=Paper=
{{Paper
|id=Vol-3899/paper14
|storemode=property
|title=Distributed intelligent system architecture for UAV-Assisted monitoring of wind energy infrastructure
|pdfUrl=https://ceur-ws.org/Vol-3899/paper14.pdf
|volume=Vol-3899
|authors=Serhii Svystun,Oleksandr Melnychenko,Pavlo Radiuk,Oleg Savenko,Andrii Lysyi
|dblpUrl=https://dblp.org/rec/conf/advait/SvystunMRSL24
}}
==Distributed intelligent system architecture for UAV-Assisted monitoring of wind energy infrastructure==
https://ceur-ws.org/Vol-3899/paper14.pdf
Distributed intelligent system architecture for
UAV-assisted monitoring of wind energy infrastructure⋆
Serhii Svystun1,∗,†, Oleksandr Melnychenko1,† Pavlo Radiuk1,†, Oleg Savenko1,† and Andrii
Lysyi1,†
1 Khmelnytskyi National University, 11, Institutes str., Khmelnytskyi, 29016, Ukraine
Abstract
With the rapid development of green energy, the efficiency and reliability of wind turbines are key to
sustainable renewable energy production. For that reason, this paper presents a novel intelligent system
architecture designed for the dynamic collection and real-time processing of visual data to detect defects in
wind turbines. The system employs advanced algorithms within a distributed framework to enhance
inspection accuracy and efficiency using unmanned aerial vehicles (UAVs) with integrated visual and
thermal sensors. An experimental study conducted at the “Staryi Sambir-1” wind power plant in Ukraine
demonstrates the system’s effectiveness, showing a significant improvement in defect detection accuracy
(up to 94%) and a reduction in inspection time per turbine (down to 1.5 hours) compared to traditional
methods. The results show that the proposed intelligent system architecture provides a scalable and reliable
solution for wind turbine maintenance, contributing to the durability and performance of renewable energy
infrastructure.
Keywords
wind turbine inspection, UAV, intelligent systems, distributed architecture, defect detection, renewable
energy maintenance, automated monitoring1
1. Introduction
In recent decades, the growth of renewable energy sources, such as wind and solar power, has increased
the importance of proper maintenance and efficient operation of these technologies. Wind turbines, a
pivotal component in wind energy generation, pose unique challenges due to their substantial size and
intricate mechanical structures [1]. At the same time, they tend to take different forms of degradation
over time, including blade cracks and mechanical failures, which can compromise performance or lead to
complete shutdowns. Failure to detect such issues promptly can result in significant repair costs and loss
of energy production [2, 3], emphasizing the critical need for continuous and reliable monitoring systems
[4].
Conventional inspection methods, such as manual visual assessments and ground-based equipment
checks, often fall short in efficiency, speed, and cost-effectiveness. These approaches can be labor-
intensive, time-consuming, and sometimes risky for personnel due to turbine components’ elevated and
exposed locations. The advent of unmanned aerial vehicles (UAVs) [5, 6], coupled with advancements in
visual data processing technologies [7, 8], has opened new avenues for automated inspection solutions
[9]. These technologies facilitate real-time identification of potential problems, thereby minimizing
downtime and extending the operational lifespan of wind turbines [10].
AdvAIT-2024: 1st International Workshop on Advanced Applied Information Technologies, December 5, 2024, Khmelnytskyi,
Ukraine - Zilina, Slovakia
∗ Corresponding author.
† These authors contributed equally.
svystuns@khmnu.edu.ua (S. Svystun); melnychenko@khmnu.edu.ua (O. Melnychenko); radiukp@khmnu.edu.ua (P.
Radiuk); savenko_oleg_st@ukr.net (O. Savenko); andriilysyi@khmnu.edu.ua (A. Lysyi)
0009-0009-8210-6450 (S. Svystun); 0000-0001-8565-7092 (O. Melnychenko); 0000-0003-3609-112X (P. Radiuk); 0000-
0002-4104-745X (O. Savenko); 0009-0001-0065-9740 (A. Lysyi)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
However, existing automated systems often struggle with limitations like insufficient scalability,
inadequate real-time data processing capabilities, and challenges in integrating multiple sensor types for
a holistic assessment [11]. Hence, there is a need for an intelligent system capable of dynamically
collecting and analyzing visual data on turbine defects, effectively integrating various sensors, and
operating efficiently across expansive wind farms.
Addressing these challenges, this research presents a novel intelligent system architecture designed
to enhance the dynamic collection and processing of visual data to identify defects in green energy
infrastructures in real time, focusing on wind turbines. The proposed system leverages UAVs with
integrated sensors, including visual and thermal cameras, and utilizes advanced algorithms to
comprehensively evaluate turbine conditions. Its distributed architecture allows for the parallel operation
of different components, such as UAVs, control devices, and sensors, ensuring seamless functionality in
large-scale wind farm environments.
The primary objective of this study is to develop an advanced, automated monitoring solution that
overcomes the limitations of current systems by improving efficiency, accuracy, and scalability in wind
turbine inspections. By harnessing state-of-the-art UAV technology, sensor integration, and real-time
data processing, the system aims to provide a more reliable and effective means of maintaining the health
and performance of wind turbines. This contributes to the broader goal of sustaining and advancing
renewable energy infrastructures for long-term viability.
2. Related works
Efforts to develop effective monitoring and inspection systems for wind turbines have intensified,
focusing on early defect detection to mitigate unexpected failures and ensure continuous operation
[12, 13]. Traditional methods, such as manual visual inspections [14] and ground-based equipment
assessments [15], are increasingly considered inadequate due to inefficiency, high costs, and safety risks
[16]. This has led to a growing interest in integrating advanced technologies like UAVs, sensor fusion,
and deep learning (DL) [17] methods into inspection practices [18].
Several studies have explored the use of UAVs for wind turbine inspections. Sanati et al. introduced
an automated UAV-based system utilizing visual and thermal imaging to detect defects like blade cracks
[19]. While this method enhanced detection capabilities, it was limited by its reliance on specific imaging
modalities, potentially missing defects detectable by other sensors. Shihavuddin et al. employed UAVs
equipped with convolutional neural networks (CNNs) to improve surface defect detection [20]. Although
this approach increased accuracy and explainability rates, it often depended on single-sensor data, which
might not capture the full spectrum of potential defects.
Recently, different research groups have investigated multi-sensor fusion techniques to overcome the
limitations of single-sensor reliance. Memari et al. integrated visual and thermal sensors to develop an
intelligent system capable of detecting surface and internal defects in turbine blades [21]. This
comprehensive approach improved reliability but introduced complexity in data integration and required
sophisticated algorithms for effective processing. Similarly, Castelar Wembers et al. proposed a multi-
sensor system combining LiDAR and infrared sensors to inspect turbine towers for structural defects [22].
However, this method demanded advanced equipment and increased computational resources.
Advancements in intelligent data processing have also been significant. Tian et al. presented a DL-
based method using CNN and long short-term memory networks to analyze collected data and identify
blade defects [23]. This technique improved detection accuracy but was computationally intensive,
potentially hindering real-time application. Li et al. developed a DL model to predict turbine components’
remaining useful life, enabling predictive maintenance and reducing sudden breakdown risks [24].
However, this model required extensive datasets and high computational power, which might only be
feasible in some operational settings.
Distributed system architectures have been proposed to enhance scalability and efficiency. Jacobsen
et al. suggested a distributed UAV network allowing multiple drones to inspect large wind farms
simultaneously, significantly reducing data collection time [25]. Despite the efficiency gains, coordinating
multiple UAVs presents challenges in communication and synchronization. Mokhtar et al. integrated
cloud and edge computing into a distributed inspection system, facilitating real-time data processing and
quicker decision-making [26]. Nevertheless, relying solely on cloud infrastructure may lead to sensitive
data leaks and data latency issues.
In summary, while significant progress has been made in wind turbine inspection technologies,
challenges persist in achieving comprehensive, real-time monitoring that is both scalable and efficient.
Existing systems often face limitations such as dependency on single-sensor data, high computational
demands of advanced algorithms, and complexities in coordinating distributed systems.
To address the abovementioned issues, this study aims to enhance the dynamic collection and real-
time processing of visual data on wind turbine defects by developing an intelligent system that addresses
these challenges. The scientific contributions of this study are as follows:
• The proposed intelligent system architecture combines data from various sensors, including the
UAV’s visual and thermal cameras, into a distributed structure, providing a comprehensive
assessment of the monitoring process while maintaining scalability.
• By employing intelligent algorithms optimized for real-time processing, the system improves the
accuracy and speed of defect identification without excessive computational requirements.
• The system utilizes a distributed network of UAVs with adaptive coordination strategies to
ensure seamless operation across large wind farms, reducing inspection time and enhancing data
collection efficiency.
These contributions aim to resolve existing issues in the field by providing a more effective and
scalable solution for wind turbine monitoring and maintenance.
3. Proposed architecture of intelligent systems
The task of collecting visual data on defects in green energy objects within a three-dimensional space
necessitates the development of a comprehensive system that integrates various tools with different
functional purposes. Given the specific nature of the objects under investigation, it is advisable to
utilize and manage geographically distributed resources. Communication support is provided
through appropriate information and communication technologies. A key element of the system
should be the provision of automated management of all components, which is essential for the
rational use of the system by end-users. Consequently, the challenge arises to design the intelligent
system 𝑀𝑀𝑠𝑠 to maximize the automation of data collection and analysis processes in real-time.
The components of the designed intelligent system 𝑀𝑀𝑠𝑠 are presented in the general structural
diagram and include various elements. The diversity of system components and their distribution
significantly impact the system’s operation. Therefore, the primary tasks in designing it should focus
on ensuring the system’s integrity during operation and achieving maximum resource efficiency.
To design the intelligent system 𝑀𝑀𝑠𝑠 , it is necessary to first identify the objects it targets. Primarily,
these are green energy objects, particularly wind turbines and their components. These objects are
characterized by complex structures and large sizes, which require a special approach to the
collection and processing of visual data.
The objects of study are located within a distributed three-dimensional space ℝ3 . To ensure precise
collection of visual data for further analysis of detected defects across the entire area, the study region is
defined as a set of spatial zones 𝑍𝑍𝑖𝑖 , each including an individual wind turbine and its surroundings. For
this purpose, we use the coordinates of the central points 𝐶𝐶𝑖𝑖 (𝑥𝑥𝑖𝑖 , 𝑦𝑦𝑖𝑖 , 𝑧𝑧𝑖𝑖 ) of the turbines and the
corresponding coverage radii 𝑅𝑅𝑖𝑖 , which ensure full coverage of the entire structure.
Each spatial zone 𝑍𝑍𝑖𝑖 can be represented as a sphere with center 𝐶𝐶𝑖𝑖 and radius 𝑅𝑅𝑖𝑖 :
𝑍𝑍𝑖𝑖 = {𝑃𝑃 ∈ 𝑅𝑅 𝟛𝟛 ∣ |𝑃𝑃 − 𝐶𝐶𝑖𝑖 | ≤ 𝑅𝑅𝑖𝑖 }, (1)
where |𝑃𝑃 − 𝐶𝐶𝑖𝑖 | is the Euclidean distance between point P and the central point 𝐶𝐶𝑖𝑖 .
The total coverage area A can be expressed as the union of all spatial zones:
𝑛𝑛
(2)
𝐴𝐴 = � 𝑍𝑍𝑖𝑖 .
𝑖𝑖=1
The studied space A is defined as a set of spatial vectors 𝑉𝑉𝑖𝑖 , and the vector V, which describes the
coordinates of the central point and the coverage radius for each wind turbine in the designed system
𝑀𝑀𝑠𝑠 , is represented as follows:
𝑉𝑉𝑖𝑖 = ⟨𝐶𝐶𝑖𝑖 (𝑥𝑥𝑖𝑖 , 𝑦𝑦𝑖𝑖 , 𝑧𝑧𝑖𝑖 ), 𝑅𝑅𝑖𝑖 ⟩, 𝑖𝑖 = 1,2, … , 𝑁𝑁, (3)
where 𝐶𝐶𝑖𝑖 (𝑥𝑥𝑖𝑖 , 𝑦𝑦𝑖𝑖 , 𝑧𝑧𝑖𝑖 ) are the coordinates of the central point of the i-th wind turbine, 𝑅𝑅𝑖𝑖 is the
coverage radius for this turbine, and N is the total number of turbines in the study area.
Thus, the area A can be represented as the set of all vectors 𝑉𝑉𝑖𝑖 , which determine the coordinates
and coverage radii: 𝐴𝐴 = { 𝑉𝑉𝑖𝑖 ∣ 𝑖𝑖 = 1,2, … , 𝑁𝑁 }. This set of vectors defines the complete study space in
the designed system 𝑀𝑀𝑠𝑠 . Therefore, we represent the spatial area in the form of a vector matrix:
𝑣𝑣1,𝑥𝑥1 ⋯ 𝑣𝑣𝑁𝑁𝑣𝑣 ,𝑥𝑥1 (4)
𝑀𝑀𝑣𝑣 = � ⋮ ⋱ ⋮ �,
𝑣𝑣1,𝑣𝑣3,3 ⋯ 𝑣𝑣𝑁𝑁𝑣𝑣 ,𝑣𝑣3,3
where �𝑣𝑣1,𝑥𝑥1 … 𝑣𝑣1,𝑣𝑣3,3 � are the 12 coordinates of the first vector (𝑣𝑣1 ) by formula (3).
Similarly, for the remaining (𝑁𝑁𝑣𝑣 – 1) vectors. The values obtained from formula (4) are used as
input data for the 𝑀𝑀𝑠𝑠 system. Its structural diagram is shown in Figure 1.
Figure 1: Structural diagram of the intelligent system 𝑀𝑀𝑠𝑠 .
Since data collection requires parallel functioning, where different software components of the
system operate simultaneously, and many components perform their tasks in parallel, it is advisable
to design such an intelligent system 𝑀𝑀𝑠𝑠 as a distributed one. Implementing a distributed architecture
allows multiple data processing and component management operations to be executed
simultaneously, reducing delays and improving performance. A distributed system ensures
scalability and reliability, which are critical for processing large volumes of data in real time and
maintaining uninterrupted operation of all components.
The system’s component structure includes UAVs, mobile control devices, computing complexes,
maintenance tools, and multifunctional sensors operating in an automated mode.
The diagram shown in Figure 1 illustrates the architecture of intelligent systems for collecting
visual data and assessing the condition of defects in green energy objects, particularly wind turbines.
The system consists of several levels and includes both hardware and software components that are
integrated to ensure automated data collection, processing, and analysis.
In terms of functionality, each software and hardware component are defined at any given
moment by its state, described by its functional capabilities [27]. Thus, the state of an individual
component can be represented as a vector: 𝑆𝑆𝑓𝑓,𝑠𝑠 (𝑡𝑡) = {𝑃𝑃𝑓𝑓,𝑠𝑠1 (𝑡𝑡), 𝑃𝑃𝑓𝑓,𝑠𝑠2 (𝑡𝑡), … , 𝑃𝑃𝑓𝑓,𝑠𝑠𝑛𝑛𝑓𝑓,𝑠𝑠 (𝑡𝑡)} , where 𝑃𝑃𝑓𝑓,𝑠𝑠 (𝑡𝑡)
represents the parameters of the system components at time t, defining their functional capabilities,
and 𝑛𝑛𝑓𝑓,𝑠𝑠 is the number of such parameters. The set of all component states at a given time t forms
the overall state of the system:
(5)
𝑆𝑆(𝑡𝑡) = � 𝑆𝑆𝑓𝑓,𝑠𝑠 (𝑡𝑡),
𝑓𝑓,𝑠𝑠
The system components interact through specific connections, which can be represented using
graphs. In this graph, vertices represent the system components, while edges denote the connections
between them. The inter-component interactions are captured by the graph 𝐺𝐺𝑓𝑓,𝑠𝑠 , where f defines the
functions representing the connections between components, and s indicates the states of these
components. Thus, the system’s graph is defined as follows:
𝐺𝐺𝑓𝑓,𝑠𝑠 = (𝑉𝑉, 𝐸𝐸), (6)
where V denotes the set of vertices representing the system components, and E represents the set
of edges that define the connections between these components.
Each edge 𝑒𝑒𝑖𝑖𝑖𝑖 ∈ 𝐸𝐸 connects vertices 𝑣𝑣𝑖𝑖 and 𝑣𝑣𝑗𝑗 , corresponding to the interaction between
components i and j. Each connection in the system can be described by a function that specifies how
one component influences another. The set of all connections within the system is expressed as:
𝐸𝐸 = {𝑒𝑒𝑖𝑖𝑖𝑖 �𝑆𝑆𝑖𝑖 , 𝑆𝑆𝑗𝑗 � = 𝑓𝑓𝑖𝑖𝑖𝑖 �𝑃𝑃𝑖𝑖𝑖𝑖 (𝑡𝑡), 𝑃𝑃𝑗𝑗𝑗𝑗 (𝑡𝑡)� ∣ ∀𝑖𝑖, 𝑗𝑗, 𝑘𝑘, 𝑙𝑙}, (7)
where 𝑓𝑓𝑖𝑖𝑖𝑖 represents the functions that describe how the parameters of one component influence
those of another.
Thus, the intelligent system 𝑀𝑀𝑠𝑠 is characterized by its elements and the connections among them,
defined as follows:
𝑀𝑀𝑠𝑠,𝑍𝑍 = {𝑆𝑆(𝑡𝑡), 𝐸𝐸}, (8)
where 𝑀𝑀𝑠𝑠,𝑍𝑍 represents the system 𝑀𝑀𝑠𝑠 , including Z, which refers to its architectural task according
to Figure 1.
The intelligent system 𝑀𝑀𝑠𝑠 must provide the following functions:
• Organizing the preparation and connection of UAVs to the system.
• Creating a three-dimensional software space A.
• Determining the initial position of the hardware device relative to the object under study in
space A.
• Managing the UAV’s trajectory.
• Synchronizing and processing data in real time.
• Integrating with other management and monitoring systems.
• Defining the states of the system and the object under study.
The objective of the system 𝑀𝑀𝑠𝑠 , as outlined in formula (8), is to represent all elements of the system,
along with the processes and interactions occurring among them. Since the heterogeneous components
of the intelligent system 𝑀𝑀𝑠𝑠 and the input data specified in formula (8) are not directly addressed, we
define the system 𝑀𝑀𝑠𝑠 as follows to account for these components:
𝑀𝑀𝑠𝑠𝑍𝑍 = �𝐾𝐾𝑠𝑠,𝑍𝑍,1 , 𝐾𝐾𝑠𝑠,𝑍𝑍,2 , … , 𝐾𝐾𝑠𝑠,𝑍𝑍,𝑛𝑛𝑛𝑛,𝑍𝑍 �, (9)
where 𝑛𝑛𝑓𝑓𝑍𝑍 is the number of sets indicating various system components by their characteristic
features, and 𝐾𝐾𝑠𝑠,𝑍𝑍,𝑖𝑖 is a set whose elements are homogeneous components that comprise the system
𝑀𝑀𝑠𝑠 , with 𝑖𝑖 = 1,2, … , 𝑛𝑛𝑓𝑓𝑍𝑍 . The designation Z in 𝑀𝑀𝑠𝑠𝑍𝑍 for system description refers to its architectural
definition, which includes the physical components.
The final output of the 𝑀𝑀𝑠𝑠 system is derived from the visual data gathered by external sensors
and the intelligent processing applied to achieve the target task. According to formula (4), the
operational three-dimensional space A consists of a set of vectors 𝑉𝑉𝑖𝑖 , so the data collected by the
external sensors is represented in matrix form as follows:
𝑑𝑑1,1 ⋯ 𝑑𝑑𝑁𝑁𝑁𝑁,1 (10)
𝑀𝑀𝐽𝐽𝑘𝑘 = � ⋮ ⋱ ⋮ �,
𝑑𝑑1,𝑛𝑛 ⋯ 𝑑𝑑𝑁𝑁𝑁𝑁,𝑛𝑛
where 𝑑𝑑𝑖𝑖,𝑗𝑗 denotes an element of the matrix 𝑀𝑀𝐽𝐽𝑘𝑘 ,, representing a vector that contains information
gathered by external sensors from the target areas within space (𝑖𝑖, 𝑗𝑗).
The information, collected by external sensors and stored in matrix 𝑀𝑀𝐽𝐽𝑘𝑘 ,, is processed within the
corresponding hardware module, and the processing results contribute to the final outcome. To
formalize this, we define a function 𝐹𝐹𝑗𝑗𝑗𝑗 , which, for each element of the matrix 𝑀𝑀𝐽𝐽𝑘𝑘 (referenced in
formula 10), generates the processed information from each target area. The target information is
then computed as follows:
𝑁𝑁𝑑𝑑 𝑑𝑑𝑛𝑛 (11)
𝑅𝑅𝑗𝑗𝑗𝑗 = � � 𝐹𝐹𝐽𝐽𝑘𝑘 �𝑑𝑑𝑖𝑖,𝑗𝑗 �,
𝑖𝑖=1 𝑗𝑗=1
where 𝑅𝑅𝑗𝑗𝑗𝑗 is the target result of information processing, 𝑁𝑁𝑑𝑑 and 𝑑𝑑𝑛𝑛 are the number of target areas
in the respective dimensions, and 𝐹𝐹𝐽𝐽𝑘𝑘 �𝑑𝑑𝑖𝑖,𝑗𝑗 � is the function that processes information from area (𝑖𝑖, 𝑗𝑗).
Thus, the intelligent system 𝑀𝑀𝑠𝑠 is comprehensively defined across all levels of its architecture,
according to formulas (8) and (9). This definition is essential for detailing both the system’s
components and the interactions among its various modules. As outlined in formula (8), the 𝑀𝑀𝑠𝑠
system operates as a distributed architecture, with modules integrated into software components as
specified in formula (9). The architectural elements described in formula (8) coordinate and manage
the components detailed in formula (9).
The architectural diagram of the 𝑀𝑀𝑠𝑠 system, shown in Figure 2, illustrates the modular level and
the interactions between components.
Figure 2: Modular-Level Architecture of the Intelligent System 𝑀𝑀𝑠𝑠 .
Formulas (4)–(11) provide a precise definition of the 𝑀𝑀𝑠𝑠 system, detailing the variety of system
components and their connections. The architecture presented at the modular level identifies specific
modules within the software components, ensuring a system design that accommodates all potential
complexities and enables component scalability through module reuse. Together, formulas (4)–(11)
define the format of input and output data, providing a comprehensive description of the 𝑀𝑀𝑠𝑠 system.
Further stages of 𝑀𝑀𝑠𝑠 system design require detailing each component according to these formulas.
The architecture shown in Figure 2 encompasses all components outlined in formulas (4)–(11),
including the processes for generating and processing results, which constitute the system’s output
data. The main components of the system are as follows:
1. Mobile Control System, responsible for coordinating and monitoring UAVs. It performs the
following key functions:
• Ensures accurate real-time determination of UAV coordinates and navigation in space.
• Transmits visual data collected by the UAV to the central module for further processing.
• Monitors and responds to emergency situations, ensuring the safety of UAV operations.
• Displays detected defects on an interactive interface.
2. Hardware Module, which integrates software components and provides:
• Implementation of flight algorithms and support for UAV stability in the air.
• Accurate determination and correction of UAV coordinates.
• Collection of visual information from cameras and other sensors and the transmission of this
data to the mobile control system.
3. Central Module, a key element of the system, ensures:
• Continuous system operation even in the event of partial failures.
• Analysis and storage of information received from other modules.
• Monitoring the operational states of all system components, ensuring their optimal
performance.
• Analysis of processed data and generation of recommendations or automatic decisions based
on that data.
4. Visual Data Composition Module, which performs the following functions:
• Receives video data from the mobile control system.
• Processes and composes frames from various hardware sensors to provide high-quality visual
analysis.
5. Defect Criticality Assessment Module, responsible for:
• Analyzing the data received to identify defects.
• Assessing the severity of detected defects and their impact on the object.
6. Trajectory Management Module, which performs the following tasks:
• Ensures the stability of the UAV’s trajectory.
• Predicts future positions to optimize the UAV’s route.
Therefore, the developed architecture of the intelligent 𝑀𝑀𝑠𝑠 system establishes the requirements for
the key elements and components necessary for its operation, as well as defines the interconnections
between them.
4. Results and discussion
For the experimental study of the effectiveness of the developed automated system 𝑀𝑀𝑠𝑠 , the wind
power plant “Staryi Sambir-1,” [28] located in the Carpathian region of Ukraine, was selected. This
station is equipped with Vestas V112 wind turbines, whose characteristics meet modern
requirements for renewable energy equipment. Specifically, each wind energy unit (WEU) has a
nominal power of 3 MW, a rotor diameter of 112 meters, and a tower height of 84 meters, providing
significant area coverage of 9,852 m². Blades measuring 56 meters in length can operate efficiently in
a wind speed range from 3 to 25 m/s, with a nominal speed of 12 m/s.
A comparison was conducted between the traditional manual approach and the automated
intelligent system 𝑀𝑀𝑠𝑠 for data collection and processing to detect and evaluate the criticality of
defects. Table 1 provides a comparative analysis of the key efficiency indicators for two methods:
the traditional operator-based approach and the automated intelligent system 𝑀𝑀𝑠𝑠 .
Table 1
Comparison of efficiency indicators between the traditional and automated 𝑀𝑀𝑠𝑠 data collection
systems
Method Number of Inspection Defect Coverage Critical Data
Simultaneous Time per Detection Completeness Determination Update
Inspections WEU (%) (%) Time (hours) Frequency
(N) (hours) (months)
Traditional 1 5.5–6.0 75–85 70–80 1.0–1.5 3
(operator)
System 𝑀𝑀𝑠𝑠 1 1.5–2.1 90–94 90–95 0.1–0.2 1
2 1.5–2.0 90–94 90–95 0.1–0.2 1
The conducted experiments demonstrate that the intelligent system notably improves defect
detection accuracy, achieving 90–94%, and coverage completeness, reaching 90–95%, compared to
the traditional method’s 75–85% accuracy and 70–80% coverage.
Additionally, it can be seen in Table 1 that the automated system 𝑀𝑀𝑠𝑠 reduces the inspection time
of a single WEU from 5.5–6.0 hours to 1.5–2.0 hours, enhancing the processing intensity of large data
volumes and allowing regular updates on the WEU’s condition (monthly, unlike the traditional
method, which updates data every three months). Figure 3 illustrates the overall effectiveness of the
intelligent system 𝑀𝑀𝑠𝑠 compared to the traditional data collection method.
Figure 3: Illustration of the effectiveness of the automated intelligent system 𝑀𝑀𝑠𝑠 compared to the
traditional data collection method.
Figure 3 visually underscores the advantages of the intelligent system by illustrating the increased
precision and reduced operator dependence. The use of UAVs minimizes human error and safety
risks, particularly for components located in challenging or hazardous environments, such as high-
altitude turbine blades. By enabling regular inspections with higher accuracy and coverage, the
system contributes to extending the operational lifespan of wind turbines and optimizing their
performance.
Table 2 shows how the intelligent system 𝑀𝑀𝑠𝑠 provides more accurate and faster determination of
defect criticality.
Table 2
Comparison of efficiency indicators between the traditional and automated 𝑀𝑀𝑠𝑠 data collection
systems
Defect Type Method Defect Size WEU Critical Critical
(cm) Component Assessment Assessment
(1–10) Time (hours)
Crack Traditional 5.0–7.0 Blade 7–9 1.0–1.5
(manual)
System 𝑀𝑀𝑠𝑠 . 4.5–7.5 Blade 6–7 0.1–0.2
Corrosion Traditional 10.0–15.0 Tower 7–9 1.0–1.5
(manual)
System 𝑀𝑀𝑠𝑠 . 9.0–14.0 Tower 8–9 0.1–0.2
Overheating Traditional – Generator 7–9 1.0–1.5
(manual)
System 𝑀𝑀𝑠𝑠 . – Generator 6–7 0.1–0.2
For example, for blade cracks, the system reduces the criticality assessment time from 1.0–1.5
hours to 0.1–0.2 hours. This approach significantly enhances precision, allowing for prompt
criticality assessments of defects such as tower corrosion and generator overheating, thereby
reducing delays in decision-making regarding necessary repairs.
The automated system also significantly reduced inspection time, as shown previously in Table
1. The average inspection duration per WEU decreased from 5.5–6.0 hours using traditional methods
to 1.5–2.0 hours with the intelligent system. This reduction enhances operational efficiency and
facilitates more frequent inspections, allowing for the timely identification and mitigation of
potential issues.
In addition to quicker inspections, Table 2 highlights the system’s ability to reduce the time
required for criticality assessments. For example, blade cracks were assessed within 0.1–0.2 hours
using the automated system, compared to 1.0–1.5 hours with manual methods. Similar improvements
were observed for other defect types, such as tower corrosion and generator overheating.
Figure 4 illustrates the narrowing of the criticality assessment range when using the intelligent
system. Traditional methods, which rely heavily on subjective judgment, produced broader criticality
ranges (e.g., 7–9 for blade cracks). In contrast, the intelligent system leveraged data-driven
algorithms to achieve more precise assessments (e.g., 6–7 for the same defect type).
Figure 4: Results of defect criticality assessments using traditional and intelligent methods.
The experimental results also demonstrate the system’s potential to improve safety and reduce
operational costs. By minimizing the need for manual inspections in hazardous environments, the
system mitigates risks to personnel while achieving better inspection outcomes. Moreover, the
automation of data collection and processing reduces dependency on highly skilled operators,
making the solution more accessible and cost-effective.
Despite its advantages, the proposed system has certain limitations that warrant further
investigation. For instance, the reliance on advanced algorithms and high-quality sensors may pose
challenges in terms of initial deployment costs and computational requirements. Additionally,
adverse weather conditions, such as strong winds or poor visibility, could affect UAV performance
and data accuracy. Addressing these challenges through algorithmic optimizations and robust
hardware design will be critical for ensuring the system’s reliability under diverse operating
conditions.
5. Conclusions
The proposed intelligent system architecture greatly enhances the dynamic collection and analysis of
visual data for detecting defects in wind turbines. Experimental results from the “Staryi Sambir-1” wind
power plant show that the system boosts defect detection accuracy to 90–94%, compared to 75–85%
achieved with traditional manual inspections. The inspection time per WEU is reduced from
approximately 6 hours to as little as 1.5 hours, with coverage completeness increasing to 90–95%.
Additionally, the system enables faster criticality assessments, cutting evaluation time from up to 1.5
hours to just 0.1–0.2 hours. Despite these advancements, the proposed approach has limitations, such
as the initial costs of deploying UAVs and sensors, the need for sophisticated algorithms to manage
multi-sensor data fusion, and potential challenges in coordinating UAVs under adverse weather
conditions. Moreover, reliance on high-quality data processing infrastructure may pose accessibility
challenges in remote locations.
Future research will focus on mitigating these limitations by optimizing the system’s cost-
effectiveness, enhancing algorithmic performance under diverse environmental conditions, and
refining UAV coordination mechanisms.
Declaration on Generative AI
During the preparation of this work, the authors used Grammarly in order to: grammar and spelling
check; DeepL Translate in order to: some phrases translation into English. After using these
tools/services, the authors reviewed and edited the content as needed and take full responsibility for
the publication’s content.
References
[1] M. Civera, C. Surace, Non-destructive techniques for the condition and structural health
monitoring of wind turbines: A literature review of the last 20 years, Sensors 22.4 (2022) 1627.
[2] J.-Y. Hsu, Y.-F. Wang, K.-C. Lin, M.-Y. Chen, J. H.-Y. Hsu, Wind turbine fault diagnosis and
predictive maintenance through statistical process control and machine learning, IEEE Access
8 (2020) 23427–23439. doi:10.1109/access.2020.2968615.
[3] Y. Xu, N. M. M. Nascimento, P. H. F. de Sousa, F. G. Nogueira, B. C. Torrico, T. Han, C. Jia,
P. P. Rebouças Filho, Multi-sensor edge computing architecture for identification of failures
short-circuits in wind turbine generators, Appl. Soft Comput. 101 (2021) 107053.
[4] T. Hovorushchenko, V. Bachuk, Y. Hnatchuk, I. Zasornova, H. E. Bouhissi, Overview of the
methods and tools for environmental components monitoring, Comput. Syst. Inf. Technol. No.
3 (2024) 51–57. doi:10.31891/csit-2024-3-7.
[5] Z. Zhang, Z. Shu, Unmanned aerial vehicle (UAV)-assisted damage detection of wind turbine
blades: A review, Energies 17.15 (2024) 3731. doi:10.3390/en17153731.
[6] M. H. Nordin, et al., Collaborative unmanned vehicles for inspection, maintenance, and repairs
of offshore wind turbines, Drones 6.6 (2022) 137.
[7] A. Shihavuddin, M. R. A. Rashid, M. H. Maruf, M. A. Hasan, M. A. u. Haq, R. H. Ashique,
A. A. Mansur, Image based surface damage detection of renewable energy installations using a
unified deep learning approach, Energy Rep. 7 (2021) 4566–4576. doi:10.1016/j.egyr.2021.07.045.
[8] S. Shahsavarani, F. López, C. Ibarra-Castanedo, X. P. V. Maldague, Semantic segmentation of
defects in infrastructures through multi-modal images, in: N. P. Avdelidis, G. Ferrarini, F. López
(Eds.), Thermosense: Thermal infrared applications XLVI, SPIE Digital Library, Bellingham, WA,
USA, 2024, pp. 199–208. doi:10.1117/12.3013884.
[9] G. Helbing, M. Ritter, Deep Learning for fault detection in wind turbines, Renew. Sustain.
Energy Rev. 98 (2018) 189–198. doi:10.1016/j.rser.2018.09.012.
[10] S. Vyas, S. S. Verma, A. Prasad, Study of UAV management using cloud-based systems, in:
Advanced Technologies and Societal Change, Springer Nature Singapore, Singapore, 2023,
pp. 97–110. doi:10.1007/978-981-19-5723-9_7.
[11] O. Melnychenko, L. Scislo, O. Savenko, A. Sachenko, P. Radiuk, Intelligent integrated system for
fruit detection using multi-uav imaging and deep learning, Sensors 24.6 (2024) 1913.
doi:10.3390/s24061913.
[12] M. Hossain, A. Abu-Siada, S. Muyeen, Methods for advanced wind turbine condition monitoring
and early diagnosis: A literature review, Energies 11.5 (2018) 1309. doi:10.3390/en11051309.
[13] O. Melnychenko, O. Savenko, A self-organized automated system to control unmanned aerial
vehicles for object detection, in: Proceedings of the 4th International Workshop on Intelligent
Information Technologies & Systems of Information Security, CEUR-WS.org, Aachen, 2020,
pp. 589–600.
[14] K. Kong, K. Dyer, C. Payne, I. Hamerton, P. M. Weaver, Progress and trends in damage
detection methods, maintenance, and data-driven monitoring of wind turbine blades – A review,
Renew. Energy Focus 44 (2022) 390–412. doi:10.1016/j.ref.2022.08.005.
[15] O. Melnychenko, O. Savenko, P. Radiuk, Apple detection with occlusions using modified
yolov5-v1, in: 2023 IEEE 12th international conference on intelligent data acquisition and
advanced computing systems: technology and applications (IDAACS), IEEE, Inc., New York, NY,
USA, 2023, pp. 107–112. doi:10.1109/idaacs58523.2023.10348779.
[16] L. Dubchak, Modern renewable energy sources and methods for detecting their defects, Comput.
Syst. Inf. Technol. No. 2 (2024) 21–26. doi:10.31891/csit-2024-2-3.
[17] Y. LeCun, Y. Bengio, G. Hinton, Deep learning, Nature 521.7553 (2015) 436–444.
doi:10.1038/nature14539.
[18] I. Gohar, A. Halimi, J. See, W. K. Yew, C. Yang, Slice-aided defect detection in ultra high-
resolution wind turbine blade images, Machines 11.10 (2023) 953.
doi:10.3390/machines11100953.
[19] H. Sanati, D. Wood, Q. Sun, Condition monitoring of wind turbine blades using active and
passive thermography, Applied Science. 8.10 (2018) 2004. doi:10.3390/app8102004.
[20] A. Shihavuddin, X. Chen, V. Fedorov, A. Nymark Christensen, N. Andre Brogaard Riis,
K. Branner, A. Bjorholm Dahl, R. Reinhold Paulsen, Wind turbine surface damage detection by
deep learning aided drone inspection analysis, Energies 12.4 (2019) 676. doi:10.3390/en12040676.
[21] M. Memari, M. Shekaramiz, M. A. S. Masoum, A. C. Seibi, Data fusion and ensemble learning for
advanced anomaly detection using multi-spectral RGB and thermal imaging of small wind
turbine blades, Energies 17.3 (2024) 673. doi:10.3390/en17030673.
[22] C. Castelar Wembers, J. Pflughaupt, L. Moshagen, M. Kurenkov, T. Lewejohann, G. Schildbach,
LiDAR‐based automated UAV inspection of wind turbine rotor blades, J. Field Robot. 41.4 (2024)
1116–1132. doi:10.1002/rob.22309.
[23] W. Tian, X. Cheng, G. Li, F. Shi, S. Chen, H. Zhang, A multilevel convolutional recurrent neural
network for blade icing detection of wind turbine, IEEE Sens. J. 21.18 (2021) 20311–20323.
doi:10.1109/jsen.2021.3093726.
[24] X. Li, X. Lei, L. Jiang, T. Yang, Z. Ge, A new strategy: Remaining useful life prediction of wind
power bearings based on deep learning under data missing conditions, Mathematics 12.13 (2024)
2119. doi:10.3390/math12132119.
[25] R. H. Jacobsen, L. Matlekovic, L. Shi, N. Malle, N. Ayoub, K. Hageman, S. Hansen, F. F. Nyboe,
E. Ebeid, Design of an autonomous cooperative drone swarm for inspections of safety critical
infrastructure, Appl. Sci. 13.3 (2023) 1256. doi:10.3390/app13031256.
[26] B. Mokhtar, AI-enabled collaborative distributed computing in networked uavs, IEEE Access 12
(2024) 96515–96526. doi:10.1109/access.2024.3425523.
[27] A. Gaeta, V. Loia, F. Orciuoli, A comprehensive model and computational methods to improve
Situation Awareness in Intelligence scenarios, Appl. Intell. 51.9 (2021) 6585–6608.
[28] Staryi Sambir-1 (Ukraine) - Wind farms - The Wind Power. URL:
https://www.thewindpower.net/windfarm_en_98_staryi-sambir-1.php.