Deep Learning and NLP-Based Trend Analysis in Actuators and Power Electronics

Abstract

Actuators and power electronics are fundamental components of modern control systems, enabling high-precision functionality, enhanced energy efficiency, and sophisticated automation. This study investigates evolving research trends and thematic developments in these areas spanning the last two decades (2005–2024). This study analyzed 1840 peer-reviewed abstracts obtained from the Web of Science database using BERTopic modeling, which integrates transformer-based sentence embeddings with UMAP for dimensionality reduction and HDBSCAN for clustering. The approach also employed class-based TF-IDF calculations, intertopic distance visualization, and hierarchical clustering to clarify topic structures. The analysis revealed a steady increase in research publications, with a marked surge post-2015. From 2005 to 2014, investigations were mainly focused on established areas including piezoelectric actuators, adaptive control, and hydraulic systems. In contrast, the 2015–2024 period saw broader diversification into new topics such as advanced materials, robotic mechanisms, resilient systems, and networked actuator control through communication protocols. The structural topic analysis indicated a shift from a unified to a more differentiated and specialized spectrum of research themes. This study offers a rigorous, data-driven outlook on the increasing complexity and diversity of actuator and power electronics research. The findings are pertinent for researchers, engineers, and policymakers aiming to advance state-of-the-art, sustainable industrial technologies.

1. Introduction

The evolution of industrial systems toward greater autonomy, enhanced energy efficiency, and improved responsiveness has significantly highlighted the importance of actuators. Serving as critical nodes that connect control algorithms with tangible motion, actuators play an indispensable role across multiple sectors. In smart manufacturing, actuators achieve sub-micron-level motion control, enabling the operation of high-precision robotic arms and automated production processes. For electric vehicles (EVs), actuators are essential for steering, braking, and thermal management, while power electronics govern inverters and converters that determine both energy efficiency and range. Within the biomedical domain, soft actuators, integrated with power electronic modules, facilitate the operation of minimally invasive surgical tools and enable assistive healthcare devices. For renewable energy applications, actuators are used to continuously adjust wind turbine blade angles, while power converters maintain stable energy delivery for smart grid integration. These technological solutions address central challenges such as energy optimization, instant responsiveness, and operational robustness across distinct fields. Ongoing advancements in power electronics, which enable efficient energy conversion and high-precision, real-time control, have played a central role in enhancing the performance and intelligence of actuator systems.

Recent technological progress in this field has promoted the adoption of wide-bandgap semiconductors, embedded digital control systems, and advanced real-time monitoring approaches, collectively resulting in notable improvements in switching performance, thermal characteristics, and energy density. These developments position power electronics as a foundational element for future actuator innovation [1]. Moreover, integrating artificial intelligence (AI) and machine learning (ML) within power electronic platforms has presented new opportunities for autonomous diagnostics, advanced fault-tolerant operations, and adaptive real-time system configuration [2]. As an example, Luan et al. [3] demonstrated that a predefined-time sliding-mode controller, combined with an observer, markedly strengthens control accuracy and fault tolerance in permanent-magnet motors. In parallel, Zhang et al. [4] introduced a time-varying damping model that enables adaptive vibration responses in structures subject to clearance-induced nonlinearities.

Such integrated solutions have been effectively deployed in a wide array of industrial domains. In the renewable energy sector, intelligent converters heighten the dependability of wind energy transformation systems under fluctuating operating states. For transportation, electric drives utilize AI-enhanced inverters to optimize both torque control and energy utilization. In the medical sector, soft robotic actuators driven by power electronic modules support accurate motion control in minimally invasive surgical tasks [5,6].

Nevertheless, although considerable technical research has addressed component-level enhancements, there is a lack of a thorough understanding regarding the co-evolution of actuators and power electronics. The majority of earlier investigations have predominantly emphasized innovations at the level of individual devices, with limited attention to how research priorities have transitioned between fields or how contemporary technological convergence—notably AI-assisted control—has redirected emphasis within the domain. The lack of meta-level analysis restricts the ability of scholars, policymakers, and industry professionals to anticipate future advancements and strategically manage resource allocation.

To address this deficiency, this study aims to systematically identify, organize, and examine the long-term trends in research that involve both actuators and power electronics. Specifically, we assess how core research domains have evolved throughout the past two decades (2005–2024), analyze the role of AI in fostering novel research avenues, and investigate how the convergence of control, power conversion, and actuation is shaping changes in system architecture.

In pursuit of these aims, we implement a deep learning-based text mining framework that integrates Bidirectional Encoder Representations from Transformers (BERT)-based sentence embeddings, Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction, Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) for unsupervised cluster analysis, and BERTopic for topic modeling and interpretability. We further use class-based TF-IDF (c-TF-IDF) to distinguish key terms representing each cluster and to monitor the variation in topic prominence over time. The analysis is performed on a collection of 1840 peer-reviewed abstracts sourced from the Web of Science database.

This work extends existing knowledge by introducing a transparent, data-driven methodology to clarify the evolution of actuator and power electronics research. The findings offer practical insights for academic scholars, system engineers, and R&D strategists who need to track technological advancements and strategically align their efforts with emerging trends in innovation.

The structure of this paper is organized as follows: Section 2 reviews the existing literature concerning actuators and power electronics. Section 3 explains the research approach, covering both data acquisition procedures and analytical techniques adopted for trend assessment. Section 4 discloses the primary outcomes of temporal trend analyses and insights obtained from topic modeling. Section 5 evaluates these findings from academic, industrial, and policy-related perspectives, underlining relevant practical impacts. Finally, Section 6 concludes the paper by discussing limitations and proposing future directions for research.

2. Literature Review

To clarify the research landscape, this literature review is divided into three distinct sections. Section 2.1 presents fundamental technological advancements, application domains, and prevailing challenges within actuators and power electronics. Section 2.2 evaluates published trend analyses from both technological and bibliometric viewpoints. Section 2.3 examines recent methodological progress in topic modeling, with an emphasis on the practical application of BERTopic which serves as the core analytical framework for this study. This layered review synthesizes technical depth with methodological precision, supporting the rationale for a data-driven examination of research patterns.

2.1. Topical Review: Actuators and Power Electronics

In modern electromechanical systems, actuators and power electronics operate in unison to enable advancements across diverse industrial domains. Actuators play a foundational role by converting electrical signals into precise mechanical motion, facilitating their integration in robotics, aerospace, automotive, energy, and biomedical systems. Power electronics are indispensable for the efficient conversion, control, and distribution of the electrical energy powering these actuators, critically influencing energy efficiency and overall system performance [7,8,9].

Over the past few decades, actuator technologies have transitioned from traditional electromechanical devices to sophisticated systems employing advanced materials such as piezoelectrics, magnetostrictive materials, and electroactive polymers. These new-generation actuators provide ultra-fine precision at the nanometer scale, accelerated response times, and lower power requirements, which prove advantageous for high-precision applications in semiconductor fabrication and biomedical engineering [10,11,12,13]. Additionally, smart actuator designs increasingly incorporate embedded sensors, real-time control units, and communication modules, supporting features such as self-diagnostic capabilities, predictive maintenance, and instantaneous analytics—essential functions in industrial automation and intelligent manufacturing environments [14,15,16].

Parallel to actuator developments, power electronics has advanced rapidly through the implementation of wide-bandgap semiconductors like silicon carbide and gallium nitride, which support elevated voltage, frequency, and operational efficiency. These advancements have expanded the deployment of power electronics to encompass EVs, renewable energy infrastructure, and aerospace innovations [17,18,19]. Notable progress includes the development of high-efficiency inverters and DC-DC converters. Ongoing research prioritizes maximizing energy conversion, reducing component footprint and mass, optimizing thermal regulation, and harnessing AI for predictive control and real-time condition diagnostics, all of which contribute to greater reliability and autonomous performance in power electronic systems [20,21,22].

These technological advancements are fueling significant innovations across a range of industries. In the EV sector, the integration of high-efficiency inverters and regenerative braking systems has resulted in substantial improvements in driving range and energy efficiency. In the renewable energy sector, smart grids based on power electronics are enhancing the stability and performance of energy supply networks [23,24,25]. In industrial automation, the adoption of sub-micron-level linear actuators and high-speed, high-precision servo drive systems has become instrumental in increasing productivity [26,27,28].

Despite notable progress, various technical challenges persist within the field. These include the management of heat in high-power-density systems, the mitigation of electromagnetic interference resulting from miniaturization, the pursuit of long-term reliability under diverse operational environments, and the scaling of manufacturing processes for advanced material-based actuators, all requiring continued research [29,30,31]. Furthermore, emerging concerns such as engineering energy efficiency, environmentally sustainable power electronic systems, securing data exchange and establishing standardized communication protocols for smart actuators, and promoting interoperability among complex, heterogeneous systems have seen increased focus [32,33].

2.2. Existing Trend Analyses in Actuator and Power Electronics Research

Prior studies in actuator and power electronics research are generally categorized into technology-centric analyses and bibliometric trend analyses. Whereas the earlier sections of this work emphasized technical successes and prevailing challenges, numerous independent studies have employed bibliometric approaches to systematically identify research directions in this area [34,35].

A 2023 article listed in PubMed examined 111 major publications on soft robotics from 2008 to 2022, offering a structured review across three dimensions: bio-inspired design frameworks, types of control (open-/closed-loop), and biomedical applications. This study identified actuator-related technologies as key contributors, accounting for 34% of the reviewed works, and reported a mean annual growth rate of 18.7% over the past five years, with a focus on dielectric elastomer actuators and advancements in actuator structural optimization [36].

A 2011 review indexed by Semantic Scholar analyzed 109,661 IEEE database articles published between 2001 and 2010, categorizing power semiconductor devices (38.2%), converter topologies (29.1%), and control strategies (22.4%) as leading research themes. Notably, research on wide bandgap (WBG) semiconductors has demonstrated significant expansion, averaging an annual growth rate of 24.3% since 2010 [35].

A 2024 study published in the MDPI Electricity journal analyzed over 1200 articles on smart grids from 2017 to 2022. The findings indicated that power electronics technologies accounted for 68% of the research activity within energy management systems (EMSs) and demand-response frameworks. In addition, the annual growth rate for research on DC-DC converter topologies reached 15%, highlighting their rising significance as a major emerging technology [37].

Traditional technological analyses typically emphasize component performance and physical limitations, whereas bibliometric trend analyses provide a broader view of research community development, interdisciplinary collaboration, and policy impact. As an illustration, a 2023 study published in Clean Energy demonstrated that integrating both methodologies could enhance the accuracy of technological forecasting by up to 41% [38].

In this research domain, blending technical and bibliometric methodologies requires the development of a comprehensive multi-temporal analytic framework. Notably, recent progress has emphasized leveraging ML-driven dynamic keyword extraction and integrating real-time data streams. According to a 2025 foresight report from Nature Electronics, the implementation of these combined strategies could enhance R&D efficiency by as much as 57% [39].

2.3. Methodological Review: Topic Modeling and BERTopic

In recent years, topic modeling techniques—in particular, advanced approaches such as BERTopic—have gained significant traction in both academic and industrial environments for systematically mapping large-scale research trends [40]. Topic modeling serves to identify key research themes and patterns from unstructured text sources including academic papers, patents, and technical documentation, with traditional algorithms such as LDA and NMF extensively utilized in this field [7,8]. However, these conventional approaches mainly rely on word frequency and co-occurrence statistics, which can limit their ability to capture contextual nuances. This limitation is especially evident in highly specialized areas with dense technical language [41,42].

BERTopic mitigates these challenges by employing BERT-driven sentence embeddings together with advanced clustering algorithms and dimensionality reduction techniques, such as HDBSCAN and UMAP [43,44]. Unlike frequency-based models, BERTopic produces contextually informed topic representations that deliver high accuracy, even for concise documents like research abstracts. It effectively addresses issues of polysemy and synonymy, and its UMAP-supported interactive visualizations clarify topic interrelations and patterns of evolution [45].

BERTopic has been widely applied to analyze publications and patents in advanced domains such as the industrial internet, semiconductors, and biomedical engineering. It enables chronological analysis of the emergence and development of specific technical topics. In research related to actuators and power electronics, BERTopic exhibits strong capability in identifying emergent subtopics, such as “wide bandgap devices,” “smart actuators,” and “high-efficiency inverters” [46,47]. BERTopic further assists in discovering areas of convergence and potential research gaps, demonstrating more than threefold increases in topic coherence and processing speed compared to LDA when used with large-scale datasets such as abstracts and patent summaries [48,49]. Additionally, its integration with large language models (LLMs) like ChatGPT-4o enhances topic interpretation and offers interactive analytical functions [50].

This study employs BERTopic for several principal reasons. First, the fields of actuators and power electronics undergo rapid technological evolution and frequent interdisciplinary advancement. Therefore, a tool capable of precisely capturing nuanced topic variations in context is essential [51,52]. BERTopic outperforms traditional models by providing superior topic separation and effectively exposing newly emerging trends and knowledge gaps. Second, it efficiently manages large volumes of abstracts and technical research, making it especially suitable for large-scale trend analysis [53]. Third, its visualization features support researchers in clearly tracking topic relationships, evolutionary trends, and areas of convergence, thus generating actionable insights for guiding research directions and setting priorities. Fourth, the integration of state-of-the-art AI tools such as LLMs enables automated topic interpretation and real-time trend tracking, which aligns with the evolving requirements of research environments [54,55,56].

Overall, BERTopic is considered a particularly robust methodological tool for analyzing extensive research trends and technological advancement in the areas of actuators and power electronics. Building on its contextual awareness, computational efficiency, and high interpretability, this study aims to overcome previous analytical constraints and deliver a comprehensive and systematic overview of emerging research themes.

3. Materials and Methods

This study adopted a structured multi-step approach to evaluate 1840 publications on actuators and power electronics from 2005 to 2024. Publications were collected from the Web of Science and were limited to peer-reviewed articles and reviews. Following initial data preprocessing, publication trends by country were examined using Python (v3.11.8)-based analysis tools. BERTopic was employed to identify and visualize key research topics, with Sentence-BERT used for embedding generation, UMAP for dimensionality reduction, HDBSCAN for clustering, and c-TF-IDF for extracting representative keywords.

To assess shifts in research focus over time, the dataset was partitioned into two intervals (2005–2014 and 2015–2024), and identical analytical workflows were applied for comparative analysis. In addition, a distinct examination of the top 10% most cited papers was performed to identify features that differentiate these works from general research trends. The overall workflow of the research is depicted in Figure 1.

3.1. Data Collection

A systematic data collection protocol was established to improve the validity and consistency of subsequent analyses. Records were retrieved from the Web of Science and filtered to focus on studies directly related to actuators and power electronics, ensuring thematic cohesion and data reliability. The curated dataset spans a 20-year period (2005–2024) and is restricted to peer-reviewed research articles and review papers. In total, 1840 relevant records were compiled and stored in Excel format. The detailed search strategy is provided in

Table 1. Web of Science search criteria.

Search Formula

“TS = (actuatorOR “motion control” OR “servo motor” OR “electromechanical device” OR “piezoelectric actuator” OR “hydraulic actuator” OR “pneumatic actuator” OR “soft actuator” OR “smart actuator”) AND (“power electronic” OR “power converter” OR inverter OR rectifier OR “DC-DC converter” OR “motor drive” OR “switching converter” OR “voltage control” OR “PWM control” OR “power conditioning” OR “energy conversion” OR “electronic drive system” OR “high-efficiency power system”)

.

After records were collected, abstracts were subjected to text mining techniques to identify primary research themes and structural patterns. BERTopic, an advanced transformer-based topic modeling approach, was utilized to generate semantic embeddings, perform clustering, and visualize topic distributions. This approach allowed for a systematic analysis of trends in actuators and power electronics research.

3.2. Preprocessing

BERT-based topic modeling methods leverage pre-trained language models that capture complex contextual relationships, thus minimizing the need for the extensive preprocessing commonly associated with traditional statistical approaches such as LDA. These models are capable of producing robust, context-aware embeddings with relatively little text cleaning. Nonetheless, skipping preprocessing entirely can undermine the quality of embeddings and reduce the reliability of clustering outcomes.

In this study, three primary preprocessing steps were applied to ensure the input data was both consistent and semantically rich. First, duplicate sentences were removed to avoid the inclusion of repeated passages that are commonly encountered in abstract datasets sourced from multiple origins. Second, non-standard characters were filtered by excluding tokens that fell outside the basic Latin Unicode block (ASCII range). This procedure was crucial for eliminating corrupted symbols, atypical punctuation, and language-specific artifacts that are not compatible with English-language NLP frameworks. Third, abstracts deemed too short or lacking substantive content were excluded by discarding those with fewer than 30 words, as these typically do not provide adequate semantic detail for robust topic modeling.

These preprocessing tasks were accomplished using regular expressions and typical text-cleaning utilities in Python (e.g., re, string). Although BERT-based methods simplify many elements of the preprocessing pipeline, these foundational practices remain essential to preserve data quality and ensure rigor in subsequent analytical steps. It is therefore important that the preprocessing protocol be tailored to the linguistic features of the corpus and the research objectives involved.

3.3. Trend Analysis

Trend analysis was performed on 1840 academic papers collected during preprocessing, employing Python and Google Colab. The purpose was to explore how research on actuators and power electronics has progressed over time, generating insight into the technological evolution and supporting strategic decision-making. Additionally, examining long-term research trends enables stakeholders to recognize emerging opportunities, foresee possible challenges, and enhance their strategic advantage within a competitive industrial context.

The evaluation targeted three primary components: (1) Identifying the 10 countries producing the most publications each year, and (2) assessing the aggregate publication output among the 10 leading countries. The results deliver a comprehensive overview of global research activity in actuator and power electronics and assist in determining which regions contribute most prominently to advancements in this domain.

3.4. BERTopic Modeling

BERTopic modeling leverages a transformer-based pre-trained language model for the effective management of complex datasets and the facilitation of transfer learning, which is especially advantageous for integrating novel research applications [42]. In this investigation, BERTopic modeling was used to generate document embeddings, group these embeddings into clusters, extract topic features using a class-based c-TF-IDF method, and apply various analytical visualizations.

Embedding is a technique used to represent the semantic content of words or sentences as low-dimensional vectors, enabling computers to efficiently process and interpret language [57]. In this research, the transfer learning capabilities of pre-trained BERT were leveraged to embed documents concerning dry electrodes into high-dimensional vectors. Sentence-BERT (SBERT) was specifically chosen to perform this embedding. The formula corresponding to this embedding process appears in Equation (1).

$$E=SBERT\left(D\right)$$

(1)

• E = Embedding vector;
• D = Set of documents.

BERT embeddings typically comprise 768 dimensions, which results in highly high-dimensional data. To address the dimensionality challenge, UMAP is implemented to reduce the number of dimensions while retaining the nonlinear structures within the dataset. The formula related to UMAP is presented in Equation (2).

$$E_{UMAP}=UMAP\left(E\right)$$

(2)

HDBSCAN is utilized to cluster the dimensionally reduced embeddings [58]. As a density-based clustering method, HDBSCAN demonstrates effectiveness in noise filtering. The relevant clustering formula can be found in Equation (3).

$$C=HDBSCAN\left(E_{UMAP}\right)$$

(3)

C = The cluster to which a document belongs.

After extracting word frequencies using CountVectorizer (v 1.3), principal keywords for each cluster are identified through a c-TF-IDF-based procedure. The c-TF-IDF method is suitable for emphasizing the most representative terms within each cluster, as it assesses word relevance across entire topics and not solely within single documents [59,60]. Additionally, it determines word importance at the cluster level instead of focusing on individual documents only. The formula for c-TF-IDF is outlined in Equation (4).

$$c-TF-IDF\left(w,c\right)={\displaystyle \frac{TF(w,c)}{{\sum }_{w^{\prime }\in e}TF(w^{\prime },c)}}\times \log \left({\displaystyle \frac{N}{DF\left(w\right)}}\right)$$

(4)

• TF = The frequency of word w in a specific cluster c;
• DF(w) = The number of documents in which the word w appears;
• N = The total number of documents.

To deliver an in-depth analysis of topic structures, several visualization tools were implemented, including the BERTopic Intertopic Distance Map, hierarchical clustering, and examinations of document allocation across topics.

3.5. Comparative Topic Evolution Analysis

At this step, the research data was segmented into two periods—2005–2014 and 2015–2024—to quantitatively analyze the evolution of research topics within actuators and power electronics. This division facilitates a meaningful comparison between two distinct time spans: a preliminary phase likely defined by foundational advancements and the initial integration of materials and control, and a subsequent period characterized by increased focus on AI-based control methods, high-performance materials, and cross-industry convergence.

To maintain analytical uniformity, the same topic modeling pipeline was implemented for each dataset within a unified framework. Subsequently, a three-step procedure was used to examine the structural progression of topics over time. First, c-TF-IDF-based keyword analysis quantitatively tracked shifts in conceptual emphasis within corresponding topics across different periods. Second, intertopic distance maps were generated to assess variations in topic proximity and semantic distribution between timeframes. Third, hierarchical clustering examined how topic groupings and their semantic structures have evolved, either diversifying or becoming more refined over time.

This segmentation and comparative approach enable the framework to move beyond basic frequency analysis, facilitating a more nuanced interpretation of the changing direction and increasing semantic complexity of research topics observed during the past two decades.

4. Results

4.1. Trend Analysis

Annual publication trends in actuator and power electronics research from 2005 to 2024 reveal a steady upward pattern, with more than 100 publications each year since 2019. Publication output peaked at 193 in 2024, marking the highest value within the examined period. These results imply that research on actuator and power electronics has maintained a strong presence in the academic landscape, with recent advances likely driven by growing application fields and wider integration across various technological domains. Figure 2 displays the yearly publication numbers throughout the study period.

A breakdown of publications by country indicates that China (604 papers) and the United States (234 papers) play leading roles, reflecting their prominent positions in actuator and power electronics research. These nations are followed by Japan (121), South Korea (98), Taiwan (83), France (68), Italy (60), India (53), Iran (53), and Germany (46). Figure 3 presents the publication volumes for these top 10 contributing countries. This pattern indicates a pronounced concentration of research among a limited set of countries, with several Asian regions making significant contributions. Such clustering may signal emerging disparities in global technological leadership, highlighting the need to encourage broader international partnerships and effective knowledge dissemination.

These publication trends can be explained by multiple converging influences. Technological advancements, especially the integration of AI into control systems and the increasing use of wide-bandgap semiconductors, have broadened the research landscape. The needs of the industry—in sectors such as electric mobility, automation, and renewable energy—have significantly intensified scholarly efforts in these areas. Additionally, government-driven policy initiatives, like China’s “Made in China 2025,” the U.S. Inflation Reduction Act, and Europe’s Green Deal, have expanded the allocation of research funding and fostered greater academic engagement. Together, these driving forces have resulted in the marked increase in publication volume observed after 2015.

4.2. c-TF-IDF Analysis

4.2.1. c-TF-IDF Analysis: 2005–2014 Period

This study examines the thematic landscape of actuator and power electronics research between 2005 and 2014 by applying BERTopic modeling techniques to abstract-level text data. The analysis produced eight distinct research topics, each represented by characteristic keywords with elevated class-based TF-IDF (c-TF-IDF) values. The following discussion details these core keywords and explains their relevance to each topic. Figure 4 visually summarizes the primary results of the c-TF-IDF analysis for this time frame.

Topic 0: Core keywords such as piezoelectric, magnetic, frequency, ultrasonic, and electrical reflect an intensive research emphasis on energy conversion techniques leveraging smart materials and high-frequency electromagnetic actuation. This topic covers investigations into actuator systems engineered to transform electrical energy into highly controlled mechanical movement, usually via magnetostrictive or piezoelectric effects. Notably, research has progressed from using traditional bulk ceramic parts toward developing integrated multilayer architectures and compact control modules. Considerable advancements have also occurred in the integration of control electronics within actuator platforms.

Topic 1: This theme features heightened research attention directed at adaptive and robust control methods for actuator systems encountering uncertainty and dynamic operating environments. Salient keywords—including drive, adaptive, friction, uncertainties, and algorithm—signal a strong research orientation toward motion control solutions that address nonlinear system characteristics and mitigate the effects of uncertainty. From the mid-2000s onward, scholarly focus shifted from traditional PID controllers to more sophisticated model-based strategies, such as Lyapunov-based and backstepping control schemes, facilitating real-time adaptation of parameters and robust disturbance rejection.

Topic 2: Topic 2 focuses on soft actuator technologies utilizing electroactive polymers (EAPs), as evidenced by keywords such as dielectric, electrical, polymers, membrane, and polymer. These actuators produce deformation in response to electric fields, supporting highly flexible and lightweight motion systems that are particularly relevant for next-generation robotics and biomedical devices. This topic highlights an increasing emphasis on innovations at the material level, especially in dielectric elastomers and stretchable composites. Since the early 2010s, the research trend has moved from rigid EAP devices toward more compliant and hybrid material configurations. Progress in dielectric strength, compliant electrode architectures, and advancements in encapsulation methods has enhanced both actuation efficiency and mechanical durability.

Topic 3: The electrification of transportation systems has promoted ongoing research aimed at integrating inverter-based electric drives and power electronics, particularly within the aerospace and automotive sectors. The frequent occurrence of keywords such as aircraft, inverter, induction, electrical, and rectifier signifies an emphasis on electromechanical energy conversion for propulsion and control purposes. This trend signals a larger transition away from centralized mechanical frameworks in favor of distributed electric actuation systems, enabled by progress in semiconductor switching devices, notably, insulated-gate bipolar transistors (IGBTs). These technological advances have resulted in notable gains in energy efficiency, system performance, and weight savings—considerations that are especially important for aerospace applications where reliability and compact design are crucial.

Topic 4: This topic addresses research related to fluid-powered actuation systems, with an emphasis on hydraulic and pneumatic mechanisms, as suggested by recurring keywords such as valve, engine, hydraulic, scroll, and pressure. These actuation systems achieve mechanical motion through fluid compression or regulated flow and are extensively used in industrial automation as well as mechanical control. The 2000s saw significant progress in the miniaturization of components, the refinement of seal design, and the incorporation of servo valves with digital controllers. Such developments have made it possible for fluid-based actuators to be deployed in more compact and mobile industrial contexts.

Topic 5: Within vehicle dynamics research, there has been a growing focus on actuator systems engineered for motion and stability control, underscored by the frequent usage of terms such as vehicle, yaw, yaw moment, roll, and simulation. This literature underscores the significance of rotational motion regulation, comprehensive system-level modeling, and the assimilation of actuators in managing dynamic stability. Over this timeframe, the deployment of high-fidelity vehicle dynamics simulators has expanded, supporting the design and evaluation of advanced strategies like electronic stability control, steering assist, and active suspension systems. Moreover, the integration of real-time feedback control and model predictive control (MPC) has further improved both road holding and passenger safety, thereby advancing the responsiveness and intelligence of contemporary vehicle platforms.

Topic 6: This topic investigates advancements in energy generation systems, focusing on synchronous machines and pitch control techniques, with central keywords including generator, permanent-magnet (PM) synchronous, synchronous generator, and pitch. The studies primarily address electromechanical energy conversion in power generation for wind turbines and other precision generator systems. In the late 2000s, a notable transition occurred from asynchronous to permanent-magnet synchronous machines (PMSMs), prompted by their higher energy efficiency, reduced size, and increased reliability. At the same time, developments in digital control systems permitted improved precision in managing generator behavior through adjustments in pitch angle and voltage regulation. These developments highlight ongoing efforts to maximize power conversion while maintaining operational stability and flexibility under varying operating conditions.

Topic 7: Initial research in smart materials has emphasized advanced functional materials that react to external inputs such as light, heat, or mechanical forces. Frequently cited terms—graphene, polymer, carbon, photomechanical, and stress—reflect a strong interest in innovative actuation approaches arising from these responsive characteristics. The adoption of graphene served as a pivotal advancement in the field due to its outstanding mechanical properties, high conductivity, and notable flexibility. When used to reinforce polymer matrices, graphene enabled the development of stretchable, light-activated actuators with significant implications for soft robotics and biomedical engineering. This line of research has established foundational progress in photomechanical actuators and graphene-based smart materials, setting the groundwork for future intelligent actuator technologies.

4.2.2. c-TF-IDF Analysis: 2015–2024 Period

This section analyzes recent developments in actuator and power electronics research by applying BERTopic modeling to publications spanning 2015 to 2024. Eleven distinct research themes were identified through c-TF-IDF-weighted keywords, capturing the dominant directions present in the literature. Concise summaries and the associated research implications for each theme are provided below. Figure 5 displays the results of the c-TF-IDF analysis conducted for this timeframe.

Topic 0: This research area prioritizes the creation of advanced control systems that sustain reliable operations under uncertain and complex conditions, with a particular focus on dynamic and nonlinear domains such as vehicle dynamics. Central keywords—such as vehicle, adaptive, uncertainties, robust, and disturbances—underscore the pursuit of resilient and adaptive control methodologies able to uphold system performance despite the challenges posed by nonlinearities and external disturbances. The 2010s saw a marked shift towards model-based adaptive control schemes as well as H-infinity robust controller implementations, both of which achieved greater system robustness in fluctuating environments. Additionally, real-time optimization strategies and data-driven control methods have been introduced to improve system adaptability in environments characterized by unpredictability. This research stream advances robust engineering solutions aimed at ensuring stability and high performance for vehicle control systems operating in complex, real-world settings.

Topic 1: Recent studies have increasingly addressed the design and fabrication of soft, stretchable, and lightweight actuator materials, as indicated by recurring words including graphene, polymer, fibers, liquid, and flexible. A central focus of this topic is on innovating material systems based on nanocomposites and liquid–metal integration, with the aim of advancing actuator technologies for safe and adaptive operation in dynamic applications. In recent years, the domain has transitioned from rigid, polymer-based actuators towards flexible, multifunctional solutions integrating functions such as energy storage, real-time sensing, and self-healing properties. These advances are crucial to the development of next-generation soft actuators for use in wearable technology, biomedical systems, and soft robotics. Broadly, this research direction mirrors a continuing evolution towards human-compatible, high-performance actuation devices constructed from flexible and nanoscale materials.

Topic 2: This topic centers on signal-responsive actuator systems, with a focus on those driven by piezoelectric and dielectric mechanisms, as illustrated by key terms including piezoelectric, dielectric, circuit, frequency, and ultrasonic. The investigation highlights state-of-the-art signal processing methods and the integration of intelligent actuation technologies. Current efforts are aimed at further miniaturizing actuator elements and embedding sensors and feedback controls directly into actuator modules, thereby facilitating real-time responsiveness and improving precision. Significant progress includes new methods for resonance frequency adjustment, the design of advanced multilayered dielectric materials, and enhancements in circuit architecture, all of which support greater energy efficiency and precise actuation. This area exemplifies the intersection of smart materials engineering and advanced control systems, driving the development of high-performance actuators for applications such as precision robotics, biomedical instrumentation, and microelectromechanical systems (MEMSs).

Topic 3: Research in this domain focuses on fault-tolerant and adaptive control strategies within actuator systems for renewable energy contexts. Central keywords such as fault tolerant, solar, sensor, WECS (wind energy conversion system), and adaptive indicate an emphasis on increasing reliability and adaptability in solar and wind systems. Early methods for managing faults were primarily passive; however, recent advancements prioritize predictive diagnostics and real-time adaptive countermeasures. State-of-the-art actuator platforms frequently integrate advanced sensing, system redundancy, and real-time observation to ensure operational stability under fluctuating and hazardous environmental conditions. These developments are critical to the evolution of WECSs and photovoltaic (PV) systems, supporting continuous, effective energy generation and system robustness.

Topic 4: Ongoing research highlights advancements in mechanical and fluidic actuation systems, as confirmed by recurring keywords such as hydraulic, turbine, valve, engine, and pressure. This body of work demonstrates the continued importance of hydraulic actuators in power generation and industrial process automation. While the fundamental principles remain well-understood, the adoption of mechatronic devices and advanced sensors has elevated classic designs into intelligent systems capable of real-time diagnostics and predictive support. These technological improvements have boosted performance and operational reliability in demanding industrial settings, underpinning the evolution of fluid power systems across energy and manufacturing fields.

Topic 5: Keywords such as robot, clutch, gear, mechanism, and muscle reflect ongoing research in mechanical and bio-inspired actuation systems for robotics. This topic covers the design and development of both rigid and compliant actuators, which support articulated movement and emulate biomechanical systems. The evolution of this field from traditional rigid-link mechanisms toward soft robotics and tendon-driven platforms has led to actuation systems that more effectively reproduce human musculoskeletal behavior. Such biomimetic designs improve safety and adaptability during interactions with uncertain or dynamic environments. Additionally, integrating feedback control and embedded sensing technologies has markedly improved motion accuracy, responsiveness, and safety in a range of robotic use cases.

Topic 6: Terms such as aircraft, electrical, insulation, machines, and degradation highlight sustained research on the reliability and longevity of electromechanical systems in aerospace. This topic specifically examines how electrical insulation performs and degrades during extended periods under high-stress operational environments, a concern that is central to more-electric and all-electric aircraft systems. As the aerospace sector advances toward electric propulsion solutions, the need for superior thermal management and innovative insulation materials has grown considerably. To meet these challenges, researchers are developing high-performance polymers and nanodielectric compounds that improve dielectric properties, thermal endurance, and the ability to resist electrical breakdown throughout prolonged operation.

Topic 7: With keywords such as plasma, jet, discharge, tokamak, and vertical, this topic addresses research in plasma actuator technologies and advanced control for use in high-energy physics, including fusion power and propulsion. Early studies targeted jet flow manipulation and aerodynamics in low-density environments, but the focus has broadened to cover plasma confinement, boundary management, and operational reliability in severe electromagnetic conditions. The development of fusion systems—particularly those based on tokamak reactor technology—has increased demand for fast-reacting, robust actuators able to operate under high heat and electromagnetic loads. This area now represents the intersection of plasma science and actuator technology, supporting the design of advanced control approaches for complex aerospace and energy infrastructure.

Topic 8: Characterized by critical terms such as piezoelectric, ferroelectric, phosphorene, strain, and polarization, this topic demonstrates notable progress in developing smart materials with coupled electromechanical responses. The research prioritizes advancing strain-sensitive materials capable of converting mechanical strain into electric polarization, with particular focus on ferroelectric materials and novel two-dimensional (2D) compounds. The field has transitioned from conventional bulk ferroelectric ceramics to atomically thin materials such as phosphorene, MoS₂, and hexagonal boron nitride (hBN), which display exceptional piezoelectric and flexoelectric behavior. These developments have expanded the prospects for nanoscale actuator technologies, providing enhanced precision, flexibility, and integration opportunities for advanced sensors, actuators, and energy-harvesting devices.

Topic 9: Encompassing terms like magnetic, superelasticity, microwire, hysteresis, and strain, this topic centers on actuator systems utilizing magnetic fields and elastic phase transitions, such as shape-memory and superelastic effects. These systems achieve reversible mechanical shape changes when subjected to external magnetic or mechanical inputs, making them well-suited for high-precision control and adaptable structures. The discipline has progressed from conventional magnetostrictive actuators to include advanced materials like NiTi-based superelastic wires, Fe-Ga alloys, and magnetic microwires. Current investigations consider hybrid actuation concepts that combine magnetic manipulation and shape recovery, with the aim of improving performance, adaptability, and multifunctionality in rapidly changing operational contexts.

Topic 10: Terms such as ethernet, imagery, unit, operations, and EtherCAT illustrate the increasing integration of actuator systems with real-time communication protocols and embedded control platforms. This topic underscores the growing significance of deterministic networking solutions—particularly EtherCAT—in orchestrating distributed actuator networks within industrial automation. Recently, control system architectures have evolved from traditional centralized controllers to distributed actuator networks supported by high-speed, low-latency communications. This shift has supported the seamless convergence with machine vision, edge computing, and cyber-physical system frameworks, enabling greater responsiveness, adaptability, and scalability in automation systems.

4.2.3. c-TF-IDF Analysis: Top 10% Most Cited Publications (2005–2024)

To further clarify the technological domains that have demonstrated the most significant academic and industrial impact, this section conducts a c-TF-IDF-driven analysis of the top 10% most cited publications in the dataset from 2005 to 2024. BERTopic modeling was implemented on this citation-weighted subset to identify leading thematic research trends. Eight discrete topics emerged, represented by keywords indicative of substantial contributions to smart materials, robust control methodologies, vehicular actuation, and plasma-driven systems. Figure 6 depicts the topic word scores associated with these influential publications.

Topic 0: Fundamental terms such as piezoelectric, electromechanical, electrical, dielectric, and material indicate a pronounced focus on smart material utilization for precise actuator engineering. This topic emphasizes electromechanical transduction processes involving piezoelectric and dielectric materials, which facilitate nanoscale motion manipulation, enhance energy efficiency, and allow integration with biomedical and micro-robotic devices.

Topic 1: Noteworthy keywords such as uncertainties, adaptive, robust, disturbance, and nonlinear emphasize scholarship targeting robust control systems that can accommodate dynamic and uncertain conditions. This topic exemplifies the development of advanced robust and adaptive control techniques aimed at maintaining system stability in the presence of parameter variations and external disturbances.

Topic 2: Key descriptors, including vehicle, yaw, in-wheel, allocation, and regenerative, point to research on automotive actuation and control architectures. Investigations in this domain address in-wheel motor synchronization, regenerative braking optimization, and advanced yaw management for electric as well as hybrid vehicles, thereby improving operational safety and performance metrics.

Topic 3: Recurring keywords such as graphene, carbon, photothermal, pnmms, and chemical reflect continuing investigations into nanostructured and photothermally active materials. This topic encompasses state-of-the-art studies on carbon-based actuation mechanisms responsive to thermal or chemical triggers, mainly within soft robotics and wearable technology applications.

Topic 4: Lexical items like microgrids, secondary, networks, fault-tolerant, and restoration signify research geared towards distributed energy infrastructures with inherent fault resiliency. This topic pertains to the development of actuator-embedded secondary control solutions that provide expedited system recovery, precise voltage control, and robust operation in smart grid and microgrid contexts.

Topic 5: Essential terms such as aircraft, insulation, electrical, machines, and hydraulic identify aerospace investigations where actuator dependability and insulation integrity are paramount. The topic focuses on electrical actuation frameworks for aircraft, with emphasis on high-capacity electrical drive systems, analysis of insulation degradation, and the evolving shift from hydraulic to electromechanical actuators.

Topic 6: Representative keywords such as induction, isolation, residuals, inverters, and residual indicate a focus on high-power inverter systems and associated isolation techniques. This topic examines safety protocols, strategies for mitigating electromagnetic interference, and methods of current suppression within industrial and vehicular inverter applications.

Topic 7: The inclusion of terms like plasma, velocity, jet, discharge, and separation signals research centered on plasma-based actuator technologies for fluid and energy management. Studies in this topic address flow separation mitigation, propulsion system enhancement, and jet control through plasma discharge mechanisms, with demonstrated applications in aerospace and fusion energy contexts.

These results emphasize the pivotal impact of smart materials, resilient control architectures, and energy-efficient actuator technologies which are discussed in the most influential publications. In particular, the significant roles of Topics 0, 1, and 2 demonstrate that interdisciplinary methodologies integrating materials science, control theory, and mechatronics have driven key advances within the field. Figure 6. Topic-specific c-TF-IDF keyword scores calculated from BERTopic analysis of the top 10% most cited papers (2005–2024). The chart highlights the central themes in highly cited research, including smart actuator materials, robust adaptive control, vehicle systems, and plasma-driven flow control technologies.

4.3. BERTopic Modeling Analysis

4.3.1. BERTopic Modeling Analysis: 2005–2014 Period

Within the BERTopic modeling process, an intertopic distance map was generated from the 2005–2014 dataset. This two-dimensional representation displays the semantic relationships among topics based on c-TF-IDF embeddings using UMAP (Uniform Manifold Approximation and Projection) for dimensionality reduction. Each circle signifies a topic, with its size reflecting the volume of documents associated with it. Figure 7 visually summarizes the topic landscape and their semantic proximities during the period analyzed.

More specifically, the axes (D1 and D2) are abstract semantic dimensions created by UMAP that maintain original high-dimensional relationships between topics. As such, the axis values do not represent direct variables but rather signify the semantic similarities or distinctions between topics.

The intertopic distance map is presented in four quadrants, each of which may highlight different thematic domains: Central region (quadrant intersection): Topics located at the center, such as Topic 0 and Topic 1, correspond to foundational, widely researched domains throughout the assessed period. The close grouping of Topics 0 (energy-efficient actuator control) and 1 (adaptive actuator systems) indicates substantial methodological and application overlap, especially within control system optimization and actuator engineering.

Upper-right quadrant (D1+, D2+): Encompasses Topics 3 and 6, emphasizing electrical and aerospace actuation systems, such as inverter-based drives and synchronous generator controls. Their close positioning suggests shared technological approaches or common application sectors, particularly within aerospace and vehicular electric domains.

Left quadrant (D1−, centered D2): Topics 2 and 7 are grouped here, emphasizing studies on advanced materials and piezoelectric technologies. Their marked separation from other clusters highlights a focused, material-science-centric research emphasis—for example, an emphasis on dielectric elastomer actuators and graphene-based smart materials.

Lower central region (centered D1, D2−): Topic 4 appears in isolation, denoting hydraulic- and fluid-driven actuation technologies. Its distance from other topics points to a lower semantic affinity with electric- and material-based actuator research, emphasizing distinctive engineering challenges and related application spaces.

Topic 0 continues as the leading research focus, as indicated by its large representation, signifying a significant volume of work on energy-efficient conversion and refined control techniques. Topic 1 is also prominent, evidencing considerable attention to adaptive control approaches under uncertain operational scenarios. In contrast, the lower-represented topics—such as Topics 5 (vehicle dynamics control), 6 (generator control), and 7 (photomechanical materials)—reflect more specialized or nascent research directions during this interval.

In summary, this intertopic distance map demonstrates a moderate degree of thematic differentiation, with material-intensive actuator research (Topics 2 and 7) clearly separated from those oriented towards system and control approaches (Topics 0, 1, 3, and 6). The visualization thereby delineates the disciplinary boundaries and overlaps central to the evolution of actuator and power electronics scholarship between 2005 and 2014.

4.3.2. BERTopic Modeling Analysis: 2015–2024 Period

To investigate contemporary actuator and power electronics developments, an intertopic distance map was constructed from the 2015–2024 data. In this map, each circle corresponds to a topic, with circle size indicating the quantity of documents within that topic. Figure 8 visually conveys the semantic relationships, topical diversity, and trend convergence among topics.

The D1 and D2 axes of the map are latent dimensions produced by UMAP, designed to maintain semantic separation among high-dimensional topic vectors. Accordingly, the axes reflect semantic relatedness or divergence across topics, rather than mapping to explicit physical features.

Interpretation by quadrants and clusters can be summarized as follows: Central area (quadrant intersection): Dominated by Topic 0, this region captures essential research within actuator control systems and demonstrates extensive thematic intersections with neighboring topics. The prominence of Topic 0 highlights its pivotal status and extensive relevance across diverse actuator technologies and practical implementations.

Upper-right quadrant (D1+, D2+): Clearly identifies Topics 1, 8, and 9, which are primarily related to advanced materials and evolving actuator technologies, including graphene-based composites, ferroelectric materials, and photomechanical actuation. Their distribution highlights a particularly focused and differentiated research trajectory that contrasts with conventional actuator control system research.

Left quadrant (D1−, centered D2): Consists of Topics 2, 3, and 6, which emphasize domains such as electromagnetic actuation, renewable energy integration, and advanced simulation strategies. Their spatial proximity on the map suggests their methodological overlap and shared application areas, particularly within energy-related systems and detailed actuator modeling approaches.

Lower-right quadrant (D1+, D2−): Incorporates Topics 4 and 7, most likely representing mechanical-, hydraulic-, and plasma-based actuation approaches. Their placement delineates a focused thematic grouping, indicating distinct engineering strategies and unique obstacles encountered in high-energy and fluid-driven actuator implementations.

The increased spatial dispersion observed in this map relative to the 2005–2014 period points to a marked rise in topic diversification and domain specialization in contemporary research. The clear delineation of clusters, including those oriented toward materials (Topics 1, 8, 9) and those centered on energy/control (Topics 2, 3, 6) emphasizes the widening interdisciplinary nature and complexity within actuator and power electronics research. Additionally, topics located centrally (0, 5, 10) represent progressive enhancements in core research directions, underscoring continuous foundational development in actuator system studies.

4.3.3. BERTopic Modeling Analysis: Top 10% Most Cited Publications (2005–2024)

To explore the semantic structure underlying influential scientific contributions, an intertopic distance map was constructed via BERTopic modeling applied to the top 10% most cited studies published from 2005 to 2024. This two-dimensional map presents the spatial arrangements among topics based on UMAP-reduced c-TF-IDF representations, with each circle’s size corresponding to the relative impact of its topic in the citation-weighted corpus. Figure 9 illustrates the resulting configuration.

Topic 0 is centrally and prominently located with the largest bubble size, indicating its status as the primary domain for high-impact publications. Its proximity to Topics 4 and 5 implies strong thematic interrelationships with bordering areas, potentially encompassing actuator design integrated with system-level control architectures and power conversion frameworks.

Topic 1 and Topic 2 are closely positioned in the upper-right quadrant, signifying a cohesive sub-cluster that emphasizes robust control strategies and adaptive regulation amid uncertain system dynamics. Their close arrangement highlights conceptual convergence in algorithmic methods for actuator optimization and enhancing system resilience.

By contrast, Topic 6 and Topic 7 are found in the lower-right quadrant, distinctly set apart from the main cluster. This spatial distinction underscores their semantic uniqueness and suggests associations with emergent or niche technologies such as plasma actuation, cutting-edge propulsion systems, or alternative energy mechanisms. Although these topics are less prevalent, their isolation indicates a high level of specialization and the introduction of innovative conceptual models.

Topic 3 is positioned further to the left of the central axis, signifying a peripheral thematic area likely concentrated on materials science topics such as photothermal actuators or carbon-based devices. Its minimal overlap with central topics emphasizes its distinct role in actuator advancements through nanomaterial incorporation.

In summary, the map displays a somewhat centralized but progressively varied thematic landscape. The dominant clusters represent leading topics in control systems and electromechanical integration, while outlying topics highlight specialized breakthroughs with substantial citation influence in targeted fields. This arrangement indicates that influential work in actuator and power electronics is driven by key engineering innovations as well as pioneering research at the intersections of materials and energy systems.

Figure 9 Intertopic distance map generated from BERTopic analysis of the top 10% most cited publications (2005–2024). Circle sizes denote topic prevalence, while spatial distances indicate semantic dissimilarities. Core topics (e.g., Topics 0–2) exhibit strong interconnectivity, whereas peripheral topics (e.g., Topics 6–7) represent emerging research frontiers.

4.4. Hierarchical Clustering Analysis

4.4.1. Hierarchical Clustering Analysis of Topics: 2005–2014 Period

Hierarchical agglomerative clustering (HAC) was performed on topic embeddings generated by BERTopic to clarify the hierarchical organization of topics from 2005 to 2014. Figure 10 displays the dendrogram which depicts the semantic relationships among topics. In this visualization, the x-axis represents linkage distance, reflecting semantic similarity among topics—shorter linkage distances indicate closer semantic affinity. The y-axis presents topics and their associated keywords, facilitating straightforward thematic analysis.

The dendrogram reveals two primary clusters: Materials cluster: Encompasses Topic 7 (graphene, polymer, carbon) and Topic 2 (dielectric, electrical, polymer). This cluster underscores a robust foundational relationship in advanced materials research, with particular emphasis on innovative polymers, graphene, and technologies related to soft actuators. The proximity of these topics signifies shared research interests in material properties and advancements in smart materials.

Systems and control cluster: Comprises the other six topics, centering on actuator system design, control strategies, and mechanical actuation. This cluster can be further divided into three well-defined subgroups: Control and signal processing: Topics 0 (piezoelectric, magnetic, frequency) and 1 (drive, adaptive, friction) create a subgroup focused on actuator responsiveness and sophisticated control algorithms under varying operational conditions.

Electrical drive systems: Topics 3 (aircraft, inverter, induction) and 6 (generator, pm synchronous) are closely linked, reflecting a concentration on electric propulsion, power conversion processes, and generator system technologies, particularly within aerospace and automotive domains.

Mechanical and hydraulic systems: Topics 4 (valve, engine, hydraulic) and 5 (vehicle, yaw, yaw moment) illustrate research centered on fluid-powered actuation and vehicle dynamic controls, emphasizing the engineering and operational principles of mechanical systems.

This hierarchy delineates two central research axes from 2005 to 2014: the advancement of materials for actuators and the holistic integration and control of actuator systems. The presence of distinct subgroups within these clusters points to targeted research domains embedded in broader themes, suggesting an early phase of integrating innovative material technologies with conventional actuator systems, preceding the wider thematic expansion observed in subsequent research.

4.4.2. Hierarchical Clustering Analysis of Topics: 2015–2024 Period

Latent hierarchical relationships among current topics in actuator and power electronics research were examined by applying HAC to BERTopic-derived topic vectors for the 2015–2024 dataset. The resulting dendrogram (Figure 11) depicts the semantic closeness among the eleven topics, based on keyword patterns and vector similarities. In this dendrogram, the x-axis represents the linkage distance to indicate the level of semantic correlation, while the y-axis catalogs topics with their key representative terms.

The clustering outcome identifies two major clusters that are both prominent and thematically well-defined:

Cluster A includes Topics 1 (graphene, polymer, fibers), 2 (piezoelectric, dielectric), 8 (piezoelectric, ferroelectric), and 9 (magnetic, superelasticity). These topics form a closely connected cluster, reflecting a concentrated emphasis on advanced functional materials and innovative actuator mechanisms. Research in this cluster demonstrates substantial progress in smart polymers, nanostructured piezoelectrics, ferroelectric compounds, and magnetostrictive technologies. The semantic closeness of these topics indicates a unified material-science-driven research direction focused on electromechanical interactions and emerging actuation technologies.

Cluster B, which is larger and more structurally varied, includes Topics 0 (vehicle, adaptive, uncertainties), 3 (fault-tolerant, solar, sensor), 4 (hydraulic, turbine, valve), 5 (robot, clutch, gear), 6 (aircraft, insulation, degradation), 7 (plasma, jet, discharge), and 10 (ethernet, imagery, unit). These topics represent applied research themes encompassing robust control under uncertainty, energy-efficient systems, hydraulic and mechanical actuation, robotics, aerospace applications, and real-time networked industrial systems. Notably, Topic 10, although semantically related, shows relative isolation, suggesting a specialized subfield in embedded systems and industrial Ethernet-based control networks.

In comparison to the 2005–2014 timeframe, this clustering analysis reveals a more nuanced and multilayered topic structure, highlighting a transition toward increased interdisciplinarity and technical specialization. Earlier research was primarily focused on piezoelectric materials and control methodologies, while the 2015–2024 era produced two distinct thematic domains: (1) advanced functional materials for actuator engineering (Cluster A), and (2) complex system integration involving AI-driven control, robotics, and distributed actuation frameworks (Cluster B). Furthermore, stronger interrelations between Topics 0 and 3 signify heightened research activity in fault-tolerant control for uncertain or distributed systems, emphasizing the field’s increasing sophistication and application-oriented trajectory.

4.4.3. Hierarchical Clustering Analysis: Top 10% Most Cited Publications (2005–2024)

For a deeper investigation of the semantic organization of highly cited research, HAC was performed on BERTopic-based topic embeddings for the top 10% most cited publications from 2005 to 2024. The corresponding dendrogram in Figure 12 illustrates how these top-cited topics are organized by their semantic relatedness. Relative to the overall dataset, the observed clustering structure here is more concentrated and exhibits deeper hierarchical nesting, indicating tighter thematic alignment among highly influential studies.

The clustering analysis identifies three main groupings. The first cluster consists of Topic 0 (piezoelectric, electromechanical, electrical), Topic 1 (uncertainties, adaptive, robust), and Topic 2 (vehicle, yaw, in-wheel). These topics collectively emphasize the integration of smart materials with advanced control methodologies in dynamic systems. Their strong semantic alignment indicates a unified research trend prioritizing adaptive and robust control methods, with particular relevance to vehicular systems utilizing electromechanical actuators.

The second cluster connects Topic 3 (graphene, carbon, photothermal) and Topic 7 (plasma, velocity, jet), both representing innovative or emerging actuation technologies. Despite stemming from distinct scientific disciplines—materials science and high-energy physics—these topics converge in their investigation of novel physical mechanisms for actuation. This cluster illustrates the significance of advanced materials and physical mechanisms, such as photothermal actuation and plasma discharge, in driving technical advancements in actuator technology.

The third major cluster encompasses Topic 4 (microgrids, secondary, networks), Topic 5 (aircraft, insulation, electrical), and Topic 6 (induction, isolation, residual). Although these topics display greater diversity within the cluster, they all pertain to reliability-driven applications in energy and aerospace domains. Investigations in this group include smart grid management, the durability of electrical insulation, and the operation of high-power inverters—areas where power electronics and actuator solutions converge to achieve dependable system operations.

Overall, this hierarchical arrangement indicates that the most influential research in actuator and power electronics centers around three core themes: (1) intelligent actuation systems combining robust control techniques and smart materials, (2) novel actuation platforms utilizing unconventional physical phenomena, and (3) power system applications emphasizing energy efficiency and reliability. This structure highlights the importance of interdisciplinary innovation and the integration of system-level approaches in shaping leading research contributions in this area.

4.5. Comparative Topic Evolution Analysis

4.5.1. c-TF-IDF Keyword Comparison

To assess shifts in research emphasis within the actuator and power electronics domains, we performed a comparative topic analysis for two intervals: 2005–2014 and 2015–2024. Utilizing BERTopic-derived c-TF-IDF models, this analysis uncovers significant thematic and technological developments that mark the field’s dynamic progression.

During the 2005–2014 period, research was predominantly centered on core actuation technologies and the fundamental properties of materials. The frequent appearance of keywords such as piezoelectric, magnetic, ultrasonic, and dielectric suggests a concentrated interest in energy conversion mechanisms and signal responsiveness using established smart materials. Core research topics encompassed piezoelectric and electromagnetic actuators, developments in adaptive control algorithms, and the initial exploration of electroactive polymer systems. Investigations during this era generally implemented rigid structures with first-generation materials, predominantly targeting improvements in actuator efficiency and control precision for conventional mechanical applications.

The 2015–2024 period, by comparison, presents a marked evolution toward multifunctional capabilities and the miniaturization of materials. Newly prominent terms like ferroelectric, phosphorene, superelasticity, and strain signal significant advancements in 2D nanomaterials and strain-coupled composites. Research topics such as graphene-enabled systems, stretchable polymers, and hybrid material platforms emphasize a shift toward integrating sensing, actuation, and energy harvesting in flexible and biocompatible designs—ideal for applications in wearable technology and soft robotics. Such developments illustrate the intersection of advanced actuator engineering and cutting-edge material science.

Simultaneously, the range of application domains broadened considerably. In earlier years, studies typically focused on established sectors such as aerospace, automotive, and industrial automation, as indicated by keywords like vehicle, yaw, inverter, and pitch that highlighted advancements in propulsion, dynamic control, and motor drive technologies. In the following decade, research extended into high-energy and multidisciplinary fields, notably including plasma actuation (with keywords like plasma, discharge, jet, tokamak) and fusion technology. This shift implies increased research activity targeting challenging operational environments and integrating actuator innovations in fusion energy, high-voltage platforms, and advanced space systems.

A significant trend was the emergence of robotics and biomechanics as major research directions for actuators in the latter period. Whereas these themes were largely missing from the prior decade, keywords such as robot, muscle, clutch, and mechanism formed new research clusters between 2015 and 2024. This emergence demonstrates the growing adoption of bio-inspired actuator designs in soft robotics and wearable devices, highlighting the expanding interdisciplinary scope of the field. Concurrently, the increased appearance of terms like ethernet, ethercat, and imagery indicates a rising focus on real-time industrial communication networks, evidencing the integration of actuation systems with advanced networking and cyber-physical infrastructures.

Another notable transition is observed in the development of control strategies. Initially, research with a control-oriented focus addressed adaptive mechanisms suited for dynamic environments, employing terms such as adaptive, friction, and vehicle. This focus later expanded to encompass the themes robust, uncertainties, disturbances, and fault-tolerant, reflecting increased complexity. These emerging topics highlight the advancement of resilient, real-time, and self-diagnosing control architectures that maintain operational reliability despite uncertain conditions, especially within autonomous systems and renewable energy platforms. The increasing integration of predictive algorithms, digital twins, and sensor fusion underscores that actuators are progressively evolving into intelligent units capable of adaptation, learning, and anticipatory fault management.

Collectively, this comparative analysis reveals a distinct progression from approaches centered around materials and control to an intricate research spectrum that incorporates smart materials, system-level integration, and intelligent control methodologies. Research on actuators and power electronics has advanced notably, now addressing both the innovation of novel devices and their efficient deployment within multifaceted, data-centric, and interconnected environments spanning diverse practical domains.

Table 2. Cross-theme shifts in actuator and power electronics research.

Cross-Theme	2005–2014 Emphasis	2015–2024 Shift
Materials	Conventional piezoelectric and dielectric	Graphene, ferroelectric materials, 2D hybrid structures
Control	Adaptive methods, nonlinear compensation strategies	Robust, AI-supported, fault-tolerant
Applications	Aerospace engineering, industrial applications	Soft robotics, renewable energy systems, biomedical engineering
Integration	Mechanical–electrical coupling	Mechatronic–cyber-physical– biological convergence

depicts the shifting research priorities in actuator and power electronics from 2005–2014 to 2015–2024, focusing on four principal dimensions: materials, control strategies, application areas, and system integration.

Initial research efforts concentrated on widely used materials such as piezoelectric and dielectric compounds, while more recent work has placed emphasis on advanced materials including graphene, ferroelectrics, and 2D hybrids, which provide enhanced flexibility and multifunctional capabilities. Simultaneously, control methodologies have evolved from primarily adaptive and nonlinear techniques to advanced, AI-driven, and fault-tolerant approaches suitable for intricate and unpredictable operational scenarios.

There has been substantial diversification in application domains. Whereas early studies targeted aerospace and industrial contexts, more recent research has expanded to include soft robotics, renewable energy, and biomedical applications, reflecting changing technological opportunities and emerging societal needs.

System integration practices have developed from basic mechanical–electrical coupling to sophisticated architectures that incorporate mechatronic, cyber-physical, and biological systems, facilitating greater interactivity and autonomy in current actuation technologies. This evolution signifies both the increased maturity of the field and its broadened interdisciplinary engagement.

Table 3. Impact-oriented comparison of research topics: general vs. highly cited publications.

Topic Area	Comprehensive Dataset (All Publications)	Top 10% Highest Cited Publications	Remarks/Interpretation
Smart Materials	Piezoelectric, dielectric, ferroelectric, phosphorene	Piezoelectric, electromechanical, dielectric, material	Common theme in energy-efficient nanoscale actuator material technologies
Robust Control	Adaptive, uncertainties, disturbances	Adaptive, robust, disturbance, nonlinear	Priority given to control algorithms that maintain resilience in uncertain conditions
Vehicular Systems	Vehicle dynamics, yaw control, robotics	In-wheel motor, yaw, regenerative braking, allocation	Principal works address integration strategies for EV control systems
Plasma Actuation	Plasma, tokamak, jet, discharge	Plasma, velocity, discharge, separation	Critical domain for fusion and aerospace propulsion studies
Photothermal Nanomaterials	Not significant	Graphene, photothermal, chemical stimuli	Essential in soft robotics and wearable electronics
Microgrid and Energy Systems	Referenced through WECSs or renewable energy control	Microgrid, restoration, secondary control, fault tolerance	Focus on actuator functionality in distributed energy resilience
Insulation and Reliability	Degradation, insulation properties in aerospace	Aircraft, insulation, degradation, hydraulic	Reliability of high-voltage and aerospace systems
High-Power Converters	Inverter, pitch, frequency control	Induction, inverters, residual, isolation	Emphasis on EMI suppression and fault detection within industrial drive systems

provides a comparative analysis of topic areas identified from both the complete dataset and the top 10% most cited publications within actuator and power electronics research. This comparison uncovers significant variations in thematic focus and research influence, pinpointing technological fields that have achieved notable academic impact and industrial applicability.

Smart materials consistently emerge as a dominant theme across both datasets, with frequent references to piezoelectric, dielectric, and ferroelectric materials. In high-impact publications, there is increased emphasis on electromechanical integration and nanoscale material characteristics, highlighting ongoing interest in energy-efficient and multifunctional actuator technologies.

In robust control, the frequent appearance of terms such as adaptive, disturbance, and nonlinear in the most cited works signals a strong focus on robust control algorithms suitable for uncertain environments. This trend indicates a shift towards control systems designed for dependable operation within complex and dynamic conditions.

Vehicular systems form a prominent high-impact research cluster, especially in studies addressing in-wheel motors, yaw dynamics, and regenerative braking. These research topics underscore the critical role of integrated electric vehicle control frameworks and advancements in motion allocation methods.

Plasma actuation, while moderately represented in the overall dataset, exhibits increased emphasis within the highly cited publications, particularly regarding fusion propulsion, discharge mechanisms, and velocity regulation. This points to a specialized yet significant research area relevant to aerospace and high-energy applications.

Importantly, photothermal nanomaterials—although relatively rare in the overall dataset—feature prominently in the highly cited literature through terms such as graphene, photothermal, and chemical stimuli. These materials are crucial in soft robotics and biomedical engineering, signaling new research directions.

Microgrid and energy systems are addressed more directly in the most cited 10% of publications, especially with respect to restoration, secondary control, and fault tolerance. This demonstrates expanding interest in actuator integration within distributed energy systems and in enhancing system resilience. Studies on insulation and reliability are mainly concentrated in aerospace applications, where challenges related to degradation, insulation, and hydraulic control are key. Highly cited works accentuate long-term system performance and high-voltage protection in critical mission environments.

Finally, high-power converters are frequently associated with terms such as inverter, induction, and isolation in the most cited studies, highlighting focused efforts on electromagnetic interference (EMI) mitigation and fault detection in industrial drive systems. Together, this citation-based analysis demonstrates that while certain topics—such as smart materials and robust control—form the backbone of the literature, others, like plasma actuation, photothermal nanomaterials, and energy system resilience, achieve greater representation in high-impact publications. These observations inform strategic priorities for future research seeking to merge technological progress with increased academic visibility.

4.5.2. Topic Space Visualization

To more thoroughly understand the evolution of research themes in actuator and power electronics studies, we analyzed the intertopic distance maps generated by BERTopic modeling for two periods: 2005–2014 and 2015–2024. This comparative analysis highlights how the organization, significance, and distinctness of research topics have shifted over time.

In the 2005–2014 period, topic structure was comparatively compact, featuring two major topics (Topic 0 and Topic 1) that were centrally located and accounted for a significant proportion of research. These leading topics maintained strong semantic ties to adjacent themes, underscoring a concentrated emphasis on areas like actuator control and power conversion. While secondary topics were somewhat dispersed, they remained connected to the main cluster, demonstrating that the field was unified and centered on several core research avenues during this time.

In contrast, the map corresponding to 2015–2024 shows a wider spatial distribution, suggestive of an increased complexity in thematic variety. While Topic 0 maintained its central and dominant role, several topics—notably Topic 1, Topic 8, and Topic 9—transitioned to more peripheral areas of the semantic map. These outlying topics signify the development of specialized subfields, such as advanced material-based actuation, including those utilizing graphene composites and photomechanical systems. The pronounced separation between these clusters indicates a movement in the field toward a more segmented organization, bringing clearer distinctions between application-focused studies and material science advancements.

Moreover, an increase in the number of discernible clusters is evident in the recent period. Specifically, three principal groups can be identified: (i) control systems and industrial applications (Topics 0, 4, 5), (ii) advanced materials and nano actuation (Topics 1, 8, 9), and (iii) energy- and signal-based actuation systems (Topics 2, 3, 6). The formation of these clusters signifies a notable shift from a generalized research context to one that is increasingly specialized and cross-disciplinary.

In addition, the emergence of semantically isolated topics, such as Topic 6 and Topic 10, over the past decade signifies the rise in specialized or nascent research areas that demonstrate minimal conceptual overlap with established domains. This occurrence is associated with the introduction of innovative technologies, including plasma actuators or real-time networked control systems, which have not yet been fully integrated into mainstream academic discourse.

Overall, the comparative structural analysis of intertopic distances demonstrates a transition from a centrally clustered, thematically unified research landscape to a more diverse and multidimensional framework. This evolution reflects broader developments in the field, driven by rising technological complexity, advancements tailored to specific applications, and increased interdisciplinary collaboration.

4.5.3. Topic Clustering Comparison

To assess the changes in semantic structure within actuator and power electronics research, we evaluated the hierarchical clustering of topics from two distinct periods: 2005–2014 and 2015–2024. The dendrograms, developed using BERTopic embeddings in conjunction with agglomerative clustering, reveal substantial modifications in cluster dynamics, thematic alignment, and disciplinary boundaries.

In the 2005–2014 timeframe, the hierarchical structure exhibited relatively shallow depth, resulting in two principal clusters. The first cluster pertained to areas such as smart materials, specifically piezoelectric- and dielectric-based actuators (Topics 2 and 7). The second cluster comprised system-focused topics that emphasized drive control, hydraulic systems, and vehicle dynamics (Topics 0, 1, 3, 4, 5, and 6). The brief linkage distances among these system-related topics reflected strong thematic coherence and conceptual unity. This organizational model suggests that early research in this field was highly cohesive and predominantly centered around core engineering challenges.

Conversely, the 2015–2024 dendrogram displays a more complex and deeply nested topology, characterized by increased topic dispersion and more pronounced subgroup stratification. The materials science cluster has grown to encompass four closely linked topics (Topics 1, 2, 8, and 9), all focused on advanced smart materials, including ferroelectrics, superelastic mechanisms, and nanocomposite systems. This development highlights greater internal complexity and diversification within material-centered actuator research.

Concurrently, a secondary significant cluster has emerged centered on control systems, robotics, and embedded platforms, absorbing a wider range of topics (e.g., Topics 0, 3, 4, 5, 6, 10). This cluster illustrates the increased volume of research focused on application-driven domains, covering robust adaptive controls, real-time networking (e.g., EtherCAT), and autonomous technologies. The appearance of additional topic branches and greater intertopic separation within this cluster demonstrates a movement toward increased specialization and semantic differentiation among application-oriented subfields.

Additionally, subjects such as plasma actuators (Topic 7) and industrial communication systems (Topic 10)—which were previously missing or scarcely observed—have now evolved into either independent or loosely connected research themes. This development signifies the emergence of novel research trajectories within the discipline. The broader range of topics and the more complex hierarchical structure seen in the 2015–2024 period underscore a structural evolution, as the field shifts away from a focus on unified core themes to greater diversification and the promotion of interdisciplinary collaboration.

5. Discussions

The findings of this study reveal a marked thematic evolution in actuator and power electronics research over the last two decades. In the 2005–2014 period, research predominantly focused on well-established areas, including piezoelectric materials, adaptive control algorithms, and mechanical actuation systems. These domains, characterized by keywords such as piezoelectric, adaptive, and hydraulic, represent a stage aimed chiefly at improving system stability, accuracy, and efficiency through proven materials and methodologies.

In contrast, the 2015–2024 period features a transition towards a more diverse and interdisciplinary set of topics. There has been elevated interest in emerging material systems (such as graphene, ferroelectric, and superelasticity), the integration of robotics, and advances in embedded industrial control systems. This shift reflects an increased demand for actuators offering multifunctionality which are able to operate in progressively complex and dynamic application scenarios.

Research in actuator and power electronics has experienced transformative advancements across four principal domains: materials, control strategies, application areas, and system integration. In materials, there has been a progression from traditional piezoelectric and dielectric compounds to advanced nanomaterials such as graphene, ferroelectric polymers, and 2D hybrids, facilitating greater flexibility, enhanced multifunctionality, and improved scalability. Recent developments in control systems have expanded beyond adaptive techniques to incorporate AI-driven and fault-tolerant architectures that provide reliable performance amid uncertain and rapidly changing environments. The application scope has broadened from classical aerospace and industrial implementations to encompass soft robotics, biomedical devices, and renewable energy systems. These emerging areas require actuators that are compact, energy-efficient, and intelligent, designed to enable advanced human–machine interactions and support sustainable technology solutions. With respect to integration, the field has evolved from simple mechanical–electrical coupling to sophisticated mechatronic, cyber-physical, and biologically embedded system architectures. Collectively, these multidimensional innovations underscore not only the field’s advanced technological state but also its increasing synergy with materials science, embedded computing, and AI.

Findings from BERTopic modeling and clustering indicate a significant shift in the structuring of research topics and organizational patterns over time. In the preceding decade, topic distributions were characterized by a few dominant clusters, resulting in a highly concentrated and vertical research topology. However, during the 2015–2024 interval, the research landscape transitioned to a more decentralized and distributed structure, characterized by the emergence of several distinctive and specialized topic clusters. These clusters are characterized as follows:

• Cluster A: Advanced materials and nanostructures;
• Cluster B: Adaptive and robust control systems;
• Cluster C: Robotic and bio-inspired mechanisms;
• Cluster D: Networked actuator systems and industrial communication protocols.

The increasing diversity of research topics reflects the growing complexity and deepening specialization within the domain, transitioning away from broad thematic investigations to innovations explicitly tailored to distinct applications. Several originally minor or previously underrepresented research areas became central between 2015 and 2024. Notably, topics such as plasma actuation (plasma, discharge, tokamak) and advanced industrial networking technologies (ethercat, ethernet) rose in prominence, representing new frontiers driven by the technological demands of fusion energy, aerospace, and intelligent manufacturing. Progress in these topics demonstrates the discipline’s capacity to respond to evolving industry needs, including operating in extreme environments, ensuring ultra-fast data communication, and supporting intelligent, autonomous systems.

The country-level analysis revealed that research outputs are heavily concentrated within a limited group of nations, notably China and the United States. This concentration reflects regional dominance in advancing actuator and power electronics technologies. However, it also highlights potential disparities in the global exchange and collaboration of knowledge. Broadening international collaborations and supporting inclusive innovation strategies could help achieve more balanced technological advancement worldwide.

A citation-based BERTopic analysis of the top 10% most cited papers from 2005 to 2024 demonstrates the prominence of strategic, system-level topics within high-impact research. These papers frequently emphasize multifunctional smart materials—especially those designed for biomedical and nanoscale actuation—alongside advanced control strategies suited to uncertain and mission-critical situations. Applications in vehicular systems, such as in-wheel motor control and regenerative braking, as well as plasma-based actuators used in aerospace and energy sectors, receive particular attention. Additionally, recent research themes like fault-tolerant controls in microgrids, electrical insulation reliability, and photothermal graphene-based actuators underscore the increasing interdisciplinarity of high-impact work. Such research directions integrate advancements in materials science, control methodologies, and the design of applications, illustrating a pronounced focus on solutions to practical challenges. Further trends—such as enhancing power converter safety, fostering distributed energy system resilience, and advancing real-time communication—reflect the growing need for application-oriented actuator technology.

In light of these findings, future investigations should directly address the most critical challenges facing industries. For instance, actuators required to function reliably under severe thermal conditions are vital in fields like aerospace, nuclear power, and heavy industry, thus prompting research into thermally robust materials such as ceramic composites or high-entropy alloys. As sustainability rises in importance, developing actuators capable of harvesting energy via piezoelectric or electromagnetic mechanisms offers compelling opportunities for powering autonomous and isolated systems. Additionally, the integration of AI-based methods—including predictive maintenance, real-time anomaly detection, and autonomous system calibration—holds considerable promise for improving reliability and performance within smart manufacturing. Pursuing these avenues not only supports validated priority areas but also enables greater synergy between academic research and industrial practice in the evolution of actuator technologies.

Several future research directions are identified based on emerging trends: Cross-disciplinary integration: Future investigations should prioritize interdisciplinary collaborations among material science, mechanical engineering, and embedded systems to facilitate the development of multifunctional, adaptive actuators. Industrial validation: There is a growing need to address the dependability, scalability, and real-time performance of actuators in industrial settings.

Sustainability focus: Subsequent actuator designs ought to consider energy efficiency, recyclability, and environmental impact, especially in alignment with sustainable development objectives. AI integration: Integrating AI methods—such as predictive maintenance, autonomous calibration, and optimization of design—has the potential to significantly enhance actuator intelligence and broaden their industrial applicability.

This study delivers important insights from multiple perspectives. Academic: The trend toward increased specialization and topic segmentation highlights the demand for integrative frameworks linking diverse research areas. Researchers are encouraged to explore convergences between smart materials, embedded control systems, and robotics to tackle the increasing complexity of actuator technologies. Industrial: Growth in areas such as real-time communication protocols, robotics, and autonomous control systems points toward the evolution of advanced manufacturing and cyber-physical systems. Industrial decision-makers should leverage these insights to guide R&D efforts and support product innovation initiatives targeting advanced actuator solutions. Policymaking: The evident regional concentration of research activity underscores the importance of international collaboration. Policymakers should promote cross-border research partnerships and the equitable distribution of innovation resources, and drive the advancement of standardization to promote interoperability for actuator systems globally.

By synthesizing academic, industry, and policy-related perspectives, this study contributes a comprehensive analysis of the actuator and power electronics research domains and suggests key strategic directions for ongoing and future development.

6. Conclusions

This study set out to identify and assess research trends in the actuator and power electronics domains between 2005 and 2024, with a focus on thematic evolution and structural change. As these systems become increasingly vital for advanced control operations, robotics, and sustainable energy infrastructure, charting the progression of academic research becomes essential for steering future technological advances.

Utilizing a natural language processing-based approach employing BERTopic, we analyzed 1840 peer-reviewed abstracts from the Web of Science database. The use of c-TF-IDF weighting, intertopic distance visualization, and hierarchical clustering facilitated the extraction of key research topics, structural patterns, and emerging directions within the academic landscape.

Comparative topic analysis revealed a clear thematic evolution over the two decades. In 2005–2014, key domains such as piezoelectric actuators, adaptive control, and hydraulic systems remained prominent. In contrast, the 2015–2024 period displayed a broader array of topics, including graphene-based soft actuators, fault-tolerant energy systems, robotic mechanisms, and networked control systems utilizing industrial protocols like EtherCAT. These trends emphasize greater interdisciplinarity, technological progress, and expanded application spaces in actuator and power electronics research.

These thematic shifts are further characterized across four fundamental dimensions: materials, control, application, and system integration. The research trajectory moved from an initial focus on conventional piezoelectric materials and adaptive control mechanisms toward advanced nanomaterials, AI-enabled strategies, and cyber-physical systems. The range of applications now extends beyond aerospace and industrial contexts to encompass areas such as soft robotics, biomedicine, and renewable energy, signaling greater interdisciplinarity and integration across smart systems engineering.

Visualization using BERTopic with hierarchical clustering showed a structural transformation within the topic space. During the earlier period, topics exhibited a centralized structure defined by two major axes: control systems and innovations in materials. In the more recent decade, topic organization became more fragmented and layered, including semantically unique clusters like advanced functional materials, resilient control strategies, robotics, and embedded industrial systems. This multidimensional growth, corroborated by intertopic distance mapping, reflected a shift from a unified research core to a more diffuse and specialized landscape. Research areas such as plasma actuation and high-speed industrial communication systems emerged as separate clusters, illustrating the progression toward new research domains.

Hierarchical clustering further substantiated the change in organizational structure from a less complex and cohesive form to a richer, more stratified hierarchy. Earlier research grouped topics under wide domains, notably system control and intelligent materials. More recently, the growth in clustering depth and precision reflects increasing variety, especially in the spheres of materials science and control-oriented research.

Geographically, the analysis indicates that only a few nations—particularly China and the United States—maintain a substantial share of global research production. While this high concentration demonstrates regional dominance, it also highlights the pressing need for enhanced international collaboration to mitigate technological disparities and foster more balanced innovation ecosystems.

Drawing on these findings, several avenues for future research are proposed. Emphasis should be placed on cross-disciplinary integration that merges advances in material science, control engineering, and embedded systems, facilitating the development of multifunctional actuator technologies. For industrial validation, particular attention should be given to real-time performance, operational durability, and actuator scalability within actual application environments. Sustainability must be an essential criterion in actuator design, ensuring alignment with environmental priorities such as energy efficiency and recyclability. Furthermore, the integration of AI-driven control systems featuring ML and autonomous optimization offers notable potential to enhance predictive maintenance and adaptive actuation capabilities.

A citation-based analysis of the top 10% most cited papers demonstrated that impactful research is often centered around integrative and application-oriented topics. These include multifunctional smart materials, resilient and fault-tolerant control system architectures, and critical application areas such as autonomous vehicles and energy system management. Such results underscore the importance of orienting future research toward strategically significant topics that address pressing real-world challenges.

This study additionally offers insights with practical significance into multiple sectors. For academic institutions, the identification of emerging, segmented research themes points to a growing need for comprehensive curricula and interdisciplinary research initiatives. For industry stakeholders, a nuanced understanding of topic evolution and thematic complexity can inform R&D investment strategies and support the commercialization of new technologies. For policymakers, the data-driven findings provide a foundation for advancing international research partnerships, equitable funding allocation, and the establishment of unified technology standards within the actuator domain.

Crucially, the delineated topic clusters correspond with innovation goals across high-priority industries, such as aerospace (e.g., plasma actuators for flow control), biomedical engineering (e.g., soft and stretchable actuators for physical rehabilitation), and advanced manufacturing (e.g., distributed actuator control systems employing industrial communication protocols). These results highlight the importance of creating actionable roadmaps that synchronize academic innovation with industry-specific R&D needs, thus facilitating technology deployment and commercial adoption.

Additionally, this research informs national and international R&D strategies, particularly those associated with the Industry 5.0 agenda and the United Nations Sustainable Development Goals. The observed thematic trends and structural complexity can serve as a reference for designing funding mechanisms and policy instruments that advance global partnerships and foster responsible, cross-border innovation.

From a methodological standpoint, future research could expand this framework by incorporating LLMs for regression-based analyses, modeling cross-topic influences, or predicting topic emergence. These advancements are expected to improve both the interpretability and forecasting capabilities of trend analysis in multifaceted and interdisciplinary areas.

The methodology introduced here exhibits several notable strengths. First, combining BERTopic with c-TF-IDF and hierarchical clustering facilitates context-sensitive topic modeling, which supports the comprehensive tracking of research developments and structural evolution across time. Second, its scalability allows for the efficient handling of extensive datasets, such as thousands of abstracts. Third, the approach’s integration of intertopic mapping and visual clustering delivers accessible and meaningful perspectives for research assessment and strategy.

However, some methodological limitations must be recognized. Sole reliance on abstract-level data may fail to capture the granular details that are present in complete texts. Furthermore, although BERTopic performs well for unsupervised topic extraction, it may not effectively represent interdisciplinary intersections or citation-driven influences unless complemented by additional analytical techniques. Future research could address these gaps by integrating hybrid methods that leverage full-text mining, citation network analysis, or supervised learning models within the current framework.