Received December 18, 2019, accepted January 15, 2020, date of publication February 11, 2020, date of current

version February 18, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.2971711

Ocean Wave Buoy Based on Parallel

Six-Dimensional Accelerometer
YUNPING LIU 1, XIN WANG 1, JINGJING YOU 2, AND CHENG CHEN 3
1 Jiangsu Province Atmospheric Environment and Equipment Technology Collaborative Innovation Center, Nanjing University of Information Science and
Technology, Nanjing 210044, China
2 College of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing 210037, China
3 Nanjing Automation Institute of Water Conservancy and Hydrology, Nanjing 210012, China

Corresponding authors: Yunping Liu (liuyunping@nuiste.edu.cn) and Jingjing You (youjingjing251010@126.com)

This work was supported in part by the Natural Science Foundation of China under Grant 51875293, and in part by the National Key
Research and Development Program of China under Grant 2018YFC1405703.

ABSTRACT Ocean wave parameters, including wave height, direction and period, are necessary to improve
forecasts of ocean wave conditions. Their accurate observation contributes to safety of both offshore
operations and military operations. However, conventional wave observation methods that combine the
three-axis accelerometer-based wave sensor with the gyroscope may show problems such as inner arm error,
transverse sensitivity effect, and gyroscope drift. The first two are due to non-coincidence of centroids inside
the accelerometer. Such problems will reduce reliability of data. To solve the problems, a new type of parallel
six-dimensional accelerometer is developed to reduce principle errors and improve stability of wave buoys.
The solution features a single mass and 12-chain redundancy. When compared with the performance of high-
precision sensors in the wave pool, the wave buoy based on six-dimensional accelerometer is proved to be
more accurate in wave height measurement than 3.75 % × measurement value, with wave direction accuracy
of ± 1◦ , and the wave period measurement accuracy of ± 0.1s.

INDEX TERMS Ocean wave sensor, parallel structure, high precision, wave buoys.

I. INTRODUCTION space-time range and has a wide range of observation periods.

Ocean wave characteristics, including wave height, direc- Since its appearance, it has been the main mean of ocean wave
tion and period, are vital for ocean development, waterborne observation [11].
transport, national defense and marine disaster prevention, There are two current methods of buoy wave measurement
so it is highly valued by marine scientific research [1]. There based on accelerometers, one is based on gravity accelerom-
are many kinds of wave observation methods, such as manual eter, and the other is based on three-axis accelerometer [12].
measurement, instrumental measurement and remote sensing The first method uses a vertical accelerometer and a tilt
inversion measurement [2], but each method has its limi- sensor to measure the accelerations in the vertical direction
tations. The accuracy of manual measurement is general, and the azimuth angle of the buoy tilt respectively. After
the continuity is poor, and its observation results have human data processing, the wave height, wave period, and wave
errors. Instrumental measurement methods, such as the wave direction are obtained. Representative products include the
measuring rod [3], pressure-type [4], [5] and acoustic wave US ENDECO/YSI 1156 wave direction trajectory buoy [13],
measurement [6], [7], cannot measure the wave direction, and the Dutch Datawell company’s wave knight Directional
are easy to be affected by environmental factors and measur- Waverider MKIII [14], [15], the Norwegian Fugrooeeanor
ing devices. The accuracy of remote sensing inversion (such wavescan buoy [16] and the Chinese SZF wave buoy [17],
as radar [8] and satellite altimeter measurement [9], [10]) is [18]. The Directional waverider MKIII measures the wave
poor and the devices are very expensive. However, the wave height by an accelerometer and wave direction by combining
buoy using accelerometers in this paper has high accuracy in the horizontal accelerometer with an electronic compass, thus
observation methods [2]. Moreover, it is not affected by the forming a complete sensor unit.
The second method estimates the wave characteristics
The associate editor coordinating the review of this manuscript and by measuring the accelerations on three axes and rota-
approving it for publication was Usama Mir . tion angles (heading angle, pitch angle and roll angle)

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
VOLUME 8, 2020 29627
 Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

of the buoy, such as the AXYS TRIAXYS Directional and Rockwell International Corporation jointly designed an
Wave Buoy [19], [20]. TRIAXYS Directional Wave Buoy inertial sensor including three micro gyroscopes and three
consists of three accelerometers (three sensors are placed micro accelerometers [26]. Wang and Yuan [27] proposed a
orthogonal), three gyroscopes, and an electronic compass. six-dimensional accelerometer based on a coplanar combi-
It provides continuous wave sampling of six parameters nation of six-degree-of-freedom accelerometers. Wang and
(X, Y, Z-axis acceleration and dip), under the motion of 8 Hz. Yuan [28] proposed a ‘‘low cost, gyro-free, simple mass’’
A PC-104 microprocessor in the buoy could take samples, accelerometer. Li et al. [29] and You [30] designed a six-
process data, and store information at the same time. Pro- dimensional accelerometer based on 9-chain, but this scheme
cessed wave data files are logged onto a PCMCIA Flash had the disadvantages of output signals distortion and lack
RAM. The on-board data processor uses a special itera- of real-time due to structural limitations and generalized
tive algorithm developed by the Canadian Hydraulics Centre momentum. Six-dimensional accelerometers have developed
(CHC) of the National Research Council of Canada (NRC) rapidly, but most of them have narrow operating frequency
for time-domain analysis and frequency-domain spectrum band and low precision [31], which cannot meet the require-
analysis. The values of wave height, average wave height, ments of measuring waves (ocean waves have strong ran-
maximum wave height, effective wave period, effective wave domness in frequency and are mostly low-frequency waves
height at spectrum analysis m0 (m0 is the zero moment near the coast). Therefore, there is no report on applying six-
of wave spectrum) time, spectral analysis of all frequen- dimensional accelerometers to wave buoys currently.
cies (mean wave periods), spectral analysis of peak wave To deal with these problems, this paper has designed a
period and time-domain/frequency-domain parameters could new 12-chain parallel accelerometer for the monitoring of
be well obtained. Compared with the wave measurement of wave accelerations. The main advantages of applying the six-
a single accelerometer, the combination of three accelerom- dimensional accelerometer on wave buoy are as follows:
eters and gyroscopes can better reflect the movement track (1) In this paper, the parallel six-dimensional accelerome-
of the buoy in three-dimensional space, and can measure ter is used as the acquisition unit. It is designed as a tetrahedral
the inclination angles to get more accurate displacement redundant parallel structure with single mass. The design has
information [21]. a compact topological structure, good decoupling property,
However, the accuracy of three single-axis accelerometers wide operating frequency band, and low processing cost,
in conventional wave buoys was often affected due to uneven avoids the lever arm effect and lateral sensitivity effect of the
positions (X, Y, and Z planes) and inconsistent starting points combination of three accelerometers. Besides, it improves the
of the units. In addition, lever-arm effect and lateral sen- measurement accuracy.
sitivity error also increased the measurement errors, which (2) According to the piezoelectric theory, the deformations
inevitably accumulated and magnified after two integrations. of 12 branches are inversed and calculated. Next, the motion
Gyroscope drift also diminished the accuracy. It is necessary amount of the inertial mass relative to the shell is calculated
to improve the wave sensor unit. Early studies focused on by using the kinematics theory of the parallel structure, and
three-dimensional linear accelerometers. K. Okada devel- the motion amount of the mass relative to the inertial coor-
oped a piezoelectric tri-axial accelerometer [22], and suc- dinate system is deduced. Finally, six parameters (three line
cessfully measured three-dimensional linear accelerations. accelerations and three angular accelerations on the X, Y, and
The 3355 three-dimensional accelerometer [23] designed by Z axes) are calculated by dynamic modeling and decoupling
EG&G placed three independent sensor units in three sen- algorithm. The decoupling algorithm has the characteristics
sitive directions orthogonally. Its principle was similar to of strong universality, high precision and good real-time
that of the TRIAXYS Wave Buoy, and measurement errors performance.
occurred due to the inconsistency of centroids of three sen- (3) The sensor adopts a gyroscope-free design, which
sors. Takao et al. [24] designed a micromechanical, piezore- avoids the problem of zero drift error when the conventional
sistive three-dimensional accelerometer, using an integrated buoy uses gyroscopes.
multidimensional sensor scheme. By sharing a mass in each The parallel six-dimensional accelerometer in this paper
sensitive direction of the accelerometer, both sensitivity and could well replace the combination of a gravity accelerom-
accuracy were improved, but its dimensional coupling degree eter and a tilt sensor or the combination of three single
was large. accelerometer and gyroscopes in conventional wave buoys.
There was little research on angular accelerometers. After introducing the wave feature algorithm, high-precision
With the assistance of Teledyne Company, JAXA (Japan and high-stability wave feature monitoring are achieved.
Aerospace Exploration Agency) developed a three-axis angu-
lar accelerometer [25], which successfully detected three II. OVERALL DESIGN OF SIX-DIMENSIONAL
angular accelerations of aircraft.In order to solve the prob- ACCELEROMETER-BASED WAVE BUOY
lems such as large volume, high cost, and low precision A wave buoy has three components: the mechanical structure,
in measuring six-dimensional accelerations by using the the sensor system, and the system control module. 1) The
‘‘accelerometers + gyroscopes’’ method, six-dimensional stainless steel spherical buoy is equipped with lead-acid bat-
accelerometers began to develop. For example, Draper Lab teries, a control box (at the gravity center of the buoy control

29628 VOLUME 8, 2020

Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

FIGURE 1. System diagram.

cabin, with a parallel six-dimensional accelerometer, a signal

processor, a system controller, and other peripheral devices),
and navigation lights. 2) The sensor system consists of a par-
allel piezoelectric six-dimensional accelerometer and a signal
processor, collecting wave acceleration signals. 3) The sys-
tem controller includes an FPGA processor, an A/D converter,
a GPS module, a GPRS module, and a SD card to process,
transmit, and save wave data. The buoy is also equipped
with an electronic compass to collect the wave direction. The
system diagram is shown in Fig.1.
The parallel six-dimensional accelerometer inside the buoy
body moves with the wave. It exports wave-acceleration-
related charge signals. These signals, which will be processed
into effective voltage signals by the signal processor. Then the FIGURE 2. General flow chart of wave statistical algorithm.

system uses dynamic modeling techniques and decoupling

algorithm to obtain effective accelerations. The accelerations
are integrated into speeds signals, which will be removed
of the trend item and the effective speeds are obtained. The
effective speeds are then integrated into the displacement
signals, which will be removed of trend term. Finally, we have
the effective displacement data.
Wave period and height can be obtained at the same time by
applying zero-crossing process to Z-axis displacement. After
the electronic compass gets angles, Kalman filtering is used
to compute the main wave direction by merging the above-
mentioned data with the acceleration sequences of the X and
Y axes. The whole procedure is shown in the flow chart of FIGURE 3. Digital prototype of Six-dimensional accelerometer.
Fig. 2.

III. HARDWARE DESIGN OF SENSOR SYSTEM of each branch is a piezoelectric ceramic, and the two ends
A. DESIGN OF ACCELEROMETER are elastic ball joints. The accelerometer has the advantages
The buoy system adopts a redundancy parallel structure of large measuring range, high sensitivity, high precision and
with tetrahedral geometry of 12-SPS (12 is the number wide working frequency bandwidth.
of branches, S represents the ball pair, and P the mobile The accelerometer consists of a mass, elastic ball hinges,
pair). This structure is to form elastomer of the accelerom- piezoelectric ceramics, pre-tightening columns, preload
eter [29], [30]. The digital and physical prototype of this plates, sub boards, and a shell. And it has one inertial mass
new six-dimensional accelerometer is shown in Fig.3 and and twelve identical branches. Each branch is formed by
Fig.4, respectively. The six-dimensional accelerometer con- connecting a cylindrical piezoelectric ceramic and two elastic
tains three parts, including 1) a mass, 2) a box-shaped outer ball hinges located at both ends in series. The connection
casing (including four side plates, a bottom plate, and a top between the piezoelectric ceramics and the elastic ball hinges
beam, all made of duralumin) and 3) 12 branches. The middle is by means of a conductive film and AB glue. One side of a

VOLUME 8, 2020 29629

 Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

FIGURE 5. Structure of the signal processor.

with a bottom circle diameter of 5 mm and a thickness

of 5 mm.
The accelerometer is fixed at the center of the buoy control
box and the box is rigidly fixed at the center of the buoy
FIGURE 4. Physical prototype of six-dimensional accelerometer. shell. When the accelerometer senses the movement of the
buoy, it will move with the ocean waves. The 12 branches
will be compressed or stretched along with the mass under the
bare copper sheet of the conductive film is bonded to one elec-
action of inertial forces, and the piezoelectric ceramics will
trode of the piezoelectric ceramic with AB glue, and the other
be subjected to corresponding axial forces [32], [33], which
side is bonded to the elastic ball hinge with AB glue. A wire
trigger the positive piezoelectric effect of the 12 ceramics and
is welded to the exposed copper sheet at the other end of the
produce polarized charges. The electric charge (its amount
conductive film (the wire is connected to the signal processor
and polarization) generated by each piezoelectric ceramic is
through the M5 interface), so that the charges generated on
proportional to the axial force it receives, and the propor-
the polarization surface of the piezoelectric ceramic can be
tional coefficient is the piezoelectric constant of the ceramic.
derived. Each end of the piezoelectric ceramic is connected
According to the piezoelectric theory, the deformations of the
with an elastic ball hinge. The branches are equally divided
branches are deduced. Moreover, the motion amount of the
into 6 groups (each two branched chains are one group),
mass relative to the shell is calculated by the kinematics
and the groups are vertically arranged with each other and
theory of the parallel structure. Therefore, the motion amount
share one composite ball hinge. Six composite ball hinges are
of the mass relative to the inertial reference system is cal-
respectively fixed on the midpoints of the upper rear edge,
culated. Finally, the accelerations collected by the parallel
the upper left edge, the right rear edge, the lower front edge,
six-dimensional accelerometer are obtained. At this time,
the lower right edge, and the left front edge of the inertia
the obtained accelerations are the accelerations of the buoy,
mass. The other end of each branch is fixed on the sub board
which are the accelerations of the ocean wave.
through an elastic ball hinge by using a positioning nut. Six
sub boards are fixed by adjusting the pre-tightening columns
(in the center of the boards) and the preload plates, so that B. DESIGN OF SIGNAL PROCESSOR
the pre-tightening and anti-loosening of each branch chain In order to reduce the influence of the accelerometer itself
are achieved. In addition, there are four through holes on the and external interference, a signal processor is designed.
bottom plate, which make it easy to connect the accelerometer The signal processor processes charge signals output by the
and the buoy control box firmly. accelerometer to realize charge-to-voltage, reduce interfer-
The specific parameters of the sensor are as follows: ence, and amplify the signals.
Performance parameters: the working frequency is The signal processor includes a charge converter, two
0.1 Hz–100 Hz, and the range of linear acceleration is −21.5g buffers, a filter, and an amplifier. Its inputs are charge signals
to +21.5g (g is local gravitational acceleration), which meets output by the parallel six-dimensional accelerometer, and its
the design requirements of ocean wave buoys. Specific outputs are voltage signals. The charge converter converts
dimensions: the side length of the cube inertial mass is charge signals into voltage signals. The buffers coordinate,
60 mm, the length of the inner large hinge is 55 mm, the length buffer, and isolate signals to improve signal quality. The
of the inner small hinge is 64 mm, the length of the outer hinge filter removes the interference signals and retains the required
is 20 mm, the side length of the outer shell is 146 mm, and signals. And the amplifier amplifies the signals and ensure
the total mass is 5.15 kg. In order to meet the requirements of that the range of the signals output by signal processor
rigidity and housing, the mass and pre-tightening screws are meets design requirements and prevent signal distortion. The
all made of ordinary steel, the shell is made of hard aluminum, schematic is shown in Fig. 5.
and the elastic ball hinges are all made of elastic steel 65Mn.
The flexible part of the elastic ball hinge is designed as IV. WAVE STATISTICAL ALGORITHM
an arc shape, in which the arc radius is 3 mm, the wide The voltage signals output from the signal processor will be
angle is 180◦ , and the minimum thickness is 1 mm. The processed by the FPGA processor. After applying dynamics
selection of piezoelectric ceramics is PZT-5, P-51 series of modeling and decoupling algorithm to the signals, effective
lead zirconate titanate (PZT ceramics), which is cylindrical, accelerations are obtained. The accelerations are integrated

29630 VOLUME 8, 2020

Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

and the pedestal (shell) of the six-dimensional accelerometer,

respectively. Therefore, the kinematics feature of the parallel
structure is an important standard for an appropriate elastic
structure of the six-dimensional accelerometer.
The inputs of the parallel six-dimensional accelerometer
are six independent components about the shell accelerations,
and the outputs are the axial forces of the 12 branches. All out-
puts will change together with every input. Therefore, the
accelerometer is a multi-input, multi-output, non-linear, and
strongly coupled system. ‘‘Forward dynamics solution’’ is to
deduce outputs from inputs via dynamic equations. On the
contrary, ‘‘reverse dynamics solution’’ is to extrapolate inputs
from outputs, also known as ‘‘decoupling’’. This paper uses
the motion of the mass relative to the shell as a ‘‘bridge’’ to
build a mathematical relation between six inputs and twelve
outputs of the six-dimensional accelerometer. It has achieved
full decoupling of the six-dimensional acceleration via the
dynamic equations.
As lightweight piezoelectric ceramics have a very good
rigidity and both ends are connected to the spherical pairs,
the 12 branches can be seen as a two-force bar without defor-
mation. Thus, the forces acting on the mass block include the
axial force fi (i is the branch number) and gravitational force
mg, where m is the mass of the mass block and g is the local
gravitational acceleration.
FIGURE 6. Detailed flow chart of wave statistical algorithm. Here, the pre-matrix and post-matrix are in the form of:

s0 −s3 s2 s1
 
into speeds, which will be removed of the trend item to cal- + −
 s
3 s0 −s1 s2 
. (1)
culate the effective speeds. The effective speeds are integrated S =
 −s s
to obtain the displacements. The displacement signals are 2 1 s0 s3 
removed of trend item again to get effective displacement −s1 −s2 −s3 s0
 
data. s0 s3 −s2 s1
The wave period and wave height can be obtained simul- −  −s3 s0 s1 s2 
S = . (2)
taneously by applying zero-crossing processing to Z-axis  s2 −s1 s0 s3 
displacement sequences. Kalman filtering is used to compute −s1 −s2 −s3 s0
the main wave direction by merging the angles obtained by
the electronic compass and the acceleration sequences of the The elements s0 , s1 , s2 , s3 are real numbers, and the column
X and Y axes. The flow chart of the algorithm is shown vectors consisting of elements s0 , s1 , s2 , s3 are recorded as S.
in Fig. 6. Wherein, matrix with positive superscript (‘‘+’’) is the pre-
matrix, and that with negative superscript (‘‘−’’) is the post-
A. ACCELERATION DECOUPLING matrix.
During the study of parallel six-dimensional accelerometers, The conversion relationship between the angular velocity
it is found that decoupling algorithm is crucial to the redun- vector ω and the attitude matrix R is:
dancy parallel 12-SPS structure with tetrahedral geometry.
ω̂ 0
 
Therefore, how to improve the efficiency and adaptability
Ṙ = R. (3)
of the algorithm is the core problem for the study. As the 0T 0
topological configuration of the sensor’s elastomer deter-
mines the decoupling algorithm, designing a suitable parallel − + +
structure to be the elastomer will be the best option. The basic where, R = (3)T 3, 3 is the pre-quaternion matrix and
−
decoupling principle of the six-dimensional accelerometer is 3 is post-quaternion matrix. They are matrices where the
to estimate the accelerations of the shell relative to its inertial values of the elements of equation (1) and equation (2) are
frame of reference. The estimation is based on motion param- the real part q0 and imaginary parts q1 , q2 , q3 of quaternions
eters of the mass relative to the shell (including displace- respectively.
ments, velocities, and accelerations). The upper and lower Two dynamic formulas of the parallel six-dimensional
platforms of the parallel structure correspond to the mass acceleration sensor system are derived by Newton-Euler

VOLUME 8, 2020 29631

 Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

method: Substituting the above formula into formula (4), the atti-
tude matrix and formula (9) are arranged to obtain:
   
  f1 + f3 − f7 − f9 g  
a 1  −f4 + f 6 + f 10 − f 12 
  0 f5 − f6 + f11 − f12
= R +  (4)
0 m −f 2 − f5 + f 8 + f 11 0 3  f1 − f2 + f7 − f8 .
ω̇∗ =
   
(16)
0 0 2nm −f3 + f4 − f9 + f10 

0
 
f5 − f6 + f11 − f12
ε
 
=
3
R
 f1 − f2 + f7 − f8 
. (5) As ω̇∗ has a unique certain value at each sampling instant,
0 2mn  −f3 + f4 + f9 + f10  it can be replaced by V : ω̇∗ = V .
0 Thus, two first-order ordinary differential equations are
In above formulas, n is the half-length of the mass edge, derived via the trapezoidal method for the dynamic equations.
a is the three-dimensional linear acceleration vector, and ε is h
ω(N
∗
) = ω(N −1) +
∗
V(N −1) + V(N ) .


(17)
the three-dimensional angular acceleration vector. 2
ω(N −1) h ∗ ω(N )
 ∗   ∗ 
Define h ∗
3(N ) = 3(N −1) + 3 + 3 . (18)
 T 4 (N −1) 0 4 (N ) 0
W = ωT , 0 . (6)
In above formulas, h is the acquisition step size of the
After combining the above-mentioned formulas (1) (2) (3) output. The right subscripts (N ) and (N − 1) respectively
(6), it is obtained: indicate the times N and N − 1.
 T
+  T + Since most carriers are changed from static to dynamic,
+ − T + − − T +
   
• + − • the initial amount of auxiliary angular velocity is zero vector
2 3
  3 +2 3 3 = W 3 3−W 3 3 . (7)
in general. In addition, in the initial state, the carrier coordi-
nate system coincides with the inertial coordinate system, so:
The multiplication of the pre-matrix and the post-matrix
3(0) = (0, 0, 0, 1)T . (19)
satisfies the commutative law:
+
+− −+ Premultiply the above formula by 3 )T to obtain:
SS =SS. (8) (N )

In addition, the pre-four-element matrix and the post-four- T h ∗ !

− ω(N

+
element matrix also satisfy: 3(N ) a(N −1) = 4 ) . (20)
1
+ + T
 T
− − T
     T
+ + − −
3 3 = 3 3=3 3 = 3 3 = E4 . (9) ω∗
 
+
In the formula, a = (α1 , α2 , α3 , α0 )T = 3 +3 h
.
4
 T 0
+− + − Swap the two terms on the left of equation (20) to get:
3S 3 = S. (10)
h ∗ !
− ω(N
In formula(9), E4 is a fourth-order identity matrix. Q(N −1) 3(N ) = 4 ) . (21)
According to formulas (8) (9) (10), formula (7) is calcu- 1
−α0 −α3 α2 α1
 
lated as:
 T  α3 −α0 −α1 α2 
ω
 
− • Of which, Q = 
= 2 3 3.  −α2 α1 −α0 α3 .Its determinant is

(11)
0
α1 α2 α3 α0
And calculated as follows:
ω ω∗
     2
=R . (12) |Q| = − aT a . (22)
0 0
Decomposing formula (20) shows that a is not a zero
Therefore, the following formula is derived:
vector. And matrix Q is nonsingular. Therefore, the nonho-
1 ∗ ω∗
 
• mogeneous linear equations of equation (21) have and only
3= 3 . (13)
2 0 have one solution, as shown in equation (23).
The first-order derivation of formula (12) is combined with h ∗ !
− ω(N
 
1 T
formula (5) to obtain: 3(N ) = Q 4 ) . (23)
aT a N −1 1
ω̇ ω̂ 0 ω ω̇
     ∗  ∗
= R +R . (14) The attitude matrix of the carrier relative to the inertial
0 0T 0 0 0
system is derived by substituting the formula above into the
Since ε = ω̇, and according to formulas (9) and (14), it is attitude matrix. The attitude matrix and the axial forces are
obtained that: substituted into equations (4) and (5) to obtain a unique
ε ω̇ 0 ω ω̇ ω̇ array {a, ε}. So far, the full decoupling of the six-dimensional
      ∗  ∗
= +R =R . (15)
0 0T 0 0 0 0 acceleration is completed [34]–[36].

29632 VOLUME 8, 2020

Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

" n−1 n−1 n−1

#
B. ACCELERATION-DISPLACEMENT X X X
= vi , vi ti , · · · , vi tim . (30)
INTEGRAL ALGORITHM
i=0 i=0 i=0
After the decoupling, the effective accelerations are obtained.
The next step is to acquire effective displacement sequence. Primary trend item is:
This can be done through integrations and the removal of the 1
X
trend item. It is based on the integral relationship of acceler- f1 (t) = pk t k = p1 t + p0 . (31)
ation, velocity, and displacement. Compound trapezoid for- k=0
mula is adopted to perform the integration, and the trend
Of which,
item is removed to reduce errors during the acceleration-
displacement integration [37], [38]. t = i/fs i = 0, 1, 2, fc, · · · , n − 1. (32)
The acceleration-displacement integration is:
Define the acceleration signals in the z-axis direction as To:
a (j) (j = 1, 2, · · · , n), then the speed signals are: n−1
 
P #T
 n ti    "Xn−1 n−1
v (i) = v (i − 1) + [a (i) + a (i + 1)] /2fs  p0 =
X

 n−1
i=0
n−1  p1 vi , (i/fs ) vi . (33)
i = 1, 2, · · · , n − 1; v0 = 0. ti2
P P
(24) ti i=0 i=0
i=0 i=0
Which contain primary trend items ε(i/fs ) + δ + v0 .
The displacement signals are: To:
n−1 n−1 n−1 n−1
s (k) = s (k − 1) + [v (k) + v (k + 1)] /2fs (i/fs )2 (i/fs )
P P P P
vi − vi
i=0 i=0 i=0 i=0
k = 1, 2, · · · , n − 1; s0 = 0. (25) p0 = " #2
n−1 n−1
Which contain quadratic trend items: (i/fs ) −2
(i/fs )
P P
n
i=0 i=0
ε(i/fs ) + (δ + v0 ) (i/fs ) + (η + s0 ) .
2
(26)
n−1 n−1 n−1
(i/fs )
P P P
In equations (17) and (18), v0 is the initial velocity and n vi − vi
i=0 i=0 i=0
s0 is the initial displacement. In addition, parameter ε is p1 = " #2 . (34)
the DC component generated by the accelerometer to collect n−1 n−1
(i/fs )2 − (i/fs )
P P
the acceleration signals. Parameters δ and η are constants n
i=0 i=0
produced by parameter ε after the primary and secondary
integration. Similarly, the quadratic trend item can be obtained.
Define fitting polynomials as: 2
X
m
X f2 (t) = pk t k = q2 t 2 + q1 t + q0 . (35)
k
fm (t) = pk t ∈ ∅. (27) k=0
k=0
The effective accelerations are v (i) − f1 (t), and the effec-
where, coefficients pk are the polynomial coefficients; ∅ is tive displacements are s (k) − f2 (t).
a set of functions consisting of polynomials less than m(m ≤
n−1)times. The calculation of coefficients pk needs to satisfy C. WAVE STATISTICS
the following formula. Wave data mainly includes wave height, direction, and
n−1 n−1 m h i 2 period. This system adopts the method of downward zero-
" #
X X X
I= [vi −fm (ti )] =
2
vi − k
pk ti = min . (28) crossing to derive wave statistics [39]. The main sensors are
i=0 i=0 i=0 a parallel piezoelectric six-dimensional accelerometer and an
Then: electronic compass.
n−1
" m h i 2
# The main goal is to count the wave height and period.
∂I X X
k j Firstly, the average wave height of ocean wave is set to be
=2 vi − pk ti ti = 0. (29)
∂pj zero-datum line. Secondly, continuously measure the data
i=0 i=0
and label the adjacent lower span zero points in the system.
To:
  Thirdly, calculate and record the wave height and period of
n−1 n−1
n
P
ti ···
P
tim  two adjacent lower zero-cross points. Data need to be stored
in a specified array. It should be noted that the elements of

i=0 i=0  
p1

 n−1 n−1 n−1
 P P 2 P m  wave height are arranged according to certain rules (such as
ti ti ··· ti   p2 

from small to large or the opposite). Finally, the maximum

 . 
 i=0 
i=0 i=0
  .. 
 . .. .. 

 . wave height Hmax , the average wave height Hmean , the effec-
 . . .  p
 n−1  m tive wave height H1/3 , and its corresponding wave periods are
 P m n−1 P m+1 n−1
P 2m  calculated. As for the wave direction, an electronic compass
ti ti ··· ti
i=0 i=0 i=0 is used to correct the data: firstly, the acceleration sequences

VOLUME 8, 2020 29633

 Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

TABLE 1. Technical parameters of wave sensor.

TABLE 2. Technical parameters of wave sensor.

from X and Y axes are used to calculate the direction, which

FIGURE 7. Acceleration test platform.
is further corrected based on results of the electronic com-
pass. Secondly, both data are merged by Kalman filtering.
In wave direction data fusion of the electronic compass and
the accelerometer, linear Kalman filter is used. The system
pool test. The design fully considered the demand of the
state variable is the wave direction. The deviation measured
experimental platform and the schedule. After the tests, the
by the accelerometer is used as t he control input, and the
measured data were analyzed to evaluate the function and
measured value of the electronic compass is used as the
performance of the system.
system observation variable. Thirdly, the direction values are
divided into 16 intervals and the frequencies of each direction
in each interval are calculated. The direction angle with the 1) ACCELEROMETER TEST
highest frequency is the main wave direction. An experimental platform is built to test the parallel piezo-
electric accelerometer. The details are shown in Fig.7. The
V. EXPERIMENTAL VERIFICATION platform uses a Simulation Transportation Vibration Tester to
A. TECHNICAL PARAMETERS simulate wave fluctuations. During the test, a parallel piezo-
According to the Seashore Observation Specification [40], electric six-dimensional accelerometer and Ellipse-n system
[41], a standard that specifies the items and technical were placed on the tester. The tester was connected to an
requirements for seashore hydro-meteorological observation SD1432C charge amplifier. The data was collected by the
of coastal and island observation stations, when the wind USB-4716-AE data acquisition card and entered into the host
speed or wave height reach a certain value (depending on the computer for processing.
situation of sea areas), wave observation should be intensified As can be seen from Fig. 8 (a), (b), and (c), the blue
to once in an hour. The buoy should be put in water deeper line in the chart is the output from the parallel piezoelectric
than 10 m, with flat ocean floor and without areas of rapid accelerometer, and the red dotted line is the output from the
waters. Ellipse-n system. It can be observed that the acceleration
Specifications for wave height, period, and direction signals from the two sensors are basically the same. It verifies
include: the effectiveness and accuracy of the accelerometer.
Wave height(m): ± 10% for level I and ± 15% for level II.
Wave direction(◦ ): ± 5◦ for level I and ± 10◦ for level II. 2) WAVE POOL TEST
Wave period(s): ± 0.5s. A wave pool is a pool in which there are artificially generated,
Hence, wave observation has been designed with following reasonably random waves, like those of the ocean. By com-
parameters: paring the performance of both our buoy and the TRIAXYS
MINI buoy in the wave pool, the accuracy and stability of
1) TECHNICAL PARAMETERS OF WAVE SENSOR newly designed buoy system have been proved. Fig. 9 shows
2) TECHNICAL PARAMETERS OF WAVE SENSOR the photos of this field-test.
Maximum wave height, 1/10 wave height, 1/3 wave height, According to the experiment, the average wave height,
average wave height; 1/10 wave height period, 1/3 wave maximum wave height, 1/10 wave height, and 1/3 wave
height period, average wave height period; and main wave height derived by the system are in line with the results form
direction. of the TRIAXYS wave buoy. Detailed waveforms are shown
in shown in Fig. 10(a)–(d). The system exported wave data
3) TECHNICAL PARAMETERS OF WAVE SENSOR every hour for 24 hours. The average wave height ranged
B. WAVE POOL EXPERIMENT from 0.15 m to 0.27 m, with the maximum wave height
To test the wave observation buoy, two experiments were of 0.7 m. The results of wave height from our system over-
designed, including 1) a laboratory test and 2) a wave lapped those from the TRIAXYS MINI buoy, as can be shown

29634 VOLUME 8, 2020

Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

FIGURE 8. (a) is the comparison of X-axis acceleration, (b) is the

comparison of Y-axis acceleration, and (c) is the comparison of Z-axis
acceleration.

FIGURE 9. Wave pool experiment.

in the curve below. Hence, the system fully meets the design
requirements in reality.
Fig. 11 shows the wave direction data observed by the
system and the TRIAXYS MINI buoy. The wave period
FIGURE 10. (a) is the average wave height, (b) is the maximum wave
curves observed by the two buoys are basically the same. The height, (c) is the 1/10 wave height, and (d) is the 1/3 wave height.
wave directions were concentrated from 100◦ to 120◦ , and the
direction with the largest frequency was about 112◦ . There-
fore, the main wave direction was the ESE (east-southeast). C. MARINE EXPERIMENT
The comparison of parameters of the system and conven- Our observation site was around 1.33 kilometers away
tional wave buoys are shown in Table 3. from the shore of Lianyungang Harbor, Jiangsu. Based on
To conclude, the ocean wave observation buoy can accu- the onsite prospect, the water of the observation site was
rately derive wave data during the observation period and it 10-meters-deep, with stable tide and silty sand, as can be
is able to successfully monitor the wave. shown in Fig. 12.

VOLUME 8, 2020 29635

 Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

FIGURE 11. Ocean wave direction.

TABLE 3. Comparison of ocean wave observation buoys and

conventional wave buoy parameters.

FIGURE 13. (a) is wave height in September 2017, (b) is wave height in
March 2018, and (c) is wave height in September 2018.

FIGURE 14. Monthly wave characteristic graph.

3) BASIC CHARACTERISTICS OF WAVES

From the one-year data observed, 3 months of wave height
data from the total were selected to be shown in Fig. 13.
Fig. 14 shows the wave characteristic.
FIGURE 12. Observation site.
4) ROSE DIAGRAM
The 0–0.5 m wave rose diagram in March and Decem-
1) METHOD ber of 2017, as well as March and September of 2018 are
The wave buoy was used as an instrument to observe the shown in Fig. 15 (a)–(d). First, the number of main-direction
ocean. The observation mode was ‘‘regular observation’’ waves was counted based on the total wave counts during the
(once an hour). The sampling interval was every 0.5 s. observation. Second, the numbers of waves in each direction
were calculated, and categorized according to wave heights.
2) DATA COLLECTION RATE Third, the frequency of waves in all directions within a certain
The data collection rate was 100%. It means that during the height range was computed. The results are shown in wave
one-year investigation, the data were fully collected except rose diagrams, in which the main wave direction overlaps the
during the replacement of the instrument. maximum wave.
The batteries of the instrument were changed for every
three months. During the one-year observation, there were 5) DATA ANALYSIS
28 groups of missing wave data. Since the missing waves Based on the analysis, more than 97.5% of the waves in the
were smaller than microwave, the influence is minimal. observation site were smaller or equal to level III (0.5–1.25m)

29636 VOLUME 8, 2020

Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

applied to obtain the average wave height, the maximum

wave height, the main wave direction, and other wave char-
acteristics. After comparing high-precision sensors through
experiments, we found that six-dimensional accelerometer
was more accurate in wave height measurement than the
3.75 % × measurement value. It had wave direction accu-
racy of ± 1◦ , and the wave period measurement accuracy
of ± 0.1s.
We are going to study marine environment monitoring in
which unmanned ships, buoys and underwater robots are used
together for the monitoring. The goal is to measure ocean
waves and obtain meteorological information as comprehen-
sively as possible.

REFERENCES
[1] T.-J. Chu, Chinese Marine Data Buoy. Beijing, China: Ocean Press, 2001.
[2] Q.-W. Zhou, S. Zhang, and H. Wu, ‘‘Overview of OceanWave observa-
tion technology,’’ Hydrographic Surveying Charting, vol. ED-36, no. 2,
pp. 39–44, Feb. 2016.
[3] S.-P. Zhang and J. Liang, ‘‘Development and application of hydrologi-
cal telemetry system,’’ Ocean Technol., vol. 24, no. 4, pp. 5–7 and 13,
Apr. 2005.
[4] S4 Current Meter User Manual, Inter Ocean Syst., San Diego, CA, USA,
1995.
[5] SBE 26plus Manual, Sea-Bird Electron., Washington, DC, USA, 1995.
[6] Teledyne RDI’s Workhorse Waves Array, Teledyne Technol.,
Thousand Oaks, CA, USA, 2013.
FIGURE 15. (a) is wave rose chart of September 2017, (b) is wave rose [7] B. Robertson, Y. Jin, H. Bailey, and B. Buckham, ‘‘Calibrating wave
chart of December 2017, (c) is wave rose chart of March 2018, and (d) is resource assessments through application of the triple collocation tech-
wave rose chart of September 2018. nique,’’ Renew. Energy, vol. 114, pp. 166–179, Dec. 2017.
[8] F.-Y. Wang, G.-N. Yuan, and Z.-Z. Lu, ‘‘Research progress of radar-
based wave measurement technology at home and abroad,’’ Sci. Surveying
Mapping, vol. ED-33, no. 4, pp. 18–20, Apr. 2008.
and there were more waves above level III in winter than in [9] G.-Y. Wang, J.-P. Zhao, and R.-S. Song, ‘‘Estimation of wave extreme
value using satellite microwave remote sensing data,’’ Port Eng. Technol.,
summer. vol. 37, no. 1, pp. 14–21, Jan. 2000.
The main direction of waves below 0.5m in September [10] J. Yang and J. Zhang, ‘‘Validation of sentinel-3A/3B satellite altimetry
and December of 2017 were respectively NE-ENE and wave heights with buoy and Jason-3 data,’’ Sensors, vol. 19, no. 13, p. 2914,
Jul. 2019.
NE-E; the waves of the same standard during March and [11] H. E. Krogstad, S. F. Barstow, S. E. Aasen, and I. Rodriguez, ‘‘Some
September of 2018 concentrated in NE-ENE and E, recent developments in wave buoy measurement technology,’’ Coastal
respectively. Eng., vol. 37, nos. 3–4, pp. 309–329, Aug. 1999.
[12] G.-Y. Tang and Q. Kang, ‘‘Comparison of wave measurement meth-
ods with wave buoy,’’ Modern Electron. Technol., vol. ED-37, no. 15,
VI. CONCLUSION pp. 121–122, 2014.
[13] C. V. Nelson and C. A. Keller, ‘‘Comparison between wave measurements
Conventional wave observation method that combines the using a wave spar buoy and an ENDECO type 1156, WAVETRACK buoy
three-axis accelerometer-based wave sensor with the gyro- for winds less than 10 M/S: Preliminary data results,’’ in Proc. Ocean
scope show problems such as inner arm error, transverse Technol. Opportunities Pacific (OCEANS), 1991, pp. 962–967.
[14] The Directional Waverider DWR-MkIII: Over Three Years Continuous
sensitivity effect and gyroscope drift. Such problems will
Operation, Datawell B.V., Haarlem, The Netherlands, 2012.
reduce data reliability of the system. This paper introduces [15] C.-C. Chen, J.-H. Zhu, and Y.-Q. Huang, ‘‘Reprocessing of wave knight
a new type of parallel six-dimensional accelerometer that data in the elevation effective wave height test,’’ Hydrographic Surveying
Charting, vol. ED-33, no. 3, pp. 53–55, May 2013.
could be applied to the design of wave buoys. With a single
[16] S. Barstow, G. Ueland, H. Krogstad, and B. Fossum, ‘‘The wavescan
mass and 12-chain redundancy added to the system, this kind second generation directional wave buoy,’’ IEEE J. Ocean. Eng., vol. 16,
of accelerometer reduces the principle error and improves no. 3, pp. 254–266, Jul. 1991.
stability and the method can be summarized as follows: [17] G.-Y. Tang and J.-P. Wang, ‘‘SZF wave buoy system,’’ Ocean Technol.,
vol. 27, no. 2, pp. 34–36, Feb. 2008.
Firstly, the six-dimensional accelerometer was used to [18] G.-Y. Tang, J.-Y. Zhou, and S.-W. L, ‘‘Development of 3M multiparameter
collect the electric charge data about wave accelerations. wave buoy,’’ Meteorolog., Hydrolog. Mar. Instrum., vol. 36, no. 2, pp. 5–9,
After the data had gone through the signal processor and Feb. 2013.
[19] TRIAXYS OEM Directional Wave Sensor USER’S MANUAL, AXYS Tech-
FPGA processor, effective accelerations were modeled and nol., Sidney, BC, Canada, 2011.
decoupled. Next, the acceleration-displacement integral algo- [20] S. Naeth, D. Shumuk, and T. Ethier, ‘‘TRIAXYS G3 directional wave
rithm, which was adopted polynomial fitting to elimi- sensor,’’ in Proc. OCEANS MTS/IEEE Monterey, Sep. 2016, pp. 1–9.
[21] J. Zhao, L. Hui, and S.-B. Chu, ‘‘Research on wave measurement technol-
nate the trend terms. Moreover, integration algorithm of ogy based on triaxial acceleration,’’ J. Ocean Technol., vol. ED-34, no. 5,
acceleration - displacement and zero-crossing method were pp. 66–70, May 2015.

VOLUME 8, 2020 29637

 Y. Liu et al.: Ocean Wave Buoy Based on Parallel Six-Dimensional Accelerometer

[22] J. Chae, H. Kulah, and K. Najafi, ‘‘A monolithic three-axis micro-g micro- YUNPING LIU received the Ph.D. degree from the
machined silicon capacitive accelerometer,’’ J. Microelectromech. Syst., Nanjing University of Aeronautics and Astronau-
vol. 14, no. 2, pp. 235–242, Apr. 2005. tics, in 2009. He is currently an Associate Profes-
[23] K. Okada, ‘‘Tri-axial Piezoelectric Accelerometer,’’ in Proc. Int. Solid- sor with the School of Information and Control,
State Sens. Actuators Conf. (TRANSDUCERS), Aug. 2005, pp. 566–569. Nanjing University of Information Science and
[24] H. Takao, Y. Matsumoto, and M. Ishida, ‘‘A monolithically integrated three Technology. He is also the author of more than
axial accelerometer using stress sensitive CMOS differential amplifiers,’’ 20 journal articles. His current research interests
in Proc. Int. Solid State Sens. Actuators Conf. (Transducers), Nov. 2002,
include robotics, sensors, and multibody systems.
pp. 1173–1176.
[25] B. Davis, ‘‘Using low-cost MEMS accelerometers and gyroscopes as
strapdown IMUs on rolling projectiles,’’ in Proc. IEEE Position Location
Navigat. Symp., Nov. 2002, pp. 594–601.
[26] Y. Suzuki and E.-B. Lin, ‘‘Current situation and trend of angular
accelerometer,’’ Inertial Navigat. Instrum., vol. 1, no. 3, pp. 1–9, 1991.
[27] D. H. Wang and G. Yuan, ‘‘A six-degree-of-freedom acceleration sensing
method based on six coplanar single-axis accelerometers,’’ IEEE Trans. XIN WANG was born in Chengdu, Sichuan,
Instrum. Meas., vol. 60, no. 4, pp. 1433–1442, Apr. 2011. China, in 1996. She received the B.S. degree in
[28] J. W. Song and C. G. Park, ‘‘Optimal configuration of redundant inertial electrical engineering and its automation from the
sensors considering lever arm effect,’’ IEEE Sensors J., vol. 16, no. 9, Nanjing University of Information Science and
pp. 3171–3180, May 2016. Technology, Jiangsu, China, in 2018, where she
[29] C.-G. Li, J. Chen, and J.-J. You, ‘‘Fault diagnosis and repair of paral-
is currently pursuing the M.S. degree in control
lel six-dimensional accelerometer,’’ J. Vib. Shock, vol. ED-36, no. 15,
engineering. Her research interests include wave
pp. 222–229, Jul. 2017.
[30] J.-J. You, ‘‘Research on piezoelectric six-dimensional acceleration sensor buoy, parallel mechanism, and acceleration sensor.
based on redundant parallel mechanism,’’ M.S. thesis, Dept. Mechatron.
Eng., NUAA Univ., Nanjing, China, 2013.
[31] J.-J. You, ‘‘Research on piezoelectric six-dimensional acceleration sensor
based on 6-SPS parallel mechanism,’’ Ph.D. dissertation, NFU, Tiddington,
U.K., 2010.
[32] H.-Q. Zhai, C.-L. Cheng, and D.-B. Tang, ‘‘Model design of six-
dimensional accelerometer,’’ Mechatronics, vol. ED-17, no. 10, pp. 71–74 JINGJING YOU received the M.S. and Ph.D.
and 81, Oct. 2011. degrees in mechanical and electrical engineering
[33] F.-L. Ni, P.-F. Lin, T. Zou, and D.-B. Tang, ‘‘Vibration suppression of from the Nanjing University of Aeronautics and
flexible joints of large manipulators based on six-dimensional acceleration
Astronautics, Nanjing, China, in 2010 and 2013,
sensor,’’ Aerosp. Control Appl., vol. ED-44, no. 5, pp. 10–16, Oct. 2018.
respectively.
[34] J.-J. You, C.-G. Li, and H.-T. Wu, ‘‘Hybrid decoupling and error self-
compensation algorithm for parallel six-dimensional acceleration sensor,’’ He is currently an Associate Professor with
Chin. J. Sci. Instrum., vol. ED-36, no. 10, pp. 2249–2257, Oct. 2015. Nanjing Forestry University. His research areas
[35] J.-J. You, C.-G. Li, and H.-T. Wu, ‘‘Hamiltonian dynamics study of par- include parallel robot mechanism and six-axis
allel six-dimensional accelerometer,’’ J. Mech. Eng, vol. ED-48, no. 15, accelerometer.
pp. 9–17, Aug. 2012.
[36] J.-J. You, Z.-Z. Fu, and C.-G. Li, ‘‘Decoupling parameter identification
and disturbance analysis of parallel six-dimensional accelerometer,’’ J. Vib.
Shock, vol. ED-38, no. 1, pp. 134–141, Dec. 2019.
[37] F.-H. Zhou, Q. Xie, and D.-X. Wang, ‘‘Application of EEMD in detrended
measurement in acceleration wave measurement technology,’’ Meteorolog.
Hydrolog. Mar. Instrum., vol. ED-34, no. 02, pp. 77–82, Dec. 2017. CHENG CHEN was born in 1991. He received
[38] G.-B. Wang, Y.-L. Liu, and X.-H. Jin, ‘‘Trend item processing based on the B.S. and M.S. degrees from the Nanjing Uni-
the principle of least squares and its implementation of MATLAB,’’ Non- versity of Information Science and Technology,
Ferrous Metall. Equip., vol. 19, no. 5, pp. 4–5, 2005. in 2014 and 2017, respectively. He is currently
[39] J. Zhao, ‘‘Wave measurement technology based on triaxial acceleration,’’ working with the Nanjing Automation Institute
J. Ocean Technol., vol. ED-34, no. 5, pp. 66–70, Oct. 2015. of Water Conservancy and Hydrology, Nanjing,
[40] Specification of Coastal Observations, State Ocean. Admin., Beijing, China. His research interests include the rotorcraft
China, Sci. Press, 1987. navigation, tracking, stability, and robot.
[41] Y.-L. Wu, ‘‘The new specification for seaside observations,’’ Mar. Fore-
casts, vol. 5, no. 1, pp. 68–69, 1988.

29638 VOLUME 8, 2020