International Micro Air Vehicle Conference and Flight Competition (IMAV) 2017 223

Development and Design Methodology of an

Anti-Vibration System on Micro-UAVs
Zhenming Li1 , Mingjie Lao2 , Swee King Phang2∗, Mohamed Redhwan Abdul Hamid2 , Kok Zuea Tang3 , and Feng Lin2
1
Mechanical Engineering, National University of Singapore, Singapore
2
Temasek Laboratories, National University of Singapore, Singapore
3
Electrical and Computer Engineering, National University of Singapore, Singapore

A BSTRACT

As the potential applications of unmanned aerial

vehicles (UAVs) are growing, more sensors are
installed on-board. Mechanical vibration of
the UAV, which greatly hinders the accuracy
of its on-board sensors, becomes an increas-
ingly important issue. In this manuscript, an
anti-vibration framework on micro-UAVs is pro-
posed. The vibration sources of the UAV will be
investigated and identified. Then, several selec-
tions of hardware dampers are tested along with
a digital low-pass filter on actual UAV. With the
Figure 1: T-Lion developed at TL@NUS
results from different case studies, a criteria of
damper selection for micro-UAV is built to serve
as a guideline for further practice. avoid aliasing effect, one common method is to sample the
1 I NTRODUCTION data with the rate of at least 2 times of the maximum fre-
quency of the signal. However, there is a limit to how fast
An unmanned aerial vehicle (UAV), commonly known as the sampling frequency is and the relative high frequency of
a drone, may be controlled by a remote-pilot or operated au- vibration generated signals will continue to impair the accu-
tonomously without human intervention. Due to its small size racy of the useful signals. Insufficient damping to such vibra-
and remoteness during operation, modern miniature-UAVs tions will also likely result in sensor drifting such as coning
are now widely used in both commercial and military aspects and sculling motion which is the condition where the output
to fulfill the needs in different scenarios [1]. One of the pop- received by gyroscope or accelerometer becomes more inac-
ular research areas is on scouting of remote areas with lim- curate over time.
ited GPS reception (such as indoor or foliage environment) Acknowledging the importance of implementing anti-
through on-board cameras for localization mean [2]. vibration measures on-board, different dampers and isolators
To realized such applications, useful data are often col- have been considered and recommended for vibration min-
lected from the sensors such as inertial measurement unit imization. For instants, Kyosho Zeal sheet and its perfor-
(IMU) and camera images at low frequencies (1-20 Hz ap- mances have been briefly studied in [8]; Wire-rope isolators
proximately) to realize navigation or localization algorithms and rubber dampers for minimum vibration effects towards
by estimating the UAV position and bearing [3, 4, 5]. The ac- data logging during flight has been discussed in [9]. In this
curacy of these data is important for the flight control system manuscript, theoretical research will be done on a selections
to ensure good performance during flights. One major factor of dampers and isolators for the best anti-high-frequency vi-
hindering the accuracy of these data is the mechanical vibra- brations to be used in UAV. The results will be further verified
tion of the UAV during flights. Among the multiple causes to by state-of-the-art bench tests and UAV flight trials.
the vibration problem, rotating rotors and structure vibration The UAV to be used to verify the proposed anti-vibration
on its natural frequencies are of most concern [6, 7]. system is an in-house UAV developed by the Temasek Labo-
Fortunately, these vibrations are usually recorded as ratories at the National University of Singapore (TL@NUS),
higher frequency signals by IMU and they tend to pollute the codenamed T-Lion (see Fig. 1). It weighs 3 kg and is capa-
useful signals at low frequency through the phenomenon of ble to carry additional 2 kg payload. On-board system to be
aliasing. Under such effect, higher frequency signals are ob- isolated from the UAV structural vibration consists of a full
served as lower frequency signals causing inaccurate data. To IMU sensor suite weighs 120 g.
∗ Email address: [email protected] The manuscript is divided as follows: Section 1 will be

1
 224 International Micro Air Vehicle Conference and Flight Competition (IMAV) 2017

given as introduction to this work; Section 2 lists the rele- inverse linear relationship; larger static deflection will have
vant theoretical frameworks involved in the study; Section 3 a lower system natural frequency, and vice versa. As men-
discusses the sources of vibrations in UAV; case studies of se- tioned earlier, lower natural frequency leads to a larger sepa-
lected dampers are examined in Section 4; results verification ration from the disturbing frequency, thus better isolation effi-
via flight trials will be shown in Section 5 while concluding ciency. Therefore, it is important to use soft, flexible dampers
remarks will be made in the last Section. for light damping mass to ensure a fair amount of static de-
flection is present.
2 T HEORETICAL F RAMEWORKS
2.1 Vibration Damping and Vibration Isolation 2.4 Damper Selection Criteria
Isolators are used to lower the natural frequency of the Judging from the above-mentioned characteristics, a few
system to below the excitation (or disturbing) frequency essential criteria for selecting dampers to be used in our stud-
(which in this case, the vibration frequency created by mo- ies are
tors). Ultimately, the aim is to keep these two frequencies 1. Electrical insulator to avoid short-circuit;
out-of-sync by 180◦ so that to avoid effects such as resonance.
On the other hand, dampers are used to remove mechanical 2. Soft and flexible;
energy from disturbing vibration out of the system by absorb-
ing the energy and converting to other forms of energy such 3. Natural frequency outside UAV structural resonance
as heat [10]. Normally, anti-vibration devices available in the zone;
market are both isolators and dampers. For simplicity, these
are referred as dampers in this paper. 4. Low compression set and low creep;
2.2 Transmissibility Curve 5. Good resistance to outdoor conditions; and
For effective vibration isolation from undesired high fre-
quency, damper would need to have a natural frequency at 6. Easy installation and adjustment.
least less than 50% of the lowest disturbing frequency, and
3 S OURCE OF V IBRATION
optimally, less than 71%. In other words, the ratio of√disturb-
ing frequency over natural frequency is more than 2. This Two main sources of high amplitude vibrations come
can be explained by transmissibility which is defined as the from the rotating rotors, which has the same frequency to
ratio of force transmitted through the suspension apparatus to the rotating rate, and structural natural frequency vibrations,
force applied by vibration and it can be calculated through which the frequencies depend on the UAV structure. For T-
      Lion UAV, the rotors rotates at approximate 50 revolutions
 Ao   ao   Fo  per second, which translates to 50 Hz vibration on the UAV.
T =   ≡   ≡  , (1)
Ai ai Fi This section will be divided into two parts, where the struc-
tural natural frequencies will be studied in a simulation and
where Ao and Ai are the amplitudes for output and input re- the overall vibrations over a large frequency band will be ob-
spectively, ao and ai are the accelerations for output and in- tained through experimental data with actual flight.
put respectively, Fo and Fi are the forces for output and input
(applied and transmitted) respectively [11]. 3.1 Structural Vibration Analysis
Vibrations can never be completely removed from the sys- The T-Lion UAV is modeled in SolidWorks simulation,
tem, i.e., T 6= 0 and hence minimizing this ratio for a specific with the actual material properties assigned to each part of
frequency range is the goal of vibration isolation. Transmissi- the simulated model. Frequency analysis on the model was
bility curve can be divided into three regions: no effect region carried out and vibration frequencies with significant impact
(T = 1), amplification region (T > 1), and isolation region are recorded as follows:
(T < 1). In general, it is highly desirable to have the natural
frequency of the damper as far apart as possible from the dis- 1. There are obvious high magnitude vibrations on x- and
turbing frequency to achieve effective vibration isolation. In y-directions at 39.90 Hz (see Fig. 2 for mode shape
general, isolation efficiency increases as the transmissibility visualization), and on z-direction at 80.48 Hz. From
decreases along increasing system frequency. the simulation result, they are mainly caused by the
hanging payload of T-Lion UAV below the UAV cen-
2.3 Static Deflection
tral region. It is also observed that similar behaviour
Static deflection is how much the damper deflects when also exists at 160.17 Hz and 321.82 Hz, which strongly
it is subjected to the static weight of the equipment it car- suggested that they are the 3rd and 4th mode natural
ries. In general, the larger the static deflection that the damper frequencies of the payload; and
has before damping process, the better isolation effect can be
achieved. In a single degree-of-freedom (DoF) system, nat- 2. Several small amplitude vibrations are observed be-
ural frequency and static deflection are estimated to have an tween 100 to 200 Hz with lesser than 1 mm amplitude
 International Micro Air Vehicle Conference and Flight Competition (IMAV) 2017 225

C→

|ac x (f)|
←A
0.5
A→
←B
A→
0
0 50 100 150 200 250 300 350 400 450
Frequency (Hz)
0.4

0.3

|ac y (f)|
0.2

0.1

Figure 2: Mode shape of T-Lion at 39.9 Hz 0

0 50 100 150 200 250 300 350 400 450
Frequency (Hz)
1

|ac z (f)|
0.5

A→
0
0 50 100 150 200 250 300 350 400 450
Frequency (Hz)

Figure 4: Vibration of T-Lion UAV measured by ADIS IMU

Figure 3: Mode shape of T-Lion at 109.88 Hz of the UAV where the battery and certain weight source
were located at the bottom of the UAV;
(see Fig. 3 for mode shape visualization). A strong vi- 2. B: Vibration signals at B is originated from the rotating
bration of 3.15 mm amplitude is also observed at fre- rotors at approximate 50 Hz; and
quency of 273.16 Hz. They are most likely the struc-
tural natural frequencies of the UAV due to extending 3. C: Vibration of the structure of UAV, mainly on the
arms and components. platform and the extension arms.
These vibration frequency peaks are rather consistent to
3.2 Actual Flight Data
our simulated results. It is concluded that structural vibra-
To verify the vibration frequency of the UAV structure, an tions are mainly high frequency signals, and some of them
actual flight trial was carried out where the T-Lion was com- are so high that low-cost IMU sensor (with sampling rate be-
manded to hover for several seconds in the air. An additional low 100 Hz) would not detect them. They are undesired and
IMU sensor, the ADIS IMU was installed to the UAV for thus dampers will be designed to filter them.
higher sampling rate of more than 800 Hz. With this sensor,
acceleration data can be collected and converted to frequency 4 C ASE S TUDIES
domain with a frequency range up to 400 Hz. Experiments According to the damper selection criterion discussed in
are designed to identify and examine different vibration sig- Section 2, four dampers have been selected for performance
nals and their corresponding vibration locations of the UAV. evaluations. The dampers can be visualized in Fig. 5.
Result in Fig. 4 shows that there is significant amount 4.1 Silicone Ball Damper
of vibration signals at high frequencies between 40 Hz to
400 Hz. Their magnitudes are large and thus cannot be ne- Silicone ball damper has similar functionality to the rub-
glected. By comparing the frequency response of the actual ber damper, but it is made by better material. In general, sili-
system with the simulated results above, vibrations of the cone damper is softer than rubber damper leading to larger
UAV can be divided into 3 parts, demarcated clearly in Fig. 4: static deflection, which is preferable to reduce vibrations.
However, it is difficult to install and to make changes to the
1. A: These low frequency vibrations (approximately damper on UAV due to its complicated installation mecha-
40 Hz at x- and y-directions, approximately 80 Hz at nism. Furthermore, it requires high damping weight (> 200
z-direction) correspond to the vibrations at the payload g) for efficient damping while the damping mass for T-Lion
 226 International Micro Air Vehicle Conference and Flight Competition (IMAV) 2017

Figure 5: (a) Silicon Ball; (b) Silicon Foam; (c) Sorbothane

sheet; and (d) Kyosho Zeal sheet

Figure 6: Vibration test setup on vibration table

UAV is too light for this requirement (120 g). This option is
thus discarded and will not be further discussed.
4.2 Silicone Foam
Silicone Foam is the softest and the most flexible damper
among all. Therefore, it is possible to damp a light weight
IMU alone without the need to add extra mass. However, it
may have very low natural frequency and eventually cause
resonance with the useful signals at low frequency. On top of Kyosho Zeal sheet
that, it breaks easily and thus not suitable to reuse. 3 6x6x5
12x12x5
Transmissibility

12x12x10
4.3 Kyosho Zeal and Sorbothane 30 Durometer Sheets 2

Kyosho Zeal sheet and Sorbothane 30 Durometer sheet 1

are most used tape dampers recognized by many studies and
experiments. They are easy to install and relatively soft to 0
0 50 100 150 200 250 300
accommodate light masses. These can also be installed by Frequency (Hz)

clamping the mass in between two pieces of pads and com- 3

Sorbothane sheet

pressing them to about 80% of its original thickness. This in-

Transmissibility

stallation method enables the UAV to do inverse flying with- 2

out compromising on damping. Kyosho Zeal sheet is rela- 1

tively softer than Sorbothane 30 Durometer sheet, but their
performance is comparable. However, these sheets have rela- 0
0 50 100 150 200 250 300
tively higher compression set. Frequency (Hz)
Silicone Foam
4.4 Damper Performances 6
Transmissibility

Using vibration table, static vibration tests have been 4

done to verify the dampers performances subjecting to vibra-

tions for a frequency range from 10 to 300 Hz. The actual 2

loading on T-lion UAV is 120 g in total, which translates to 0

30 g per damper. Here, the stimulated damping mass used is 0 50 100 150
Frequency (Hz)
200 250 300

30 g to be consistent to the actual UAV load mass. Different

dimensions are tested for selected dampers. Fig. 6 shows an
example of the set-ups of the system on vibration table.
With Equation 1, transmissibility of each damper can be Figure 7: Transmissibility curve with different dimensions
calculated with output and input acceleration data collected. for Kyosho Zeal sheet, Sorbothane sheet, and Silicone foam
Transmissibility curve for each damper is obtained by plot-
ting the calculated transmissibilities with its corresponding
frequencies. Observed from the results shown in Fig. 7, there
 International Micro Air Vehicle Conference and Flight Competition (IMAV) 2017 227

0.5
|ac x (f)|

0
0 50 100 150 200 250 300 350 400 450
Frequency (Hz)
0.5
|ac y (f)|

0
0 50 100 150 200 250 300 350 400 450
Frequency (Hz)
0.5
Figure 9: Kyosho Zeal sheet test results on three axes of UAV
|ac z (f)|

Experiment Dimension (mm) Installation Contact

0
1 12 × 12 × 5 Single Less
0 50 100 150 200 250 300 350 400 450 2 12 × 12 × 5 Single More
Frequency (Hz)
3 12 × 12 × 10 Single More
4 6×6×5 Single More
5 6×6×5 Clamped More
Figure 8: Vibration generated in actual UAV flight recorded 6 12 × 12 × 5 Clamped More
by ADIS IMU sensor 7 12 × 12 × 10 45◦ More

Table 1: Kyosho Zeal sheet installation parameters

is a clear shifts of the transmissibility curve towards the left
when the smaller dampers are used, and this results to lower
natural frequency of the damper. Ideally, the lower the natu-
ral frequency of the damped system, the better the isolation with regarding to Kyosho Zeal sheet performances.
damper provides for system with high disturbing frequency. Fig. 9 shows that the variations of vibration amplitudes
Despite Silicone Foam has the lowest natural frequency, of x-, y- and z-axis across different configurations of Kyosho
its high flexibility makes it sensitive to disturbances which Zeal sheets at 50 Hz. According to the results, experiment
worsens its performance. Overall, Kyosho Zeal Sheet has the 4 has the best result in attenuating the vibration signals on T-
best performance among the selections. lion UAV as the vibration amplitude is the lowest at 0.0021 G,
Its effect can be further proven by ADIS IMU testing as 0.0032 G, and 0.005 G, comparing to vibration amplitude
discussed earlier. Fig. 8 shows that almost all magnitudes of without any dampers at 0.08 G, 0.13 G and 0.17 G.
vibrations are reduced after a Kyosho Zeal sheet is installed In general, the following guidelines for damper installa-
to damp the sensor. Noted that for the vibrations at higher tion provide the best result for UAV vibration reduction:
frequencies, the reduction is up to 96.8%. Also, as expected,
motor noise at around 50 Hz is magnified as suggested in the 1. More contact surface area of the damper is preferred;
transmissibility curve shown previously. However, this does 2. Kyosho Zeal sheet should be installed vertically;
not affect its effectiveness in minimizing vibrations at higher
frequencies. 3. Even Length-Width-Height (LWH) ratio (≈ 1 : 1 : 1)
gives the best results; and
5 F LIGHT E XPERIMENTS
As Kyosho Zeal sheet was found to be best suited to 4. Damper with smaller dimensions performs better for
damp the on-board sensors of weight 30 g each (120 g in light mass damping.
total) from the previous section, the on-board system which
5.2 Digital Low-Pass-Filter Design
includes IMU sensors were mounted on Kyosho Zeal sheet
for actual UAV implementations on our in-house quad-copter Actual flight experiment results show significant reduc-
codenamed T-Lion. In this section, more dimensions and dif- tions in the vibration amplitudes (up to 97.1% reduction along
ferent installation methods of Kyosho Zeal sheet were used the z-axis) with both damper and low-pass filter (cut-off fre-
and its effect on vibration reduction is studied. quency: 10 Hz) implemented. The illustration below (Fig. 10)
shows that vibration signals could be amplified with damper
5.1 Hardware Damper Design alone, but the combined effect of both low-pass filter and
Flight tests are conducted with the different configura- damper will shift the system into isolation region, and this
tions in Table 1 aiming to verify the effects of each parameter justifies the experiment results.
 228 International Micro Air Vehicle Conference and Flight Competition (IMAV) 2017

20 ious masses and dimensions of the dampers in details, so that

15 Digital Low-pass Filter
to predict its general performance and the optimal working
Kyosho Zeal sheet damper range (in terms of damping mass and natural frequency) for
10 Combined filter
each damper.
Magnitude (dB)

5
Lastly, this research work can be extended to other impli-
0 cations such as UAVs with different configurations and heli-
-5 copter drones by making use of the frameworks and evalua-
-10 tion results provided in this paper.
-15 R EFERENCES
-20
0 50 100 150 200 250
[1] K. Nonami, F. Kendoul, S. Suzuki, W. Wang, and
Frequency (Hz) D. Nakazawa. Autonomous Flying Robots. Springer,
2010.
Figure 10: Combined frequency response of hardware
damper and digital filter [2] F. Wang, S. K. Phang, J. J. Ong, B. M. Chen, and T. H.
Lee. Design and construction methodology of an indoor
uav system with embedded vision. Control and Intelli-
Despite the high flexibility and softness of the cho- gent Systems, 40(1):22–32, 2012.
sen dampers, mechanical dampers show their limitations in
[3] S. K. Phang, F. Wang, K. Wang, S. Lai, and B. M. Chen.
damping light masses and at low frequency such as 50 Hz.
An effective method for autonomous localization and
On another hand, low-pass filter also has its limitation as
navigation in unknown indoor environment using mav.
its functional frequency range is much shorter than mechan-
In International Micro Air Vehicles Conference, 2015.
ical damper. Therefore, the use of low-pass filter is essential
to lower amplified signal due to dampers while dampers are [4] J. Langelaan and S. Rock. Navigation of small uavs
used to reject the high frequency vibrations due to the UAV operating in forests. In AIAA Guidance, Navigation and
structure. Control Conference, 2004.
6 C ONCLUSION [5] B. Sinopoli, M. Micheli, G. Donato, and J. T. Koo. Vi-
This manuscript elaborates in detail the essential theoreti- sion based navigation for an unmanned aerial vehicle.
cal frameworks when considering anti-vibration measures on In IEEE International Conference on Robotics and Au-
UAVs and how these can be used to improve on the current tomation, 2001.
measures. Building on from this, criteria of damper selec- [6] T. H. Lee G. Cai, B. M. Chen. Unmanned rotorcraft
tion for UAVs with light damping mass is established to act systems. Springer, London/New York, 2011.
as guidelines for future practices.
With close examinations on vibration sources generated [7] V. Gavrilets, A. Shterenberg, M. A. Dahleh, and
on T-lion UAV through both simulation and experimental re- E. Feron. Avionics system for a small unmanned he-
sults, it is shown that structural vibration contributes much licopter performing aggressive maneuvers. In Digital
higher vibration amplitude at higher frequencies comparing Avionics Systems Conference, 2000.
to motor vibration at lower frequency, addressing the impor-
[8] A. Imam and R. Bicker. Design and construction of a
tance of considering vibration attenuation for structural vi-
small-scale rotorcraft uav system. International Jour-
bration when designing anti-vibration measures. Therefore,
nal of Engineering Science and Innovative Technology,
the relationships between different parameters of dampers
3(1):96–109, 2014.
and their damping effectiveness are closely evaluated through
vibration tests, and the best performer, Kyosho Zeal sheet, [9] G. Cai, F. Lin, B. M. Chen, and T. H. Lee. System-
further proves its effectiveness in minimizing structural vi- atic design methodology and construction of uav heli-
brations at higher frequency, despite of creating slight signal copters. Mechatronics, 18(10):545–558, 2008.
amplification for motor vibration. This also indicates the lim-
itation of mechanical damper in damping light mass at low [10] R. Simmons. Vibration isolation. ASHRAE Journal,
disturbing frequency. Nonetheless, with both digital low- pages 30–40, 2007.
pass filter and Kyosho Zeal sheet, actual flight test shows up [11] H. Fonseca. Displacement and force transmissibility in
to 97.1% improvement on lowering the vibration amplitude. structures and multilayer supports with applications to
This addresses the significance of implementing both digital vibration isolation. Master’s thesis, Universidade Tec-
filter and mechanical damper for effective damping on UAVs. nica De Lisboa Instituto Superior Tecnico, Lisbon, Por-
Further studies can be done on vibration tests to investi- tugal, 2011.
gate the pattern of how different parameters change with var-