Content uploaded by Ming Yang
Author content
All content in this area was uploaded by Ming Yang on Jul 27, 2023
Content may be subject to copyright.
9680 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023
Digital Twin-Driven Industrialization
Development of Underwater Gliders
Ming Yang , Yanhui Wang , Member, IEEE, Cheng Wang , Yan Liang , Shaoqiong Yang ,
Lidong Wang, and Shuxin Wang
Abstract—Underwater glider (UG) is the only marine
equipment in the field of ocean observation that can realize
unmanned autonomous, all-weather and wide-area contin-
uous 3-D observation. The rapid individualized design (RID)
and full lifecycle management (FLM) make two difficulties to
be solved in the industrialization development of UGs since
the UG system involves interdisciplinary multiphysics and
there exists multisource uncertainties in complex ocean
environment. With the application of new information tech-
nologies in industry, the above problems can be effectively
solved by the digital twin (DT)-driven methodology. In this
article, an architecture of DT-driven RID and FLM for UGs
is first put forward based on our previous theoretical re-
search and technique development. Then, the proposed
architecture of DT-driven methodology is researched in de-
tail in terms of its digital modeling, design optimization,
virtual verification and practical application. Finally, Petrel
developed by China is introduced, and a typical preliminary
application of DT-driven methodology is presented based
on petrel to verify its feasibility. The architecture proposed
in this article is also appropriate for other types of au-
tonomous underwater vehicles.
Index Terms—Digital twin (DT), full lifecycle management
(FLM), industrialization development, rapid individualized
design (RID), underwater glider (UG).
I. INTRODUCTION
THE advanced marine equipment plays an important role in
building a maritime power. The development of technology
Manuscript received 22 August 2022; revised 15 December 2022;
accepted 29 December 2022. Date of publication 3 January 2023; date
of current version 24 July 2023. This work was supported in part by
the National Key R&D Program of China, in part by the National Nat-
ural Science Foundation of China under Grant 51721003 and Grant
11902219, in part by China Postdoctoral Science Foundation under
Grant 2022TQ0233 and Grant 2022M722367, and in part by Aoshan
Talent Cultivation Program under Grant 2017ASTCP-OS05 and Grant
2017ASTCP-OE01 of QNLM. Paper no. TII-22-3587. (Corresponding
author: Yanhui Wang.)
Ming Yang, Yanhui Wang, Cheng Wang, Yan Liang, Shaoqiong Yang,
and Shuxin Wang are with the Key Laboratory of Mechanism Theory
and Equipment Design Ministry of Education, School of Mechanical
Engineering, Tianjin University, Tianjin 300350, China, and also with
the Joint Laboratory of Ocean Observing and Detection, Laoshan Labo-
ratory, Qingdao, Shandong 266237, China (e-mail: mingyang@tju.edu.
cn; yanhuiwang@tju.edu.cn; wangcheng01@tju.edu.cn; liangyan312@
tju.edu.cn; shaoqiongy@tju.edu.cn; shuxinw@tju.edu.cn).
Lidong Wang is with the Changshu Guorui Technology Company,
Changshu 215500, China (e-mail: kaerking@gmail.com).
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TII.2023.3233972.
Digital Object Identifier 10.1109/TII.2023.3233972
Fig. 1. UGs in ocean observations.
and equipment for three-dimensional (3-D) marine environment
observation is increasingly highlighted [1]. Thereinto, underwa-
ter glider (UG) [2] is the only marine equipment in the field of
ocean observation that can realize unmanned autonomous, all-
weather, and wide-area continuous 3-D observation, the design
optimization [3] and the practical application [4] of which have
become a research hotspot of major maritime powers.
As shown in Fig. 1, UGs can achieve vertical motion
in water columns by changing its buoyancy, and horizontal
motion through the hydrodynamic forces generated by wings,
and therefore move in a characteristic sawtooth pattern in the
vertical plane. With the continuous breakthroughs in subsurface
communication, automatic control, and computer technology,
UGs have developed rapidly and gradually become important
tools for contesting maritime rights. Due to their low noise, low
cost, remarkable endurance, small weight, and easy deployment,
various UGs have been widely applied in ocean environment ob-
servation and military target detection [5]. However, the design
and application of UGs still face some challenges, despite the
rapid development of ocean science research.
1) With the increasing requirements of diversified observa-
tion missions, individualized UGs with different design
indexes, such as operation depth, payload capability and
size, need to be rapidly designed for timely service.
However, many disciplines are concerned in the design of
UG system, such as fluid mechanics, elastic mechanics,
electromagnetism, hydro acoustics, vibration modality,
and system dynamics. There exist complex coupling re-
lationships among these disciplines and the simulations
required are time-consuming, which brings great chal-
lenges for rapid individualized design (RID) of UGs.
2) UGs are autonomous platforms traveling in the water
column beyond visual line of sight. Therefore, the full
1551-3203 © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: DIGITAL TWIN-DRIVEN INDUSTRIALIZATION DEVELOPMENT OF UNDERWATER GLIDERS 9681
lifecycle management (FLM), which involves operation
evaluation, flight parameter adjustment, and fault diag-
nosis, appears to be particularly important for improv-
ing performance or avoiding missing of the prototype.
However, due to their long-term operation in complex
ocean environment, UGs are susceptible to time-varying
multisource uncertainties, such as biofouling, ocean cur-
rent, corrosion, and seal failure. These uncertainties will
influence the sailing efficiency obviously, even resulting
in permanent sinking of the vehicle, which significantly
increases difficulty of practical application.
Nowadays, the new generation of information technologies
is increasingly applied in industry and manufacturing, mainly
including Industry 4.0 [6], Internet of Things [7], made in China
2025 [8], and deep learning [9]. The Internet of Things helps col-
lect various big data in the entire product lifecycle, the Internet
of Things is used to realize the data management and processing,
and the deep learning is used for data mining and analysis.
The era of big data-driven design, manufacturing, and service is
coming. However, the current research on product lifecycle data
mainly focuses on physical products rather than virtual models.
Besides, the data in product lifecycle are isolated, fragmented,
and stagnant, due to the lack of convergence between the phys-
ical and virtual spaces of products [10]. These problems lead to
low level of efficiency, intelligence, and sustainability in product
design, manufacturing, and application. Thus, some new meth-
ods are required to support RID and FLM of products, which
shall combine the physical data, virtual data, and connection
data of physical and virtual products. Digital twin (DT) [11],
[12] is an integrated multiphysics, multiscale, and probabilistic
simulation of complex product, which uses the best available
physical models, sensor updates, status data, etc., to mirror the
performance of its corresponding twin. DT is composed of phys-
ical factors, virtual factors, and data that tie physical and virtual
factors, and thus can realize the convergence between physical
and virtual spaces of products. Therefore, the DT of product
can be established in the design and manufacturing process and
get continuously evolved in the full lifecycle application, which
promotes more efficient, smart, and sustainable product design,
manufacturing, and service.
However, implementing the DT in design and application of
UGs still requires a conceptual theory and a sound technique
basis, including synchronizing the multisource data and design-
ing the complex UG system with multiphysics. To bridge the
above gaps, an architecture of DT-driven RID and FLM of UGs
is proposed in this article. A virtual digital model for RID of
UGs is first established, based on which the design optimization
of UGs can be performed to obtain an optimal design scheme
by importing the design indexes from a physical observation
mission. Then, the virtual verification of the design scheme is
carried out by importing some DT data from physical space.
Finally, the FLM can be realized by the fusion of multisource
physical data, such as design data, manufacturing data, trim data,
and navigation data. The main contributions and innovations of
this article can be summarized as follows.
1) An architecture of DT-driven methodology is first es-
tablished to promote industrialization development and
application of UGs in ocean science research.
2) The DT-driven RID method of UGs, involving digital
modeling, design optimization, and virtual verification, is
proposed to satisfy the increasingly diversified demands
of observation missions.
3) The DT-driven FLM method of UGs is presented to
realize the state monitoring, motion optimization, fault
diagnosis, etc., under multisource uncertainties.
4) A typical preliminary attempt (digital model-based design
optimization) is made to apply DT-driven methodology
to UGs, and its feasibility is illustrated.
The rest of this article is organized as follows. Section II
introduces the related works. Then, an architecture of DT-driven
RID and FLM of UGs is proposed in Section III, the details of
which are introduced in Section IV and Section Vrespectively.
Section VI introduces the application of DT methodology in the
Petrel UGs developed by China. Finally, Section VII concludes
this article.
II. RELATED WORKS
A. Concept and Application of DT
By far, several explanations and definitions of DT have
been proposed. Hochhalter et al. [13] thought that DT is a
life management and certification paradigm whereby models
and simulations consist of as-built vehicle state, as-experienced
loads and environments, and other vehicle-specific history to
enable high-fidelity modeling of individual aerospace vehicles
throughout their service lives. According to Reifsnider and Ma-
jumdar [14], DT integrates ultrahigh-fidelity simulation with an
on-board health management system, maintenance history, and
historical vehicle and fleet data to mirror the life of a specific
flying physical twin and enable significant gains in safety and
reliability. Glaessgen and Stargel [15] gave a general definition
of DT in 2012, which has been recognized and used by most
researchers till now. A DT is an integrated multiphysics, multi-
scale, probabilistic simulation of an as-built vehicle or system
that uses the best available physical models, sensor updates,
fleet history, etc., to mirror the life of its corresponding flying
twin. The DT is ultra-realistic and may consider one or more
important and interdependent vehicle subsystems, including air-
frame, propulsion, energy storage, life support, avionics, thermal
protection, etc.
According to above definitions, the following characteristics
of DT [10] are summarized as follows.
1) Real-Time Reflection: The virtual space is the real-time
reflection of the physical space, which can keep ultra-high
synchronization and fidelity with the physical space.
2) Interaction and Convergence: The multisource data in
the physical space and virtual space are interacted and
converged.
3) Self-Evolution: DT can update data in real-time, so that
virtual models can undergo continuous improvement by
comparing virtual space with physical space in parallel
[16].
By far, DT has been applied in many industrial fields and
demonstrated its great potential. Structural Science Center at
US Air Force Research Laboratory employed DT to establish
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
9682 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023
a realistic high-fidelity flight model and combine virtual model
data with physical data to make a more accurate fatigue life
prediction [16]. Hongxia et al. [17] effectively improved the
quality management ability of the aircraft final assembly pro-
cess by establishing a virtual model, correlating data models
of different domains to form a unified DT data source, and
applying data mining tools to locate quality problems accu-
rately. Heim et al. [18] used a DT model to aid in designing
a sufficiently stable supply chain and maintenance strategy
of aircraft. Bondarenko and Fukuda [19] described a DT that
combines continuous time-domain cycle-mean value engine
model of ship with the crank-angle resolved phenomenological
combustion model, satisfying the real-time execution constraint.
Based on the DT technology, Liu et al. [20] put forward a quad-
play CMCO (i.e., configuration design-motion planning-control
development-optimization decoupling) design architecture for
the flow-type smart manufacturing system in the Industry 4.0
context.
According to the current applications mentioned above, DT
is primarily applied for system design or failure prediction in
the fields of aeronautics, surface ships, manufacturing system,
etc. In fact, DT has high potential to solve the problems existing
in UGs, due to the synchronous linkage and ultra-high fidelity
between physical space and corresponding virtual space. This
article emphasizes its potential applications in RID and FLM of
UGs with a case study on the Petrel glider.
B. Existing Design Methods of UGs
In general, the research and development of a product refers to
the entire design process from start to finish and the steps of every
stage it contains. The professional knowledge and experience
of the individuals act as the core in traditional product design
process. Various tests must be performed at the designing stage
to constantly prove the validity and usability of the product.
UG is a complex integrated system involving multiple dis-
ciplines, such as hydrodynamic shape, pressure hull, control,
navigation, hydraulic transmission, energy, and path planning,
which requires the successive participation of designers. The
final design scheme of UG system can be obtained by repeatedly
integrating, improving and updating the design related to each
discipline. Under this circumstance, many optimization designs
for UGs have been proposed from different disciplines, such
as hydrodynamic shape [21], pressure hull [22], buoyancy unit
[23], and control parameter [24]. However, these traditional
design methods of a single discipline fail to consider the complex
coupling among two or more disciplines, which brings difficulty
to obtain a globally optimal design scheme.
Multidisciplinary design optimization (MDO) provides an
effective way in the design of the complex engineering systems,
which can coordinate all disciplines and decouple their interac-
tions in the design cycle, to improve the design and reduce the
time cost simultaneously. In recent years, Yang et al. [25] has
applied MDO to realize collaborative optimization of multiple
disciplines in UG system, which is a preliminary application
research of model-based system engineering (MBSE) [26].
However, due to the multisource uncertainties, such as complex
TABLE I
COMPARISON OF DESIGN METHODS FOR UGS
ocean environment and uncertain biofouling, the established
system model generally reflects dynamic and time-varying char-
acteristics, making it hard to accurately forecast the motion
performance of UGs in full lifecycle. Moreover, the traditional
system model cannot be reused for a new design scheme. To
sum up, although MDO can decouple the complex relationship
among multiple disciplines, it cannot realize the RID of UGs due
to the lack of a dynamic and integrated digital model, as well
as FLM of UGs due to lack of interaction between the data of
system model in virtual space and multisource data in physical
space. With the development of the new generation of MBSE, it
is possible to establish a DT for integrally describing a complex
system, which can apply to the full lifecycle of UG system,
including conceptual design, detailed design, manufacturing
process, and operation management.
The various design methods for UGs are compared and the
results are given in Table I.
III. ARCHITECTURE OF DT-DRIVEN RID AND FLM OF UGS
To solve the above problems, a new architecture of DT-driven
RID and FLM of UGs is put forward based on our previous
theoretical research and technique development. The proposed
architecture mainly concerns digital modeling, design optimiza-
tion, virtual verification and practical application, as shown in
Fig. 2.
A. Digital Modeling
Digital model is a faithful mapping of the physical UG system,
which provides the first and also the most important basis for
RID of UGs. In this process, designers need to transfer the whole
UG system into a digital model in virtual space by referring to
the data from design space, market survey, application situation,
manufacturing, etc. in physical space, which mainly concerns
digital shape, digital pressure hull, and digital buoyancy device.
Meanwhile, the data of digital model, such as dimensions and
layout, will be transmitted to the physical space to evaluate the
effectiveness of the design, manufacturing, and application. In
this way, it can perfectly guide the improvement of the model
by making full use of the feedback in physical space and the
various problems that have appeared in customers’ usage of the
previous UGs.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: DIGITAL TWIN-DRIVEN INDUSTRIALIZATION DEVELOPMENT OF UNDERWATER GLIDERS 9683
Fig. 2. Architecture of DT-driven RID and FLM of UGs.
B. Design Optimization
Design optimization of UG is carried out based on the digital
model established above to obtain an optimal design scheme for
a specific observation mission. In this process, an optimization
model can be established by coalescing multiphysics simulation,
dynamic model, deep learning network, etc. The design indexes
of the model, such as operation depth, size, and load capacity, can
be extracted from an actual observation mission in the physical
space. Finally, the MDO method is applied to decouple the
relationship among multiphysics and obtain an optimal design
scheme, which will be transmitted into the observation mission
in physical space to preliminarily check the effectiveness of the
design.
C. Virtual Verification
Virtual verification is performed before the manufacturing
to make sure that the design scheme in virtual space satisfies
the requirements in physical space. By importing the historical
data, such as hotel load, ocean current, and seawater density,
the simulation can be carried out to obtain the performance of
design scheme, such as operation time, maximum range, input-
output characteristics, and energy distribution. The performance
data are compared with the design indexes in physical space.
The design scheme will be output if the data satisfy the design
indexes, or improved if not.
D. Practical Application
Practical application is performed when the manufacturing,
assembly, and test are finished. In this process, UGs will return
the quasi-real-time data, such as conductivity-temperature-depth
(CTD) data, navigation data, and anomaly data, to the virtual
space, which will be coalesced with the trim data, manufactur-
ing data, and assembly data to perform state monitoring, fault
TABLE II
DT-DRIVEN METHODOLOGY OF UGS
diagnosis, system identification, and model correction of UGs.
After motion analysis and evaluation, the newcontrol parameters
can be obtained and transferred to UG to improve its navigation
performance or avoid vehicle loss caused by some faults in its
full lifecycle.
Table II clearly lists the research process and research content
of the DT-driven RID and FLM of UGs, and the specific research
details are given in Sections IV and V, respectively.
IV. DT-DRIVEN RID OF UGS
A. Digital Modeling of UGs
As an important part of DT, digital modeling helps accelerate
product development and improve the efficiency of design, man-
ufacturing, and service. As shown in Fig. 3, the digital modeling
of UGs adopts some characteristic parameters to express the
whole system, and the construction process can be divided
into three parts, including subsystem digitization, space layout
digitization and net buoyancy digitization.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
9684 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023
Fig. 3. Digital modeling of UGs.
The digital model of the UG system can be expressed as
D:⎧
⎨
⎩
Dkck
1,c
k
2,··· ,c
k
j
SkVk,xk,y
k,z
k k=1,2··· ,n
ΔBkGk,B
k
(1)
where Dis the digital model of the design scheme, Dkis
the digital model of the subsystem, (ck
1,c
k
2, ..., ck
j) are relevant
characteristic parameters, nis the number of the subsystems, Sk
represents the space layout of the subsystem, Vkis the volume of
the subsystem, [xkykzk] is the space position of the subsystem,
and ΔBkis the net buoyancy of the subsystem (namely the
resultant force of gravity Giand buoyancy Bk).
1) Subsystem Digitization: The whole UG system consists
of multiple subsystems, including pressure hull, hydro-
dynamic shape, battery pack, variable buoyancy device,
attitude adjustment device, etc., whose design schemes
can be expressed by some characteristic parameters. The
design scheme of pressure hull can be expressed by its
structural style, dimensions (diameter, hull length, ring
width, wing thickness, hull thickness, ring number, etc.),
and material properties (yield strength, Poisson’s ratio,
elasticity modulus, material density, etc.). The design
scheme of hydrodynamic shape can be expressed by its
body shape (diameter, forebody length, afterbody length,
total length, etc.), wing dimensions (wing root, wing tip,
wing span, wing position, sweep angle of wings, etc.), and
stabilizer dimensions (stabilizer root, stabilizer tip, stabi-
lizer span, stabilizer position, sweep angle of stabilizers,
etc.). Moreover, the digitization of battery pack can be
realized by its battery number, rated voltage, rated current,
and nominal energy. The design parameters of the variable
buoyancy device consist of power of each component,
volume of bladder and oil tank, quantity of pumps and
solenoid valve, and parameter of buoyancy compensation
device. Once the digital models of all subsystems are
constructed, the design schemes can be expressed by their
characteristic parameters.
2) Space Layout Digitization: Similarly, the volume and
position of each subsystem need to be expressed with
some design parameters for space layout digitization.
After subsystem digitization, the effective space of the
whole UG system can be deduced with dimensions of the
pressure hull and fairings, and the layout of other subsys-
tems inside them can be designed. The volume of each
subsystem can be expressed by its design parameters. For
example, the volume of the variable buoyancy device can
be expressed by the total volume of bladder and oil tank
(related to the oil volume), and the total volume of the
pumps, motors, valves and other components (related to
their model selection). To clearly describe the position
of all subsystems, the distance from their centroids to the
buoyancy center of the whole UG system can be expressed
by three coordinate values.
3) Net buoyancy Digitization: In the system design of UG,
an unmanned observation platform driven by variable
buoyancy, the net buoyancy ΔBis one of the most im-
portant factors that have to be considered. Thus, the net
buoyancy statistics of all subsystems need to be ana-
lyzed. The whole UG system is divided into three parts
to facilitate the net buoyancy digitization, including the
subsystems outside the pressure cabin, the pressure cabin,
and the subsystems inside the pressure cabin. The net
buoyancy of the subsystems outside the pressure cabin
can be deduced by their resultant force of gravity Gand
buoyancy B. The net buoyancy of pressure cabin can be
expressed by its displacement and buoyancy factor bf.
The subsystems inside the pressure cabin do not provide
buoyancy, and the absolute value of their net buoyancy is
equal to their gravity. Specially, bladder, as a part of the
variable buoyancy device, is outside the pressure cabin, so
the net buoyancy of this subsystem needs to be calculated
separately. The net buoyancy of the whole UG system can
be obtained by summarizing the net buoyancy of above
three parts.
The real-time data interaction occurs in the construction of
the digital model of UG system. After above processes, a design
scheme of UG system in physical space can be expressed by
hundreds of design parameters in virtual space, providing a
potential for RID of UG.
B. Design Optimization of UGs
The digital model established above can reflect dynamic and
time-varying characteristics of the UG system, which can be
directly reused for a new design. Based on the digital model, the
design optimization of UGs can be efficiently performed, which
involves multiphysics simulation, determination of the design
indexes, and establishment of the MDO framework, as shown
in Fig. 4.
The design optimization of the UG system can be expressed
as
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
min f(ds,d,y)
[ds,d]=Dk,Sk,ΔBkk=1,2,··· ,n
s.t.g
ids,di,yji≤0
hids,di,yji=0i, j =1,2,··· ,m;i=j
dL
s≤ds≤dU
s,dL
i≤di≤dU
i
(2)
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: DIGITAL TWIN-DRIVEN INDUSTRIALIZATION DEVELOPMENT OF UNDERWATER GLIDERS 9685
Fig. 4. Design optimization of UGs.
where fis the objective function, the shared variable dsand the
local variable dcan be obtained from the digital model (1), yis
the coupled variable among disciplines, diis the local variable
in discipline i,mis the number of the discipline, giand hi
are inequality constraints and equality constraints of the dis-
cipline i,yji is the coupled variable in discipline iimported by
discipline j,dU
Sand dL
Sare the upper bound and lower bound of
ds, and dU
iand dL
iare the upper bound and lower bound of di.
In fact, the relationships among shared variables, local vari-
ables, and coupled variables can be expressed as
M:yij =yij ds,di,yjii, j =1,2,··· ,m;i=j
(3)
where Mrepresents the mechanism models and data models
established by multiphysics simulations.
1) Multiphysics Simulation: With the development of com-
puter technologies, simulation technology is experiencing
rapid development and getting mature. Modeling and
simulation make it possible for designers to evaluate the
product performance through virtual methods rather than
physical experiments, so as to rapidly obtain real-time
feedback about the current design scheme. Specifically,
the multiphysics simulation, which is particularly impor-
tant for multidisciplinary system design, helps realize
more comprehensive design space exploration and more
optimal design scheme. Based on the digital model of
UG established above, the multiphysics simulation of
UG system can be carried out to analyze the specific
performance, which mainly concerns fluid dynamics,
structural mechanics, hydraulic transmission, system dy-
namics, electromagnetic field, sound field, heat transfer,
and vibration mode. By multiphysics simulation, the cor-
responding mechanism models or data models can be
established for the rapid design optimization.
2) Design Indexes: The development of UGs is aimed at
serving ocean science research. Currently, a single de-
sign scheme of the UG system can no longer satisfy
the increasingly diversified requirements of ocean sci-
ence research, and UGs with different operation depths,
mission sensors, dimensions, weight, and operation time
and distance are being developed. In actual engineering
application, the design indexes of a UG system are ex-
tracted from a specific observation mission in physical
space, which are then imported into the relevant sub-
system. Specially, the operation depth, dimensions, and
weight are important indexes for pressure hull design,
and the mission sensors will influence the hydrodynamic
shape design and determine the load capacity of UGs in
net buoyancy digitization. Therefore, the design indexes
determined by the observation missions bring some con-
straints for design optimization of UG system.
3) MDO Framework: MDO provides some efficient meth-
ods to solve optimization problems of the UG system,
and the proper MDO methods can be selected by system
analysis and coupling analysis to decouple interdisci-
plinary relationships. For example, the surrogate model
established by deep learning network can transform the
complex simulation process into a mathematical model
while ensuring certain accuracy, which can efficiently
reduce the calculation cost and rapidly obtain the specific
performance of a design scheme. Three design principles
are proposed here to build an optimal design scheme,
listed as follows.
a) Space balance: The effective space of the UG system
matches properly with all the subsystems designed,
leaving no extra space.
b) Net buoyancy balance: The whole UG system needs
to be neutrally buoyant at the sea surface.
c) Moment balance: The moments of all subsystems to
the buoyancy center need to be balanced. Finally,
a framework of MDO is established by integrating
the digital model, multiphysics simulation, design
indexes, and MDO method, and some algorithms
are selected to solve the optimization and obtain an
optimal design scheme.
C. Virtual Verification of UGs
The traditional methods generally evaluate the validity and
feasibility of a design scheme through some tests, which not only
prolongs the production cycle, but also greatly increases the cost
of time and money. DT-driven virtual verification takes full use
of the data of equipment, environment, and material as well as
history data of the last generation, and therefore can detect the
design defect and find the causes to realize a fast and convenient
redesigning. Also, it can greatly improve the design efficiency
by avoiding tedious verification and testing. The RID of UGs
is completed by virtual verification, which realizes performance
prediction based on data of ocean environment and physical
parameters, as shown in Fig. 5.
The core concept in virtual verification of the UG design
scheme is expressed as
Design optimization ⇒D⇔Yv(D,M,O,P)vs Y0
(4)
where Yvis performance indexes of the current design scheme,
Y0is the design indexes, Ois the historical data of ocean
environment, and Pis the historical data of physical parameters.
By design optimization and contrasting Yvand Y0for certain
times, the virtual verification can be efficiently realized.
1) Performance Prediction: The design parameters of the
current design scheme are imported into the mechanism
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
9686 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023
Fig. 5. Virtual verification of optimization results.
models and data models established in multiphysics sim-
ulation, to rapidly predict the performance of design
scheme, such as gliding velocity, energy distribution, load
capacity, and operation time. If the performance satisfies
the design indexes required by the specific observation
mission, the optimal design scheme will be output and
the manufacturing of prototype begins. If not, the design
parameters need to be updated and the optimization be
performed again until the design indexes are satisfied.
2) Ocean Environment: Since UGs travel with a low velocity
and small driving force, their motion state is suscepti-
ble to the ocean environment, including ocean current,
seawater density, seawater temperature, ocean wave, and
sea area. For example, the variation of seawater density
and seawater temperature will influence the net buoyancy
of UGs, and the variation of ocean current, ocean wave,
and sea area will change the gliding velocity and optimal
control parameters of UGs. Thus, the historical data of
ocean environment in physical space need to be imported
into the multiphysics simulation to improve the prediction
accuracy.
3) Physical Parameters: Moreover, the performance of UGs
is also influenced by some physical parameters mea-
sured in laboratory, such as material property, discharg-
ing efficiency, pump efficiency, manufacturing accu-
racy, and hotel load. For example, the performance of
the pressure hull is influenced by the material prop-
erty and manufacturing accuracy, and the operation time
and distance are obviously affected by the discharging
efficiency, pump efficiency, and hotel load. Therefore,
some physical parameters measured previously can be
directly used in virtual model for verifying the design
scheme.
By integrating the historical data of ocean environment and
physical parameters, designers can create vivid simulation sce-
narios to effectively apply simulation tests to a design scheme
and predict its actual performance as accurately as possible.
Finally, a virtual model of UG can be established after digital
modeling, design optimization and virtual verification, which
will be used and updated in its FLM.
Fig. 6. DT-driven FLM of UGs in practical applications.
V. D T- D RIVEN FLM OF UGS
The engineering prototype can be built after manufacturing,
assembly and testing of the design scheme obtained by RID of
UGs. The DT-driven FLM of UGs begins when the prototype is
applied in specific observation missions, as shown in Fig. 6.
The core concept in the FLM of a UG system is
T=[Tin,Tmea ,Tout]
YF(D,M,Tin,Tmea ,L,Do)versus Tout
(5)
where Tis the sea trial data, Tin,Tmea, and Tout are input
parameters (pitch angles, oil volume, etc.), data measured by
sensors (density, temperature, etc.), and output parameters (ve-
locities, attitudes, etc.), YFis the output parameters obtained
by simulations, Lis the laboratory test data, and Dois the
environment data obtained from other sources. The optimal YF
is obtained by contrasting YFand Tout or searching Tin, and then
the FLM of a UG system can be realized.
In practical applications, four kinds of data can be used in
FLM of UGs, summarized as follows.
1) Sea Trial Condition T: In the sea trial, UGs can re-
turn quasi-real-time data about sea trial condition to the
control center by satellite communication, consisting of
navigation data, energy estimation, fault warning, actual
trajectory, and mission sensor data.
2) Laboratory Test Condition L: This part of data consists of
pressure test data, assembly error, trim condition, manu-
facturing error, vibration modal test data and so on.
3) Data of Design Scheme D: The FLM of UGs is performed
based on the data of digital model and simulation obtained
from RID of UGs in virtual space.
4) Environment Data Obtained From Other Sources Do:
Some environment data, such as the time-varying ocean
current and the typhoon situation, need to be obtained
from other sources due to the limited load capacity
of the UG platform. The above data are put into the
DT methodology, which helps understand degradation
and anomalous events and predict unknowns. Then, the
FLM can be performed for product users, summarized as
follows.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: DIGITAL TWIN-DRIVEN INDUSTRIALIZATION DEVELOPMENT OF UNDERWATER GLIDERS 9687
a) Real-time motion state monitoring: Based on the DT
methodology, the twin of physical UG can be updated in
real-time by employing sensor data and communication
technology. The real-time sea trial data of UG are trans-
mitted to the virtual UG to realize the synchronous link-
age and ultrahigh fidelity between the physical UG and
the corresponding virtual UG. With the acquired real-time
data and other data mentioned above, the operation state
of UG can be understood in real-time, and the following
work can be conducted with the virtual UG and relevant
data.
b) Biofouling evaluation and prediction: Due to their low
velocity and remarkable endurance, UGs are susceptible
to uncertain biofouling, which will increase their drag
and result in extra energy consumption (EC). In addi-
tion, biofouling will change the net buoyancy of UGs,
even causing permanent sinking. Thus, it is significant to
obtain the real-time biofouling condition for improving
the navigation efficiency or avoiding vehicle loss. With
DT methodology, the real-time biofouling condition can
be evaluated by contrasting the error between the actual
motion state of UG in physical space and its virtual
motion state in simulation. Therefore, the growth trend of
biofouling can be predicted with the historical biofouling
data, so as to avoid vehicle loss caused by biofouling.
c) EC prediction and optimization: The high-accuracy EC
analysis and prediction provide the basis for the UG to
realize the longest observation distance or operation time.
With the ultrahigh fidelity of virtual UG between physical
UG and virtual UG, the EC information of UG can be
monitored in real-time. Relevant statistical analysis can
be performed based on the real-time and historical EC
data as well as laboratory test data, which may involve
the EC distribution and total energy carried by the UG.
Therefore, the real-time dump energy can be accurately
evaluated to make plans for timely recovery. Besides, the
EC optimization can be performed to reduce the energy
consumed by major components.
d) Fault diagnosis and operation evaluation: During the
long-time operation in complex ocean environment, UGs
break down frequently due to sensor failure, seal failure,
pump failure, hull corrosion and so on, which will lead
to mission interruption or vehicle loss. Therefore, timely
fault diagnosis and operation evaluation is significant for
avoiding vehicle loss. However, some failures, such as
seal failure and hull corrosion that change gradually, are
difficult to be discovered with traditional methods. With
DT methodology, the minor variation of motion state can
be detected by comparing the motion data of UG with
those of its virtual twin, and the timely fault diagnosis can
be realized. Moreover,the operation state can be evaluated
to decide whether the UG can continue to work normally.
e) Motion optimization and strategy adjustment: The gliding
range of UGs for a single cycle is frequently influenced
by the ocean current and ocean wave due to the underac-
tuation and low velocity, which will change the optimal
motion parameters and control strategy for realizing the
maximum gliding range. With DT methodology, the data
of ocean current and ocean wave in the physical space,
obtained from weather forecast or certain observation
equipment, can be imported into the modeling and sim-
ulation of UGs in virtual space. Therefore, the motion
optimization and strategy adjustment can be performed
with some optimization algorithms, which will guide
the control parameter adjustment of UGs in observation
missions.
f) Model correction and cooperative control: Currently, the
collaborative observation with multiple UGs has become
one of the most important growth trends in ocean observa-
tion. However, influenced by multisource uncertainties,
such as trim error, manufacturing error and assembly
error, UGs of the same type may show different motion
characteristics, even with consistent control parameters,
which makes it difficult to perform collaborative obser-
vation. With DT methodology, the time-varying digital
model of each prototype can be corrected and updated
in real-time, so the control parameters of each prototype
can be optimized and adjusted to realize collaborative
observation.
g) Sensor layout optimization and data processing:The
quality of the observation data collected by mission sen-
sors determines whether the observation with UGs is
effective for ocean science research. However, there exists
a complex coupling relationship between the mission
sensors and the UG platform. Specially, the motion state
variation caused by multisource uncertainties will lead to
the variation of physical fields at sensor probes, further
influence the data accuracy and result in difficulty for data
processing. With DT methodology, digital physical fields,
such as flow field and sound field, can be established by
modeling and simulation of UGs in virtual space, and
then corrected and updated in physical space with sea
trial data. Based on the digital physical fields, the data
processing becomes simple, and the optimal sensor layout
with negligible physical field variation can be obtained.
VI. DT METHODOLOGY APPLICATION IN PETREL UGS
A. Development of Petrel UGs
The research and development of UGs in China began at the
turn of 21th century, since when the UG technology has devel-
oped rapidly and many prototypes with different characteristics
have been designed and tested. Among them, the Petrel glider is
listed in the most mature UGs. After nearly 20 years of research
and development, we have designed many types of Petrel UGs
with different operation depths and gliding distances, mainly
including Petrel-I, Petrel-II, Petrel-200, Petrel-L, Petrel-4000,
and Petrel-X, as shown in Fig. 7. They have been constantly
modified and then verified by a mass of sea trials.
For the productization and application of Petrel UGs, a com-
plete business process, shown in Fig. 8, was jointly established
by our department and Changshu Guorui Technology Co., Ltd.
in 2021, mainly involving design, manufacturing, testing and
service. Petrel UGs provide an effective measure for ocean
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
9688 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023
Fig. 7. Some typical Petrel UGs.
Fig. 8. Production line for UGs.
scientists to realize ocean research with low cost. Up to now,
more than 100 Petrel UGs have been sold to users and applied
in East China Sea, South China Sea, Western Pacific, Indian
Ocean, and Bering Sea. More than 80 observation missions
have been completed, including acoustic observation, mesoscale
eddy observation, internal solitary wave observation, turbulence
observation, and typhoon observation.
B. Case Study: Digital Model-Based Design
Optimization
This article takes Petrel-L as a typical preliminary attempt
of DT-driven design and application, which aims to realize the
longest gliding distance under the weight constraint of 100 kg
and limited energy. Since the first prototype was designed in
2018, Petrel-L has been continually optimized, and broken the
maximum gliding distance record of Chinese UGs for many
times, with 2262.6 km in 2018, 3619.6 km in 2019 [27], and
4435 km in 2020 [28] respectively.
In our latest work [25], the digital model-based design opti-
mization of an improved Petrel-L is carried out, which can be
taken as a case study to preliminarily verify the feasibility of
the DT-driven methodology. Fig. 9shows the specific design
process.
1) Digital Modeling: The whole UG system was first di-
vided into five disciplines by system analysis, including
hydrodynamic shape, pressure hull, buoyancy, attitude
and energy. Then, the characteristic parameters of each
discipline are determined, and the system layout and net
buoyancy of all components of the UG system are defined.
Fig. 9. Digital model-based design optimization of the engineering
prototype. (a) Digital modeling. (b) Design optimization. (c) Virtual veri-
fication.
Finally, a digital model of UG is preliminarily established
to obtain the optimal design scheme.
2) Design Optimization: First of all, the relevant multi-
physics simulations are performed based on the digi-
tal model established above to establish the mechanism
models or data models. For the hydrodynamic shape and
pressure hull, the computational fluid dynamics and finite-
element analysis are used to collect data. Other disciplines
are expressed by the mechanism models, namely the
dynamic model and the EC model. Then, some constraints
are determined, such as weight, velocity, gliding distance,
etc. Finally, the algorithm integrating the concurrent sub-
space optimization, the penalty function method, and the
multipopulation genetic algorithm is adopted to search
the design scheme with the longest gliding distance.
3) Virtual Verification: The optimal design variables, ocean
environment data, and physical parameters are imported
into the dynamic model and EC model to calculate the
motion performance of the design scheme, involving the
motion trajectory, gliding velocity, gliding distance, etc.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: DIGITAL TWIN-DRIVEN INDUSTRIALIZATION DEVELOPMENT OF UNDERWATER GLIDERS 9689
Fig. 10. Navigation trajectory of Perel-L.
The results are contrasted which the design indexes to
determine the final design scheme.
4) Sea trial Verification: Based on the design scheme ob-
tained by above processes, an improved prototype of
Petrel-L with a weight of 93 kg was manufactured, as-
sembled and tested in our previous work. A sea trial
was carried out to verify the motion performance of the
prototype, especially the maximum gliding distance. In
this article, a prototype of about 100 kg is designed
based on the processes of the digital modeling, design
optimization, and virtual verification, whose prospective
maximum gliding distance can reach 5618 km. A sea trail
is performed in this article to test its maximum gliding
distance. As shown in Fig. 10, the prototype was deployed
near Mariana Trench in July 2020, moved across the first
and the second island chains, and was finally recovered in
South China Sea in January 2021. The maximum gliding
distance in this sea trial reaches 5506 km, which has
preliminarily verified the feasibility of DT methodology.
Based on the sea trial data and laboratory test data, we have
currently performed some preliminary studies about the FLM
of UGs, such as biofouling prediction, system identification,
trajectory optimization, and fault diagnosis. In future, a DT of
UG system with higher fidelity will be established in the virtual
space, which will provide technological support for the more
diversified FLM application of the UG system.
With the development and improvement of DT methodology,
more individualized Petrel UGs will be rapidly designed and
applied to some special observation missions. Moreover, UGs
with more remarkable performance can be designed with DT
methodology. In fact,multiple challenges remain to be addressed
before DT can be widely applied in the full lifecycle of UGs,
such as the following.
1) Due to the limitation of underwater communication, the
real-time remote transmission of Big Data cannot be re-
alized by current technologies. Therefore, breakthroughs
in underwater transmission technology need to be made
to realize ultrahigh synchronization between the virtual
and physical UGs.
2) The application of deep learning in DT-driven RID and
FLM of UGs is another urgent challenge that needs to be
addressed. Specifically, deep learning shall be employed
to improve the fidelity between the virtual and physical
UGs and the computational efficiency of DT data pro-
cessing.
3) Considering the complexity of UG system and ocean
environment, the interdisciplinary digital model with a
mass of multisource DT data needs to be constructed,
and the model shall be continually improved to realize
high-fidelity smart analysis and prediction.
VII. CONCLUSION
Currently, two urgent challenges was addressed to promote
the industrialization development of UGs, including RID of
UGs involving multiple disciplines and multiphysics, and FLM
of UGs with a mass of multisource data. In this article, an
architecture of DT-driven RID and FLM of UGs was presented
to solve these problems. The three steps in DT-driven RID of
UGs, including digital modeling, design optimization and virtual
verification, are first introduced in detail. Then, some potential
studies or applications based on the DT-driven FLM of UGs
were introduced to provide a better service. Finally, the Petrel
UGs developed by our group are introduced briefly, and a typical
preliminary attempt of DT-driven methodology based on Petrel
was given to verify the feasibility of DT methodology. The future
challenges of DT application to UGs are also summarized.
This article has preliminarily investigated the framework of
DT-driven RID and FLM of UGs. At present, the research is in
the initial stage and still requires a lot of research work. Our
future work will concentrate on the following aspects.
1) High-fidelity model and high-efficiency simulation of
complex UG systems. High-fidelity model and high-
efficiency simulation are the application basics of DT
methodology, while current model and simulation of
UG system merely consider partial design variables in
certain disciplines. In future, the more and more high-
dimensional design variables from multiphysics fields
will be imported into the model to improve its fidelity,
and some technological measures, such as deep learning,
will be researched and adopted to improve the simulation
efficiency.
2) DT-driven self-adaptation control of UG system under
multisource uncertainties. In the observation missions,
UGs are susceptible to the internal errors (such as trim er-
ror and assembly error) and external disturbances (such as
ocean current and biofouling), which heavily influences
the trajectory accuracy and endurance ability. DT method-
ology allows the UGs to identify these time-varying errors
and disturbances with limited data, therefore to adaptively
adjust their flight parameters and keep the remarkable
motion performance.
3) Cluster networking application of UGs based on DT
methodology. In actual engineering field, a single mis-
sion, such as the mesoscale eddy observation, commonly
requires a collaborative observation with multiple UGs
deployed in different positions, which is becoming a re-
search hotspot in ocean observation. It is important to re-
alize the real-time and high-precision cooperative control,
path planning, and trajectory optimization with DT-driven
methodology in cluster networking application of UGs.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
9690 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023
REFERENCES
[1] M. R. Group, “Development strategy for China’s marine engineering
science and technology to 2035,” China Eng. Sci., vol. 19, no. 01,
pp. 108–117, 2017.
[2] T. Sands, “Development of deterministic artificial intelligence for un-
manned underwater vehicles (UUV),” J. Mar. Sci. Eng., vol. 8, no. 8, 2020,
Art. no. 578.
[3] Z. Peng, J. Wang, and J. Wang, “Constrained control of autonomous
underwater vehicles based on command optimization and disturbance
estimation,” IEEE Trans. Ind. Electron., vol. 66, no. 5, pp. 3627–3635,
May. 2019.
[4] Y. Yokota and T. Matsuda, “Underwater communication using UAVs to
realize high-speed AUV deployment,” Remote Sens., vol. 13, no. 20, 2021,
Art. no. 4173.
[5] D. L. Rudnick, “Ocean research enabled by underwater gliders,” Annu.
Rev. Mar. Sci., vol. 8, pp. 519–541, 2016.
[6] J. Qin, Y. Liu, and R. Grosvenor, “A categorical framework of manufac-
turing for industry 4.0 and beyond,” in Proc. 6th Int. Conf. Changeable,
Agile, Reconfigurable Virtual Prod., 2016, pp. 173–178.
[7] P. P. Ray, “A survey on Internet of Things architectures,” J. King SAUD
Univ.-Comput. Inf. Sci., vol. 30, no. 3, pp. 291–319, 2018.
[8] L. Li, “China’s manufacturing locus in 2025: With a comparison of ”made-
in-China 2025“ and ”Industry 4.0“,” Technol. Forecasting Social Change,
vol. 135, pp. 66–74, 2018.
[9] X. Yu et al., “Deep-learning-based open set fault diagnosis by ex-
treme value theory,” IEEE T. Ind. Inform., vol. 18, no. 1, pp. 185–196,
Jan. 2022.
[10] F. Tao, J. Cheng, Q. Qi, M. Zhang, H. Zhang, and F. Sui, “Digital twin-
driven product design, manufacturing and service with Big Data,” Int. J.
Adv. Manuf. Tech., vol. 94, no. 9, pp. 3563–3576, 2018.
[11] D. Jones, C. Snider, A. Nassehi, J. Yon, and B. Hicks, “Characterising the
digital twin: A systematic literature review,” CIRP J. Manuf. Sci. Technol.,
vol. 29, pp. 36–52, 2020.
[12] Z.Lei, H. Zhou, W. Hu, G. Liu, S. Guan, and X. Feng, “Toward a web-based
digital twin thermal power plant,” IEEE Trans. Ind. Inform., vol. 18, no. 3,
pp. 1716–1725, Mar. 2022.
[13] J. Hochhalter et al., “Coupling damage-sensing particles to the digital twin
concept,” 2014.
[14] K. Reifsnider and P. Majumdar, “Multiphysics stimulated simula-
tion digital twin methods for fleet management,” in Proc. 54th
AIAA/ASME/ASCE/AHS/ASC Structures, Struc. Dyn. Mater. Conf.,
2014, p. 1578.
[15] E. H. Glaessgen and D. S. Stargel, “The digital twin paradigm
for future NASA and U.S. Air force vehicles,” in Proc. 53rd
AIAA/ASME/ASCE/AHS/ASC Struct., Struct. Dyn. Mater. Conf. 20th
AIAA/ASME/AHS Adaptive Struct. Conf. 14th AIAA, 2012, p. 1818.
[16] E. J. Tuegel, A. R. Ingraffea, T. G. Eason, and S. M. Spottswood,
“Reengineering aircraft structural life prediction using a digital twin,” Int.
J. Aerosp. Eng., vol. 2011, 2011, Art. no. 154798.
[17] C. Hongxia, Z. Wei, and Z. Zheng, “Quality management and analysis of
aircraft final assembly based on digital twin,” in Proc. IEEE 11th Int. Conf.
Intell. Hum. Mach. Syst. Cybern., 2019, pp. 202–205.
[18] S. Heim, J. Clemens, J. E. Steck, C. Basic, D. Timmons, and K. Zwiener,
“Predictive maintenance on aircraft and applications with digital twin,” in
Proc. IEEE Int. Conf. Big Data, 2020, pp. 4122–4127.
[19] O. Bondarenko and T. Fukuda, “Development of a diesel engine’s digital
twin for predicting propulsion system dynamics,” Energy, vol. 196, 2020,
Art. no. 117126.
[20] L. Qiang et al., “Digital twin-based designing of the configuration, mo-
tion, control, and optimization model of a flow-type smart manufacturing
system,” J. Manuf. Syst., vol. 58, pp. 52–64, 2021.
[21] M. Yang, Y. Wang, S. Yang, L. Zhang, and J. Deng, “Shape optimization of
underwater glider based on approximate model technology,” Appl. Ocean
Res., vol. 110, 2021, Art. no. 102580.
[22] M. Yang, S. Yang, Y. Wang, Y. Liang, S. Wang, and L. Zhang, “Optimiza-
tion design of neutrally buoyant hull for underwater gliders,” Ocean Eng.,
vol. 209, 2020, Art. no. 107512.
[23] K. Asakawa, K. Watari, H. Ohuchi, M. Nakamura, T. Hyakudome, and
Y. Ishihara, “Buoyancy engine developed for underwater gliders,” Adv.
Robot., vol. 30, no. 1, pp. 41–49, 2016.
[24] S. Wang, M. Yang, Y. Wang, S. Yang,S. Lan, and X. Zhang, “Optimization
of flight parameters for Petrel-L underwater glider,” IEEE J. Ocean. Eng.,
vol. 46, no. 3, pp. 817–828, Jul. 2021.
[25] M. Yang, Y. Wang, Y. Liang, and C. Wang, “A new approach to system
design optimization of underwater gliders,” IEEE/ASME Trans. Mecha-
tronics, vol. 27, no. 5, pp. 3494–3505, Oct. 2022.
[26] A. M. Madni, C. C. Madni, and S. D. Lucero, “Leveraging digital twin
technology in model-based systems engineering,” Systems, vol. 7, no. 1,
p. 7, 2019.
[27] M. Yang, Y. Wang, S. Wang, S. Yang, Y. Song, and L. Zhang, “Motion pa-
rameter optimization for gliding strategy analysis of underwater gliders,”
Ocean Eng, vol. 191, 2019, Art. no. 106502.
[28] S. Wang et al., “Multidisciplinary design optimization of underwater glider
for improving endurance,” Struct. Multidisciplinary Optim., vol. 63, no. 6,
pp. 2835–2851, 2021.
Authorized licensed use limited to: TIANJIN UNIVERSITY. Downloaded on July 26,2023 at 02:47:26 UTC from IEEE Xplore. Restrictions apply.
