
feasibility of the proposed method. Since the addition of the QZS device only changes the stiffness of the
system, thus only the potential energy of the rotational system needs to be rewritten here, i.e.,
2
1 1  l 
 UHC-PEH  k lt  w  L, t   2  kne  l0  e  ,
2
U  UHC-PEH  Uls  U NSD (18)
2 2  cos 

where  =w  L, t  / le . By substituting Eqs. (4) and (18) into Eq. (5), the governing equation of the system

Manuscript Page 18 of 35
Figure 11 The method by adding a negative stiffness device (NSD) to reduce the high stiffness from the linear spring and
the simulation result of an example system. (a) Schematic of the rotational HC-PEH system with a quasi-zero stiffness (QZS)
structure. H0 denotes the initial deformation of the system at x=L along the Y-axis, which depends on the equilibrium between
the harvester and the QZS structure. Additionally, H0 is different from w0 since the NSD also has a contribution to the deformation
of the harvester. When the rotating host accelerates to the target rotation speed (target), there is no deformation in the harvester
(w=0) and the stiffness of the linear spring is reduced to roughly zero. (b) Structural property of the NSD in general static
equilibrium. The linear spring is the same as before and the characteristic of the QZS structure can be adjusted by the geometric
(𝜙0 , le, and H0) and material (kne) parameters of the NSD. (c) Schematic of the previous and the current systems when H0=w0 and
the corresponding simulation result on their comparison. When target is determined, once r is sufficiently large, the system
experiences a huge constant centrifugal force around target. To protect the harvester and counter the centrifugal force at the same
time, the additional stiffness from the linear spring has to be large, causing the system significantly stiffened and hardly to vibrate
even if the adverse effect of the offset distance is completely eliminated. However, the addition of the NSD significantly improves
the RMS voltage of the harvester around target because it reduces the stiffening effect from the linear spring.

can be obtained as

 
M q q   C  Cl  q  Cn q 2 q  K   2 M q  K P0  Kl q   K n q 3  Vq  N  q   Fc  FG  Fl , (19)

Manuscript Page 19 of 35
 V
qq  C pV   , (20)
R
where

 l0  H0 H0
N  q   2kne  L  1   q , l0  , and le  .

  2
 L  q 2
 l 2
e 
 sin  0  tan  0 

Then, to validate the method of using the QZS device, a simulation of the RMS-rotation speed response
when H0 equals w0 is performed. As shown in Figure 11c (refer to Table 1 for the other parameter values),
the result clearly shows that the voltage response of the system only with the linear spring at the target
rotation speed (target) is very low and the addition of the NSD significantly increases the voltage around
target. This not only corresponds to the conclusion in Section 3 that the high additional stiffness coming

with the strategy of letting lt≤wu (for protection of the harvester from over deformations in large constant

centrifugal force cases) indeed causes the significant performance reduction but also indicates that the
method of using the QZS device works well in solving the high additional stiffness problem.

4.2 Dynamical characterization of the system

For a traditional QZS device in vibration isolations, its performance largely depends on the range of the
QZS region where the stiffness of the system is near zero and only increases along with the displacement
at a very low rate. Once outside the QZS region, the stiffness of the system significantly increases with the
displacement at a high rate due to the hardening effect. The broader the QZS region is, the performance of
the vibration isolation is better. Thus, it is important for this study to investigate the effect of the QZS
region’s variation on the harvester’s performance.
Learned from Eq. (19), the equivalent stiffness provided by the NSD is written as

 l0 
K NSD  q   2kne  L  1  . (21)

  2
 
L q 2
 l 2
e 


It reveals that this stiffness varies nonlinearly with the modal coordinate (q) and le, and linearly with kne and
l0. Since both l0 and le are decided by H0 and 𝜙0 , thus H0 or 𝜙0 is more independent when taken as a
variable in simulations. However, considering both H0 and 𝜙0 are geometric parameters of the system,
their effects on the QZS region are similar and thus only H0 is taken as a variable in later simulations. Figure
12a shows the relationship between the force and the displacement of the QZS device as H0 varies, which

Manuscript Page 20 of 35
is obtained via simulations based on the expression of [ Fl  N  q   K l q ]. It well reveals that the range of

the QZS region increases along with the H0. Then, to investigate the effect of the QZS region on the
harvester’s performance, the RMS voltage-rotation speed response is also simulated here, as shown in
Figure 12b. The result reveals that there is an optimal value of H0 when the harvester’s performance is best.
It well indicates that the widening of the QZS region does not always help improve the harvester’s
performance, which is different from its effect on vibration isolations.

Figure 12 Simulations of the rotational HC-PEH system with a QZS device. (a) The force-displacement curve of the QZS
device as H0 varies. The increase of H0 helps broaden the QZS region but the effect is small. (b) The RMS voltage-rotation speed
curves as H0 varies. It reveals that there is an optimal QZS region for improving the electrical performance of the system, which
is a little different from the general understanding of quasi-zero stiffness structures. (c) The force-displacement curve of the QZS
device as kne varies. The variation of kne has a significant effect on the QZS region. (d) The RMS voltage-rotation speed curves
as kne varies. It is noteworthy that there are regions of negative stiffness when kne increases to 105%, which can cause an instability
problem for the system and thus an electrical performance reduction.

Similarly, the force-displacement curve of the QZS device as kne varies and the corresponding RMS
voltage-rotation speed responses are also investigated here. The simulation results are respectively shown
in Figure 12c and 12d. It is observed from Figure 12c that kne has a significant effect on the variation of the
QZS device’s stiffness and there is an optimal value of kne realizing the best QZS region. Additionally, by
comparing Figure 12c with Figure 12d, it is found that the performance of the harvester significantly
reduces when kne increases from 100% to 105%, and the stiffness provided by the QZS structure is negative.
In this case, the system is more likely to become unstable and vibrate in chaotic motion [65], which causes
the RMS electrical performance reduction.
Actually, the increase of H0 also leads to an unstable situation for the system and thus the performance
reduction in Figure 12b. However, the cause is different since the negative stiffness comes from the
Manuscript Page 21 of 35
centrifugal effect. To better reveal this, the time histories of the system at taget are simulated at the target
rotation speed when H0 is 7 mm and 9mm respectively, to show the difference (see Figure 13). When H0 is
7 mm, the system vibrates periodically (see Figure 13a) and its performance is the best in Figure 12b.
Moreover, the phase trajectory of the system indicates that the system is bi-stable. As revealed by Ref. [50],
it is caused by the centrifugal effect, and the system is likely to undergo chaotic motions when the working
condition changes further. As a result, when H0 increases to 9 mm, the system vibrates in chaotic motion
(see Figure 13b). In this case, despite a small change in the voltage amplitude, the RMS voltage significantly
reduces.

Figure 13 Dynamical responses with different H0, corresponding to Figure 12a and 12b. (a) when H0 is 7 mm. The system
is bi-stable and vibrates in a regular motion with period-1. (b) When H0 is 9 mm. The system undergoes chaotic motions while
the RMS voltage of the harvester largely reduces.

In summary, although the proposed method of using the QZS mechanism enables the harvester to work
well around the target rotation speed in a large centrifugal force case, the QZS region still needs to be well
designed for avoiding the aforementioned two causes of the system’s chaotic motions. Alternatively, it is
also potentially effective to try other methods [57, 59, 60] of achieving negative stiffness to solve the high
additional stiffness problem.

Manuscript Page 22 of 35
5 Conclusion

This paper presents methods of using a tensile force provided by a pre-extended linear spring or a quasi-
zero stiffness structure to enable the vibration energy harvesting excited by gravity and rotations in offset
configurations. A distributed-parameter model of the designed energy harvester system where the tensile
force comes from a linear spring is for the first time developed and validated by experiments when the
offset distance, the stiffness, and the extension of the linear spring vary respectively. Based on the model,
the effects of the stiffness, the pre-extension, and the damping on the system’s response are further studied.
A comprehensive discussion on the methods that respectively use the magnetic repulsion and the pre-
extended linear spring is presented and reveals their common issue, i.e., vibration suppression caused by
high additional stiffness, when used in large centrifugal force cases. To address this issue, the quasi-zero
stiffness mechanism is innovatively introduced, and the corresponding theoretical model is then developed
to characterize the new system.
Both the experimental results and their corresponding simulation results show that the proposed
method of using a linear spring to provide the tensile force significantly improves the electrical performance
of the energy harvester in different cases. The maximum improvement rate of the RMS voltage around the
target rotation speed reaches 1225% and the improvement is largely related to the offset distance and the
tensile force. Via the parametric study, it is found that there are optimal values of the stiffness and the pre-
extension in improving the harvester’s electrical performance, and the reduction of the spring’s damping is
beneficial.
Based on the developed model with a Quasi-zero stiffness device, the method of applying the negative
stiffness mechanism to solve the high additional stiffness issue is verified to be significantly effective
around the target rotation speed and thus feasible. In addition, the parametric study also shows that the
region of the quasi-zero stiffness needs to be well designed, to avoid causing the chaotic motion of the
harvester and thus a reduction of its electrical performance.
Although this work is carried out based on a specific energy harvester (HC-PEH), we believe that the
results still provide guidance for other gravity-rotation-excited vibration energy harvesters and their
applications.

Acknolwedgement

The work described in this paper was supported by the National Natural Science Foundation of China under

Manuscript Page 23 of 35
Grant Nos. 12202116 and 11732005, the Postdoctoral Science Foundation of Heilongjiang Province of
China under Grant No. LBH-Z21143, and the Research Grants Council of Hong Kong under Grant Nos.
CityU 11212021 and CityU 21210619.

Reference
[1] Sun C, Chen Y, Hou L. Steady-state response characteristics of a dual-rotor system induced by rub-impact. Nonlinear
Dynamics. 2016;86:91-105.

[2] Hou L, Chen H, Chen Y, Lu K, Liu Z. Bifurcation and stability analysis of a nonlinear rotor system subjected to constant
excitation and rub-impact. Mechanical Systems and Signal Processing. 2019;125:65-78.

[3] Ma H, Lu Y, Wu Z, Tai X, Wen B. Vibration response analysis of a rotational shaft–disk–blade system with blade-tip rubbing.
International Journal of Mechanical Sciences. 2016;107:110-25.

[4] Yang Y, Cao D, Yu T, Wang D, Li C. Prediction of dynamic characteristics of a dual-rotor system with fixed point rubbing—
theoretical analysis and experimental study. International Journal of Mechanical Sciences. 2016;115:253-61.

[5] Lu Z, Hou L, Chen Y, Sun C. Nonlinear response analysis for a dual-rotor system with a breathing transverse crack in the
hollow shaft. Nonlinear Dynamics. 2016;83:169-85.

[6] Yang Y, Ouyang H, Yang Y, Cao DQ, Wang K. Vibration analysis of a dual-rotor-bearing-double casing system with pedestal
looseness and multi-stage turbine blade-casing rub. Mechanical Systems and Signal Processing. 2020;143:106845.

[7] Yang Y, Ouyang HJ, Zeng J, Hui MA, Yang Y, Cao D. Investigation on dynamic characteristics of a rod fastening rotor-
bearing coupling system with fixed-point rubbing. Applied Mathematics and Mechanics (English Edition). 2022;43:1063-80.

[8] Nakhchi M, Naung SW, Rahmati M. Influence of blade vibrations on aerodynamic performance of axial compressor in gas
turbine: Direct numerical simulation. Energy. 2022;242:122988.

[9] Caselitz P, Giebhardt J. Rotor condition monitoring for improved operational safety of offshore wind energy converters. J
Sol Energy Eng. 2005;127:253-61.

[10] Wang Y, Yang Z, Li P, Cao D, Huang W, Inman DJ. Energy harvesting for jet engine monitoring. Nano Energy. 2020:104853.

[11] Dunn B, Kamath H, Tarascon J-M. Electrical energy storage for the grid: a battery of choices. Science. 2011;334:928-35.

[12] Roundy S, Tola J. Energy harvester for rotating environments using offset pendulum and nonlinear dynamics. Smart
Materials and Structures. 2014;23:105004.

[13] Li M, Wen Y, Li P, Yang J, Dai X. A rotation energy harvester employing cantilever beam and magnetostrictive/piezoelectric
laminate transducer. Sensors and Actuators A: Physical. 2011;166:102-10.

[14] Guan M, Liao W-H. Design and analysis of a piezoelectric energy harvester for rotational motion system. Energy Conversion
and Management. 2016;111:239-44.

[15] Zhang Y, Zheng R, Shimono K, Kaizuka T, Nakano K. Effectiveness Testing of a Piezoelectric Energy Harvester for an
Automobile Wheel Using Stochastic Resonance. Sensors (Basel). 2016;16.
Manuscript Page 24 of 35
[16] Khameneifar F, Arzanpour S, Moallem M. A Piezoelectric Energy Harvester for Rotary Motion Applications: Design and
Experiments. IEEE/ASME Transactions on Mechatronics. 2013;18:1527-34.

[17] Pillatsch P, Yeatman EM, Holmes AS. A piezoelectric frequency up-converting energy harvester with rotating proof mass
for human body applications. Sensors and Actuators A: Physical. 2014;206:178-85.

[18] Xue T, Yeo HG, Trolier-McKinstry S, Roundy S. A wrist-worn rotational energy harvester utilizing magnetically plucked
{001} oriented bimorph PZT thin-film beams. 2017 19th International Conference on Solid-State Sensors, Actuators and
Microsystems (TRANSDUCERS). Kaohsiung, Taiwan: IEEE; 2017. p. 375-8.

[19] Zhao L, Zou H, Gao Q, Yan G, Wu Z, Liu F, et al. Design, modeling and experimental investigation of a magnetically
modulated rotational energy harvester for low frequency and irregular vibration. Science China Technological Sciences.
2020;63:2051-62.

[20] Fan K, Wang C, Chen C, Zhang Y, Wang P, Wang F. A pendulum-plucked rotor for efficient exploitation of ultralow-
frequency mechanical energy. Renewable Energy. 2021;179:339-50.

[21] Zou H, Li M, Zhao L, Liao X, Gao Q, Yan G, et al. Cooperative compliant traction mechanism for human-friendly
biomechanical energy harvesting. Energy Conversion and Management. 2022;258:115523.

[22] Zhao L-C, Zou H-X, Zhao Y-J, Wu Z-Y, Liu F-R, Wei K-X, et al. Hybrid energy harvesting for self-powered rotor condition
monitoring using maximal utilization strategy in structural space and operation process. Applied Energy. 2022;314:118983.

[23] Zhao L-C, Zou H-X, Wu Z-Y, Gao Q-H, Yan G, Liu F-R, et al. Dynamically synergistic regulation mechanism for rotation
energy harvesting. Mechanical Systems and Signal Processing. 2022;169.

[24] Yang K, Abdelkefi A, Li X, Mao Y, Dai L, Wang J. Stochastic analysis of a galloping-random wind energy harvesting
performance on a buoy platform. Energy Conversion and Management. 2021;238:114174-.

[25] Wang B, Luo X, Liu Y, Yang Z. Thickness-variable composite beams for vibration energy harvesting. Composite Structures.
2020:112232.

[26] Wang B, Long Z, Hong Y, Pan Q, Lin W, Yang Z. Woodpecker-mimic two-layer band energy harvester with a piezoelectric
array for powering wrist-worn wearables. Nano Energy. 2021;89:106385.

[27] Zou Y, Tan P, Shi B, Ouyang H, Jiang D, Liu Z, et al. A bionic stretchable nanogenerator for underwater sensing and energy
harvesting. Nature communications. 2019;10:2695.

[28] Zhang C, Chen J, Xuan W, Huang S, You B, Li W, et al. Conjunction of triboelectric nanogenerator with induction coils as
wireless power sources and self-powered wireless sensors. Nature Communications. 2020;11:1-10.

[29] Chen J, Wang ZL. Reviving vibration energy harvesting and self-powered sensing by a triboelectric nanogenerator. Joule.
2017;1:480-521.

[30] Ghaderi B, Nayyeri V, Soleimani M, Ramahi OM. Pixelated metasurface for dual-band and multi-polarization
electromagnetic energy harvesting. Scientific reports. 2018;8:13227.

[31] Mei X, Zhou R, Fang S, Zhou S, Yang B, Nakano K. Theoretical modeling and experimental validation of the centrifugal
softening effect for high-efficiency energy harvesting in ultralow-frequency rotational motion. Mechanical Systems and Signal
Processing. 2021;152.
Manuscript Page 25 of 35
[32] Fu H, Mei X, Yurchenko D, Zhou S, Theodossiades S, Nakano K, et al. Rotational energy harvesting for self-powered
sensing. Joule. 2021;5:1074-118.

[33] Wang Y, Yang Z, Cao D. On the offset distance of rotational piezoelectric energy harvesters. Energy. 2021;220:119676.

[34] Fang S, Xing J, Chen K, Fu X, Zhou S, Liao W-H. Hybridizing piezoelectric and electromagnetic mechanisms with dynamic
bistability for enhancing low-frequency rotational energy harvesting. Applied Physics Letters. 2021;119:243903.

[35] Gu L, Livermore C. Compact passively self-tuning energy harvesting for rotating applications. Smart materials and
structures. 2011;21:015002.

[36] Hsu J-C, Tseng C-T, Chen Y-S. Analysis and experiment of self-frequency-tuning piezoelectric energy harvesters for
rotational motion. Smart Materials and Structures. 2014;23:075013.

[37] Mei X, Zhou S, Yang Z, Kaizuka T, Nakano K. A passively self-tuning nonlinear energy harvester in rotational motion:
theoretical and experimental investigation. Smart Materials and Structures. 2020;29:045033.

[38] Mei X, Zhou S, Yang Z, Kaizuka T, Nakano K. A tri-stable energy harvester in rotational motion: Modeling, theoretical
analyses and experiments. Journal of Sound and Vibration. 2020;469:115142.

[39] Mei X, Zhou R, Yang B, Zhou S, Nakano K. Combining magnet-induced nonlinearity and centrifugal softening effect to
realize high-efficiency energy harvesting in ultralow-frequency rotation. Journal of Sound and Vibration. 2021;505:116146.

[40] Mei X, Zhou S, Yang Z, Kaizuka T, Nakano K. Enhancing energy harvesting in low-frequency rotational motion by a quad-
stable energy harvester with time-varying potential wells. Mechanical Systems and Signal Processing. 2021;148:107167.

[41] Zhao L-C, Zou H-X, Wu Z-Y, Gao Q-H, Yan G, Liu F-R, et al. Dynamically synergistic regulation mechanism for rotation
energy harvesting. Mechanical Systems and Signal Processing. 2022;169:108637.

[42] Fang S, Wang S, Zhou S, Yang Z, Liao W-H. Exploiting the advantages of the centrifugal softening effect in rotational
impact energy harvesting. Applied Physics Letters. 2020;116:063903.

[43] Fang S, Wang S, Zhou S, Yang Z, Liao W-H. Analytical and experimental investigation of the centrifugal softening and
stiffening effects in rotational energy harvesting. Journal of Sound and Vibration. 2020;488:115643.

[44] Fang S, Wang S, Mei X, Zhou S, Yang Z, Liao W-H. A centrifugal softening impact energy harvester with the bistability
using flextensional transducers for low rotational speeds. Smart Materials and Structures. 2020;29:115024.

[45] Fang S, Wang S, Miao G, Zhou S, Yang Z, Mei X, et al. Comprehensive theoretical and experimental investigation of the
rotational impact energy harvester with the centrifugal softening effect. Nonlinear Dynamics. 2020;101:123-52.

[46] Zou H-X, Zhang W-m, Li W-B, Wei K-X, Gao Q-H, Peng Z-K, et al. Design and experimental investigation of a
magnetically coupled vibration energy harvester using two inverted piezoelectric cantilever beams for rotational motion. Energy
Conversion and Management. 2017;148:1391-8.

[47] Yang Z, Zu J. High-efficiency compressive-mode energy harvester enhanced by a multi-stage force amplification mechanism.
Energy conversion and management. 2014;88:829-33.

[48] Chen K, Gao Q, Fang S, Zou D, Yang Z, Liao W-H. An auxetic nonlinear piezoelectric energy harvester for enhancing
efficiency and bandwidth. Applied Energy. 2021;298:117274.
Manuscript Page 26 of 35
[49] Wang Y, Zhao Y, Chen C, Cao D, Yang Z. Misalignment-induced bending-torsional coupling vibrations of doubly-clamped
nonlinear piezoelectric energy harvesters. Mechanical Systems and Signal Processing. 2022;169:108776.

[50] Wang Y, Yang Z, Cao D. Distributed-parameter modeling and dynamic analysis of rotational compressive-mode energy
harvesters. Nonlinear Dynamics. 2021;103:157-82.

[51] Li H, Yang Z, Zu J, Qin W. Distributed parameter model and experimental validation of a compressive-mode energy
harvester under harmonic excitations. AIP Advances. 2016;6:085310.

[52] Blevins RD. Formulas for natural frequency and mode shape. New York: Van Nostrand Reinhold; 1979. p. 108-9, 76.

[53] Yang T, Zhou S, Fang S, Qin W, Inman DJ. Nonlinear vibration energy harvesting and vibration suppression technologies:
Designs, analysis, and applications. Applied Physics Reviews. 2021;8:031317.

[54] Wang Q, Zhou J, Xu D, Ouyang H. Design and experimental investigation of ultra-low frequency vibration isolation during
neonatal transport. Mechanical Systems and Signal Processing. 2020;139:106633.

[55] Huang X, Chen Y, Hua H, Liu X, Zhang Z. Shock isolation performance of a nonlinear isolator using Euler buckled beam
as negative stiffness corrector: Theoretical and experimental study. Journal of Sound and Vibration. 2015;345:178-96.

[56] Zou D, Liu G, Rao Z, Tan T, Zhang W, Liao W-H. A device capable of customizing nonlinear forces for vibration energy
harvesting, vibration isolation, and nonlinear energy sink. Mechanical Systems and Signal Processing. 2021;147:107101.

[57] Gatti G. Optimizing elastic potential energy via geometric nonlinear stiffness. Communications in Nonlinear Science and
Numerical Simulation. 2021;103:106035.

[58] Gatti G. An adjustable device to adaptively realise diverse nonlinear force-displacement characteristics. Mechanical Systems
and Signal Processing. 2022;180:109379.

[59] Zhao F, Ji J, Ye K, Luo Q. An innovative quasi-zero stiffness isolator with three pairs of oblique springs. International
Journal of Mechanical Sciences. 2021;192:106093.

[60] Gatti G. Effect of parameters on the design of a suspension system with four oblique springs. Shock and Vibration.
2021;2021.

[61] Yan G, Zou H-X, Wang S, Zhao L-C, Wu Z-Y, Zhang W-M. Bio-inspired toe-like structure for low-frequency vibration
isolation. Mechanical Systems and Signal Processing. 2022;162:108010.

[62] Han H, Sorokin V, Tang L, Cao D. Lightweight origami isolators with deployable mechanism and quasi-zero-stiffness
property. Aerospace Science and Technology. 2022;121:107319.

[63] Gatti G, Shaw A, Gonçalves P, Brennan M. On the detailed design of a quasi-zero stiffness device to assist in the realisation
of a translational Lanchester damper. Mechanical Systems and Signal Processing. 2022;164:108258.

[64] Yin Y, Sun B, Han F. Self-locking avoidance and stiffness compensation of a three-axis micromachined electrostatically
suspended accelerometer. Sensors. 2016;16:711.

[65] Kovacic I, Brennan MJ, Waters TP. A study of a nonlinear vibration isolator with a quasi-zero stiffness characteristic. Journal
of sound and vibration. 2008;315:700-11.

Manuscript Page 27 of 35
Appendix A Mathematical formulations

Nomenclature

L : Length of the support beam;

b : Width of the support beam;

h : Thickness of the support beam;

A : Cross-sectional area of the support beam;

Density of the support beam;

E : Young’s modulus of the support;

I : Moment of inertia the support;

Rotation angle of the support;

s : Length of the support beam after deformations;

Mid-plane stretching of a bending beam;

xStretching of the bow-shaped beam along x-axis;

yStretching of the bow-shaped beam along y-axis;

aMid-plane stretching of the support beam at x =L;

FaAxial load due to the mid-plane stretching of the support beam at x =L;;

w (x, t ) Transverse deflection of the support beam;

LBWaist length of the bow-shaped beam;

BBDistance between the proof mass and the PZT plate;

HBHeight of the bow-shaped beam’s waist;

hB Original thickness of the bow-shaped beam;

Angle between the PZT plate and the waist of the support beam;

Initial angle between the PZT plate and the waist of the support beam;

k Equivalent torsional stiffness of the bow-shaped beam;

F Elastic force from each bow-shaped beam along x-axis;

Pt Transient axial force (Pt) acting on the support beam;

FpReaction force from the PZT plate;

P0 Preloaded axial force;

S Length of the central cymbal structure;

 Deformation transmission efficiency;

Softening coefficient of the nonlinear stiffness due to the connection by glue between beams

US  Elastic potential energy of the two support beams due to elastic deformations;

Manuscript Page 28 of 35
UB  Elastic potential energy of the bow-shaped beams;

UP  Elastic potential energy of the PZT plate;

U f  Total potential energy of the system in the centrifugal field;

f
US  Potential energy of the support beams in the centrifugal field;
f
UM  Potential energy of the proof mass in the centrifugal field;
f
UP  Potential energy of the PZT plate in the centrifugal field;

UPe  Electrical potential energy of the PZT plate;

UP  Total potential energy of the PZT plate;

U Total potential energy of the system;

T Total kinetic energy of the system;

W External virtual work;

M Equivalent proof mass;

Mp Mass of the PZT plate;

KP0 linear stiffness from the preloaded axial force;

pi Mechanical stress of the PZT plate along i-axis (i =x, y );

pi Mechanical strain of the PZT plate along i-axis (i =x, y ); 

Ex Electrical field strength;

Dp (i ) Electrical displacement (along i-axis, i =x, y );

V Voltage generated by the PZT plate;

Vpx Volume of the PZT plate’s bonded region;

Vpy Volume of the PZT plate’s compressive region;

Hx Length of the PZT plate's bonded area;

Hy Length of the PZT plate's compressive area;

hp Thickness of the PZT plate;

bp Width of the PZT plate;

Ep Young’s modulus of the PZT plate;

d31 Coupling coefficient of the PZT plate along its length;

d33 Coupling coefficient of the PZT plate along its thickness;

33 Electrical permittivity of the piezo plate;

Q Electric charge output of the PZT plate;

R External resistance;

cS Damping coefficient from the support beam;

cB Damping coefficient from the bow-shaped beam;

Manuscript Page 29 of 35
 r Offset distance;

 Rotation speed;

 Rotation angle;

g Gravitational acceleration;

n (x ) n-th linear vibration mode function of the support beam;

 (x ) Fundamental vibration mode function of the support beam;

qn (t ) n-th generalized coordinate;

q Generalized coordinate corresponding to W (x );

q0 Equilibrium position of the generalized coordinate corresponding to W (x );

La Lagrange’s function;

i Dimensionless parameter related to the boundary condition (i-th natural mode);

i Dimensionless natural frequency parameter (i-th natural mode);

𝑈 : Static potential energy of the system.

According to Ref. [49], the total potential of the rotational HC-PEH system consists of the elastic

potential energy from the support beam ( U S ) and the bow-shaped beam( U B ), the potential energy of the

system in the centrifugal field ( U f ), the elastic and the electrical potential energy of the PZT plate ( U P ),

which is written as (see Nomenclature for each symbol)

U HC-PEH  U S  U f  U B  U P
2 2
EAD22  L  w   EAD2 D1  L  w  2 
2
L  2w 

4 L  0  x 
EI  2  dx     dx    0   dx  V
0
 x   L   x  
2
EAD2 D0  w  L EAD12 2 2 EAD0 D1 EAD0

L 0  x  dx 
L
V 
L
V
L
 L  1   1  1 
 2   2    A  wr  w2  dx   M  M P   rw  L, t   w2  L, t   
 0  2   2  2 
2
8k D0  L  w 
2
2  2 k  L  w 2  16k D0 D1

H yHB    dx 
 x  H y H B  0  x 
   dx  
H yHB
V

0

8k D12V 2 8 k D1V L  w 

2
16k D0 2


H yHB

H yHB  0   dx 
 x  H yHB
3d 31 pyV pyV 3d33 pxV pxV  2pxV px  py
2
V py  33V 2
      2V px  V py  , (A1)
2hp hp Ep 2E p hp2

where

2k L 2 LE p hp bp tan 2 0 P0 L d 31 LE p bp tan  0 D

D + , D0  , D1   , D2  ,   1  D2 .
H EA2
B H y EA 1  D  EA 1  D  EA 1  D 
Manuscript Page 30 of 35
The kinetic energy of the rotational HC-PEH system is written as

1 L 1 1
THC-PEH  2    Aw 2 dx  2  Mw 2  L, t   M p w 2  L, t  . (A2)
2 0 2 2

The dissipation function (WHC-PEH) that is introduced to account for the effect of the primary mechanical
and electrical damping phenomena on the rotational HC-PEH system is written as
1 L 1
  Q V  P0 x
 WHC-PEH =  cS  w  wdx  cB
2 0 2 (A3)
  2M  M p  g cos  t   w  L, t  +2   gA cos  t   wdx,
L

Appendix B Expressions of the components and the coefficients in the

governing equation

The coefficients of governing equations are given as follows

U HC-PEH U S U f U B U P
   
q q q q q
2
 L  2  
2
EAD22  L   
2
 3 2 EAD2 D1  L   2 
   2EI  2  dx  q 
L  0  x 
   dx  q   0   dx  Vq
 0  x    L   x  
2EAD2 D0  L    
2
2   1  
L
  0   dx  q  2   r  0  Adx   M  M P    L 
L   x     2  
 L  1  
 2   2 q    A 2 dx   M  M P   2  L 
 0  2  
2
4 k D0  L    2  2  2 k  L   2  3 4 k D1  L    2 
H B2  0  x  H B2  0  x  H B2  0  x 
    dx  q     dx  q     dx  qV
  
 2d b h tan 0   L   2 
 3 E p tan 0  d31bp  33 p p
   0   dx  qV
 H y   x  
2
 2V px hp2 tan 2 0  2 2  2 E p tan 2 0  L   2  3
 Vpy    0   dx  q
 H x2  H y2   x  
2 E p tan 0  2V px hp tan 0   d31 2 D1 tan 0   L    2 
2 2

  V py      0   dx  qV (B1)
Hy  H 2
x  h p H y   x  
4 2  D0 E p tan 2 0  2V px hp2 tan 2 0   L    2 
 V py    0   dx  q,
H y2  H x2    x  

Manuscript Page 31 of 35
U HC-PEH U S U f U B U P
   
V V V V V
EAD2 D1  L     2 2 EAD12
2 2
2EAD0 D1 4k D0 D1 4k D1
  0   dx  q + V    V
L   x   L L H B2 H B2
 2  k D1 L   2  2 3  2d33 hp tan 0   L    2  2
 2 0  x  dx  q + 2  bp E p tan 0  d31  H y   0  x  dx  q
 H B    
d 2 D1 tan 0 
 3E p bp  2d 33 hp tan 0  d31 H y   31   V
 hp H
 y 
 2d h tan 0   V
 3 E p D0bp tan 0  d31  33 p
   2  2V px  V py  332
 Hy  hp
 E p tan 0  2Vpx hp2 tan 2 0   d31 2 D1 tan 0   L   2  2
  V py       0   dx  q
Hy  H 2
x  h p H y   x  
2
 2V E h 2 tan 2 0   2 D1 tan 0 d31 
+  E pV py + px p p2    V
 H H h
 x  y p 
 2h tan 0   2 D1 tan 0 d31 
2 2 (B2)
 2 D0 E p bp hp tan 0  1+ p   ,
 H x H y   Hy hp 

2
   L
2  2
 cB  D1  
2
 dx
 WHC-PEH 1  cB      x 
qV
L L 0
=   cS   dx  q   dx  q q 
2 H B2  0  x 
2 2
 
q 2 0   2 H 2
B

 L   
2

dx  q   2M  M p    L  g cos  t  +2 gA cos  t  dx
L
  P0    
 0
 x   0
(B3)
2
1 L  c  2  L   2  2
   cS   2 dx  q  B 2  0   dx  q q
2 0  2H B   x  
 L   
2

  2M  M p    L  g cos  t  +2
L
gA  cos  t  dx   P0     dx  q,
0

0
 x  
2
L   
cB  D1    dx
 WHC-PEH 0
 x  cB D12V
 Q  P0 D1  
qq   Q, (B4)
V 2 H B2 2 H B2

M q  2 A    dx    M
L

0
2
p  2M   2  L  , (B5)

1 L
C  cS   2 dx, (B6)
2 0
2
cB  2  L    
2

2 H B2  0  x 
Cn     dx  ， (B7)

Manuscript Page 32 of 35
  2EAD2 D0 4 2  D0 E p tan 2 0  2V px hp2 tan 2 0  4 k D0  L    2
K    py
V     0   dx
 L H y2  H 2
x  H 2
B   x 
(B8)
2
L  
2
+  2EI  2  dx,
0
 x 

 L   
2

K P0   P0     dx , (B9)
 0
 x  
2
 EAD 2 2  2 k  2 2  2 E p tan 2  0  2V px h p2 tan 2 0    L    2 
K n  
 L
2
    V py     0   dx  ,
H B2 H y2 H x2     x  (B10)
   

FG   2 M  M p    L  +2 A B dx  g cos   t  ,

L
(B11)
 0 
 L  1  
Fc  2 2 r    Adx   M  M P    L   , (B12)
 0  2  
  2d 33 hp tan  0 
  3 bp E p tan 0  d 31  
  H
 y 
2 E p tan 0  2V px h p2 tan 2 0   2 D1 tan 0 d31 
  V py      (B13)
Hy  H x2  Hy hp 
 EAD2 2  k   L   
2

H B2   0  x 
+2D1      dx,
 L
d 2 D1 tan 0 
C p =3E p bp  2d33 h p tan  0  d31 H y   31  
h Hy
 p 
 EA 2k  2 
+2   2  D1  2  2V px  V py  332 (B14)
 L HB  hp
2
 2V px E p hp2 tan 2 0   2 D1 tan 0 d31 
+  E pV py +    ,
 H 2
H h
 x  y p 

 EAD0 2k D0   2d33 hp tan 0 

Q0  2 D1     3 E D b
p 0 p tan   d
0  31  
 L H B2   Hy 
(B15)
 2h 2 tan 2 0  2 D1 tan 0 d 31 
 2 D0 E p bp hp tan 0 1+ p    .
 H H  H h
 x y  y p 

Appendix C Fabrication of the HC-PEH Prototype

The HC-PEH prototype in experiments was fabricated by a soft piezoelectric ceramic (PZT-5H plate,
original size: 40×15×1 mm3), 3D-printed steel bow-shaped beams with partially thickened waists and

Manuscript Page 33 of 35
flexible hinges (each beam is integrated with a mass block and bonded with two nuts via epoxy glue), and
two steel support beams, as shown in Figure 3. There is no limitation to the selection of the materials used
for the fabrication of the HC-PEH, which can be changed based on the working condition and requirements.
To make it simple, there were four areas respectively on both sides of each end of the PZT plate, which are
used to bond bow-shaped beams via epoxy glue. Thus, the size involving the compressive region of the
PZT plate is 36×15×1 mm3, while the size of each bonded region is 2×15×1 mm3. There is a slot in each
bow-shaped beam, and one end of a support beam is bonded with the bow-shaped beam in its slot via epoxy
glue. It is noteworthy that this design helps reduces the bending stiffness of the harvester compared to the
previous design [49, 50]. However, the connection was found to be weak in experiments and finally caused
an axial stiffness reduction for the support beam, which resulted in a decrease in the nonlinear stiffness of
the system. Since a weak connection only has fewer effects on the investigation, it only helps achieve a
better performance of the nonlinear energy harvester and thus we did not carry out this work in this study.

Appendix D Supplementary Figures and Tables

Figure D1. Schematic of the simplified boundary condition of the HC-PEH.

Manuscript Page 34 of 35
 Figure D2. Schematic of the stopper designed to limit the deformation of the HC-PEH in the initial state.

Table D1 Geometric properties and material parameters of the HC-PEH prototype used in the experiment.

Description Value
Length of the support beam 74 mm
Thickness of the support beam 0.25 mm
Width of the support beam 4.5 mm
Thickness of the thickened waist 2 mm
Original thickness of the bow-shaped beam 0.2 mm
Width of the bow-shaped beam 15 mm
Length of the thickened waist 13.37 mm
Angle between the bow-shaped beam and the PZT plate 9 degree
Length of the flexible hinge 1 mm
Young’s modulus of the support beam 200 GPa
Young’s modulus of the piezo plate 63 GPa
Density of the support beam 7850 kg/m3
Density of the piezo plate 7800 kg/m3
Size of the piezo plate’s compressive region 36×15×1 mm3
Each size of the piezo plate’s bonded region 2×15×1 mm3
Static capacitance of the piezo plate 8.73 nF
Proof mass 26.93 gram
Overall mass 58.55 gram

Table D2 Parameters of the linear spring used in the simulation

Description Value
Stiffness of the linear spring-1 (k) 16.95 N/m
Damping coefficient of the linear spring-1 (c) 0.4 Ns/m
Stiffness of the linear spring-2 (k) 37.75 N/m
Damping coefficient of the linear spring-2 (c) 1 Ns/m

Manuscript Page 35 of 35