Current Status and Outlook of Roadbed Slope Stability Research: Study Based on Knowledge Mapping Bibliometric Network Analysis

Abstract

Landslide hazards on roadbed slopes pose significant safety risks, leading to casualties, property losses, and environmental damage. With the rapid expansion of global railway and highway construction, roadbed slope stability has become a critical research focus. However, systematic reviews and prospective studies based on bibliometric analysis in this field remain limited; such a lack is likely to lead to a lag in the theoretical development of the field. To address this gap, this study analyzes 453 papers from 2014 to 2023 using the Web of Science (WOS) core collection and tools like VOSviewer, CiteSpace, and Bibliometrix R. This study focuses on the following: (i) Visualizing research trends through knowledge graphs, covering document quantity, the authors, the countries, and the keywords. (ii) The objectives, the methods, specific objects, and the environmental conditions of the literature in this field are categorized and discussed, and the limitations of numerical simulation and other research shortcomings in this field are pointed out. (iii) Future research directions, focusing on the actual working conditions and utilizing advanced and flexible subroutine functions to simulate complex conditions with multi-physical field coupling, are discussed to ensure the accuracy of this research and the sustainability of road construction development. This paper can help scholars comprehensively and quickly understand the research status and hotspots in the field of roadbed slope research, with a view to providing theoretical support for future research and exploration.

1. Introduction

As the global economy grows and urbanization speeds up, there has been a significant focus on the construction and development of railways and roads, which are crucial transportation infrastructures [1]. According to the data, by 2022, the total length of railways worldwide surpassed 1.15 million kilometers, and construction efforts continue to amplify. The railway network for both passengers and freight is expanding, particularly in developing nations and areas. The scale of road construction, mainly motorways, is also expanding, and many countries and regions around the world are actively promoting the construction and improvement of motorway networks to improve the efficiency and convenience of road transport [2,3].

As a result of large-scale railway and highway construction, road safety has also become a hot issue today, affected by engineering and geological conditions, topography, and human factors. Railway and highway routes inevitably have to go through mountains and a multitude of road slopes. In addition to the construction of tunnels and viaducts, roadbed side slopes in the terrain have to be excavated at high ground to form graben side slopes, and the terrain has to be at a low elevation to construct embankment slopes [4]. Safety accidents caused by unstable roadbed occur from time to time, such as the 7-7 Hangrui Expressway, Anhui section K124+100 (Hangward) slope collapse accident; the 5-1 Meidai Expressway, Fujian direction K11+900m near the highway pavement collapse disaster; the 8-4 Mentougou railway section landslide accident; and so on. It can be seen that landslide disasters have a huge impact on human beings, and we cannot ignore such problems [5].

Roadbed slopes can be divided into embankment slopes and cut slopes, according to previous studies. It is known that the instability of roadbed slopes is mostly caused by heavy rainfall; firstly, heavy rainfall leads to a sharp increase in soil water content [6,7], the soil becomes wet, friction between the particles decreases, and the adhesion of the soil decreases [8,9]. This change makes the soil more prone to slipping. Secondly, rainwater infiltrates into the interior of the soil body, forming a seepage field and generating seepage forces. These forces act on the soil body and increase the dynamic and hydrostatic loads, leading to a decrease in the shear strength of the soil body and a decrease in slope stability [10,11]. However, many studies have shown that train loading, vegetation cover, and excavation disturbances also have an effect on the stability of roadbed slopes [12,13,14]. Figure 1 demonstrates the various factors that contribute to the instability of roadbed slopes. Research on the mechanism of slope instability caused by a single instance of rainfall is relatively well established [15], but the reality of slope instability is a case of multiple factors coupled, so the roadbed slope is still the object of current hot research. This study focuses on the stability of all types of roadbed slope in highways and railroads, including filled slopes, excavated slopes, and composite slopes, with a view to providing a scientific basis and technical support for the design, construction, and maintenance of roadbed slopes.

In summary, the stability study of roadbed slopes is of great significance to protect lives and reduce losses to ensure the normal passage of vehicles, and in recent years, a large amount of research on the stability study of roadbed slopes has been published in academic journals, which has become an important branch of the field of geotechnical engineering [16,17]. Abolfazl Baghbani and colleagues conducted a review of the use of artificial intelligence in geotechnical engineering over the last thirty years, highlighting its extensive applications in assessing landslide risks and determining slope model parameters for road foundations [18]. Zhen and his team examined the advancements in technology for treating roadbeds in expansive soils, providing a review of the methods for highway roadbed treatment in large soil areas, although their focus was solely on technology for widening highway roadbeds [19]. Han and his group employed bibliometric network analysis to study loess slope reinforcement technology, investigating the causes of landslide disasters and identifying the future trends in reinforcement technology [20]. However, within this area of research, there is still a lack of bibliometric analysis of roadbed slope stability, as well as fewer systematic summaries and prospective outlook studies in this research direction, which may lead to researchers having difficulties comprehensively grasping the knowledge and research trends in this field and difficulties in effectively integrating the scattered research results, thus affecting the generation of new knowledge and innovation. As a result, we employ various statistical analysis tools from the literature to visualize and quantitatively assess the pertinent studies in the area of roadbed slope stability research. This paper aims to explore the following research questions (RQs).

RQ1: How has the literature on roadbed slope stability studies evolved and developed?

RQ2: What are the current hot research methods, research objects, and influential literature and researchers in roadbed slope stability research?

RQ3: What are the prospects for future research in this area?

This study aims to address these three specific questions through the use of knowledge mapping techniques. The first question is tackled by providing a summary of the publication count in the field over the last decade, along with identifying the leading countries. The second question is explored through the creation of clustering and co-occurrence maps that highlight the keywords, the institutions, the authors, and other relevant aspects of the literature. Finally, the third question is addressed by summarizing the findings from all the knowledge graphs. The purpose of this paper is to provide scholars with a way to understand the hotspots in the field, to address the lack of research in the field, and to discuss the direction of future research. Through this paper, scholars can better understand the hot issues and research trends in the field of roadbed slope stability research, which will help scholars to more accurately construct their own research direction and adopt sustainable methods to solve the roadbed slope stability problems in practical work.

2. Data and Methods

2.1. Data Sources

This paper searches for articles in the Web of Science (SCIE) core collection database using the database’s keyword search subject range function. Using the Boolean logic operation search technique, the keyword search formula is (TS = (slope of roadbed) OR TS = (subgrade slope) OR TS = (Embankment slope) OR TS = (Cut slope) AND (TS = (stability) OR TS = (risk) OR TS = (safe)) AND (TS = (highway) OR TS = (railway) OR TS = (road)). The earliest research in this category that could be searched on Web of Science was from 2010, and the number of publications from 2010 to 2013 did not fluctuate greatly, with publications in the last ten years representing 89.2% of the total literature. The finalized time frame for analysis was from 1 January 2014 to 31 December 2023. The date of search was 20 August 2024, and the types of literature searched were articles, dissertation theses, and review articles. Through careful screening, irrelevant articles were excluded as much as possible, and finally 453 documents were selected as research objects. The top 10 research areas with the highest number of publications are shown in Figure 2, namely Geosciences Multidisciplinary, Engineering Geological, Engineering Civil, Environmental Sciences, Water Resources Engineering Environmental, Meteorology Atmospheric Sciences, Materials Science Multidisciplinary, Construction Building Technology, and Multidisciplinary Sciences.

2.2. Research Methodology

In this paper, three literature analysis tools, namely CiteSpace, VOSviewer, and Bibliometrix R package, were mainly used to conduct the comprehensive analysis of the literature; in addition, this paper also used software such as Origin and Visio to further visualize the number of publication years and the percentage of publication countries in the literature. Figure 3 shows the general research framework of this paper.

Bibliometrix, VOSviewer, and Citespace are three commonly used tools for bibliometric analysis. Bibliometrix is an open source R package that provides functions such as data loading, format conversion, matrix creation, network analysis, and visualization to help researchers explore bibliometric data, reveal publication relationships, and construct author collaboration networks [21,22,23]. VOSviewer is widely used in scientific research for bibliometric analysis, knowledge graph construction, and disciplinary development trend prediction, which can help researchers quickly understand the current status and development trend of research in the field. In this paper, we used it to create a keyword co-occurrence graph, considering a network containing n nodes [24,25]. In this paper, we used it to make a keyword clustering graph. Consider a network of n nodes. Suppose we want to create a mapping or clustering of these nodes, the following is a simplified general formula for the clustering algorithm (Equations (1) and (2)) [26,27]:

$$S_{ij}={\displaystyle \frac{2m{c}_{ij}}{c_i{c}_j}}$$

(1)

$$c_i={\sum }_{j\ne i}{c}_{ij}andm={\displaystyle \frac{1}{2}}{\sum }_i{c}_i$$

(2)

c_ij denotes the number of links between nodes i and j (c_ij = c_ji ≥ 0). S_ij denotes the strength of the association between nodes i and j [28] where ci denotes the total number of links for node i, and m denotes the total number of links in the network.

Citespace is information visualization software developed based on Java language, which is mainly used for the measurement and visualization analysis of scientific literature [29]. It helps researchers to discover the trends, research hotspots, and literature relationships in academic fields and supports the drawing of maps, such as keyword co-occurrence and literature co-citation, helping researchers to fully understand the development of academic fields [30].

3. Literature Analysis and Discussion

3.1. Analysis of Publication Information

The trends in issuance and country distribution over the last decade were analyzed, as shown in Figure 4. As can be seen in Figure 4a, from 2014 to 2018, the field was in a slow development phase, with a low number of articles published each year. From 2019 to 2021, the number of publications grew rapidly, with a slight decrease in the latter two years. In the last five years, there were more than twice as many average publications than there had been in the previous five years. This suggests that 2019 is an important time point for the development of this field; it is worth noting that in China, the book “Outline of the Construction of a Strong Transportation State” was published by the People’s Publishing House in 2019, an initiative that has, to some extent, promoted and stimulated the development of the research aspect of the stability of roadbed slopes. Figure 4b shows the results of the bibliometrics of roadbed slope stability analysis by country. China produces a lot of research in this area, with a share of 46.58%, which is due to the fact that China is a country with the longest railways and highways in the world and the highest frequency of this type of disaster [31,32]. This is followed by India and the United States, with shares of 18.1% and 6.18%, respectively, and the share of the other countries reaches 22.3%, as shown in Figure 4b. Clearly, countries with a more mountainous terrain and extensive networks of railroads and highways tend to have a greater focus on research in this field. Additionally, the high volume of publications from China could be attributed to policy initiatives, such as the ‘Transportation Power’ program. However, this regional emphasis has resulted in a somewhat limited scope of research. For instance, Chinese researchers often concentrate on slope issues in monsoon climates, whereas Nordic researchers are more likely to study the effects of freeze–thaw cycles. This geographical divergence may impede the development of universally applicable theories.

3.2. Co-Occurrence Network Measurement Analysis

Co-occurrence network econometric analysis is a scientific research method that uses the co-occurrence theory to analyze feature items in the literature in order to reveal the strength of association and information structure between these feature items. Using co-occurrence network econometric analysis can systematically sort out the current research status, hotspots, and trends in the field of roadbed slope instability and provide the basis and direction for subsequent in-depth research. The following is the analysis of the keywords, the countries, and the highly cited literature in this field.

3.2.1. Co-Occurrence Analysis of National and Institutional Networks

Figure 5a shows a country cooperation network graph, with the minimum occurrence threshold set to three. Thirty-one countries and regions with cooperation relationships were finally selected for analysis. A larger sphere indicates a greater number of published articles; a wider line represents a greater number of collaborations and closer ties. The hue of the color signifies the years of collaboration, with the darker shades representing earlier partnerships. Figure 5b shows a collaborative network diagram of the institutions. The minimum occurrence threshold determined is three after manually screening out the final display of the 33 institutions of the collaborative network co-occurrence diagram (due to the maximum institution name length being 19, some institution names are not displayed in full).

As can be seen from Figure 5, countries such as France and Spain have conducted collaborative research in this field at an earlier point in time, while countries such as China and India have collaborated on research at a later point in time, but as can be seen from the size of the sphere, these two countries have contributed the largest amount of literature to this field. From this figure, it can also be seen that China and the United States cooperate the most; in addition to this, China also makes frequent contact with the United Kingdom, Canada, and other countries. The reason for this is that, on the one hand, roadbed slope instability is a global problem, which is widely found not only in China, but also in the United States, Canada, the United Kingdom, and other countries. All these countries have a large number of mountainous and hilly areas, and the stability of slopes in these areas is directly related to the safety of highways, railways, and other infrastructures [33,34]. On the other hand, China, the United States, Canada, and the United Kingdom are influential in the field of scientific research, with a large number of scientific research institutions and high-level researchers. While each country has its own focus in the field of roadway slope instability, there is complementarity in technology. Through cooperation, we can learn from each other, learn from each other’s advanced technology and experience, and jointly promote the technological progress in this field [35]. In terms of quantity, the two institutions contributing the most papers to this field are the ‘Chinese Academy of Sciences’ and the ‘University of Science and Technology of China’, and the frequency of cooperation between these two institutions is also the highest. China’s swift progress in road construction, coupled with the government’s robust backing of scientific research—particularly in geotechnical and geological engineering—creates a unique scenario. The Chinese Academy of Sciences (CAS) excels in fundamental research and theoretical advancements, while the University of Science and Technology of China (USTC) is known for its interdisciplinary and technological innovations. Their collaboration can effectively combine their strengths to address issues related to roadbed slope instability. For instance, the CAS can offer geological insights and mechanical theories, while the USTC can leverage its expertise in computational science and data analysis to create innovative research methodologies and technical tools.

3.2.2. Keyword Co-Occurrence Analysis

Figure 6 shows a keyword co-occurrence diagram of the literature in the field of roadbed slope stability research. In the process of generating the graph, we checked “co-occurrence” in the “Type of analysis” to analyze the co-occurrence relationship between the keywords, and checked “keyword plus” in the “unit of analysis” to cover a wider range of keywords. At the same time, we adjusted the minimum frequency of occurrence to nine to ensure that the keywords with a certain research basis and attention are screened out. In addition, we also excluded some meaningless keywords or keywords with weak relevance to the research topic, and finally obtained an overlay visualization graph consisting of 57 of the most relevant keywords, with a total connection strength of 1085.

From Figure 6, it can be seen that most of the keywords have a relatively stable frequency of occurrence in the last ten years, with no obvious trend of increase or decrease. This indicates that the concentration of research topics is high in this time period, the research direction is relatively stable, and there is no large-scale hotspot transfer phenomenon. Additionally, “stability”, “model”, “behavior”, “design”, “failure”, “landslide”, and other keywords appear most frequently. These high-frequency keywords reflect the core content and the key direction of current research in this field, i.e., they are mainly focused on the construction and validation of models related to the stability of roadbed slopes, the analysis of stability behavior, the optimization of the design method, and the study of the failure mechanism and the landslide phenomenon. The continuous attention on these research topics is of great significance for the in-depth understanding of the stability problems of roadbed slopes, improving the quality of design and construction, and preventing and controlling landslide disasters [20]. In contrast, “cost” and “sustainability” appear less frequently and are not even shown in the co-occurrence diagram, suggesting that the economic and environmental dimensions are not sufficiently integrated at this stage.

3.2.3. Analysis of Highly Cited Literature

Highly cited literature typically reflects the widely acknowledged and accepted research findings within a discipline, often encompassing fundamental theories, essential methodologies, and significant discoveries. By examining these works, one can swiftly understand the core knowledge framework and research trends in the field, as well as identify the current research focal points and potential future directions for development [36,37,38].

The five most frequently cited papers globally are shown in

Table 1. Most globally cited documents.

Paper	References	TC	NTC
INTRIERI E, 2019, EARTH-SCI REV	[39]	175	8.94
CHAMBERS JE, 2014, NEAR SURF GEOPHYS	[40]	121	3.41
TANG L, 2018, COLD REG SCI TECHNOL-a	[41]	108	4.70
JAMSAWANG P, 2016, COMPUT GEOTECH	[42]	105	3.93
ZHANG MY, 2017, APPL THERM ENG	[43]	99	3.56
TCs: Total Citations; NTCs: Normalized Total Citations.

, including the abbreviated titles of the literature, the year of publication, the DOI, the Total Citations, and the Normalized Total Citations. Among them, the most highly cited paper was included in Earth-Science Reviews in 2019, with an NTC value of 8.94 being the highest among these five papers. Normalized Local Citations is a bibliometric concept designed to assess the citation impact of the literature within a particular region or field using a standardized approach. This method addresses the potential biases in citation counts that may arise from variations in the research fields, the publication years, and the publication types. In this context, the NTC primarily focuses on the weighted counts of global and local citations for the literature, the authors, and the sources to derive a value, as illustrated in Equation (3):

$$C=m_{1}L+m_2(a{D}_i+b{A}_i+c{S}_i)$$

(3)

where C is the Normalized Local Citations value; L is the quantified local citations value; m₁ is the local citations parameter; D_i, A_i, and S_i are the quantified literature and author source indices, respectively; m₂ is the cluster index; and a, b, and c are the literature and author source index parameters, respectively.

Of the 453 papers analyzed, the 10 papers with the highest number of local citations are shown in

Table 2. Most locally cited documents.

Document	References	LC	NLC
SINGH R, 2014, B ENG GEOL ENVIRON	[44]	25	6.78
SIDDIQUE T, 2017, ENVIRON EARTH SCI	[45]	19	4.22
BASAHEL H, 2017, J ROCK MECH GEOTECH	[46]	15	3.33
ZHANG MY, 2017, APPL THERM ENG	[43]	15	3.33
VISHAL V, 2017, NAT HAZARDS	[47]	15	3.33
KUNDU J, 2017, J EARTH SYST SCI	[48]	13	2.89
KAINTHOLA A, 2015, GEOSCI FRONT	[49]	12	5.10
PRADHAN SP, 2020, J ROCK MECH GEOTECH	[50]	12	8.05
LIU H, 2016, COLD REG SCI TECHNOL	[51]	11	4.79
SHARMA LK, 2017, J GEOL SOC INDIA	[52]	10	2.22
LCs: Local Citations; NLCs: Normalized Local Citations.

.

The article that received the most local citations was published in the Bulletin of Engineering Geology and the Environment in 2014, but it does not have the highest standardized citation count. It is lower than 8.05, which is ranked eighth, as the NLC combines the global and local citation, accounting for the documents, the authors, and the sources. Compared to the article with the highest number of local citations, the article published by Pradhan et al. in 2020 in the Journal of Rock Mechanics and Geotechnical Engineering has a higher NLC value. Although the article was published at an earlier period of time, resulting in fewer local citations, the journal has a higher impact factor, and therefore a higher NLC value. By summarizing the highly cited literature, it is found that the highly cited literature globally focuses on numerical simulation methods, but the studies with many local citations are more focused on case validation, reflecting the disconnection between academia and the engineering community. For example, although the landslide prediction model proposed by Intrieri et al. [39] is widely cited, its applicability in complex geological conditions still lacks field validation. This is an issue that needs to be focused on in the future.

3.2.4. Author Co-Occurrence Analysis

Author co-occurrence analysis typically involves examining the collaborative network, knowledge transfer, and trends in disciplinary advancement within scientific research by studying the relationships between various authors in the same area. Additionally, when investigating authors in a specific field, those with a high volume of publications often reflect and align with the research hotspots prevalent during a particular phase in that discipline.

Table 3. Top 5 most influential authors.

Authors	Articles	AF	TC
SINGH TN	29	7.57	576
ZHANG MY	16	2.85	383
AI YW	14	1.78	225
PEI WS	11	2.01	320
PRADHAN SP	10	2.86	272
TCs, Total Citations; AF: Articles Fractionalized.

and Figure 7 show the five authors with the most publications and the co-occurrence map of the authors in the field, respectively.

Table 3 shows that SINGH TN has the highest number of articles published in this field, up to 29, and he also has the highest Total Citations (TCs value), as well as Articles Fractionalized (AF), which is a term used to refer to some form of fractionalization of the number of articles per author to reflect their actual contribution to collaborative research [53]. In collaborative research, an article may be co-authored by multiple authors, and the contribution of each author may not be equal. Therefore, by fractionalizing the article counts, the relative importance of each author in overall research output can be more accurately assessed. The synthesis of various factors, such as the number of publications, the TC value, the AF value, and so on, can be seen. The researcher SINGH TN has an important influence on the development of research in the field of roadbed slope instability. A study of author co-citation analysis (co-authorship analysis) revealed that the final co-citation map generated after the proper adjustment of the minimum frequency of occurrence contained 30 authors [54,55,56,57]. In addition, ZHANG MY and AI YW et al. [58,59] are influential authors who usually publish as collaborators. In contrast, some authors complete their research and publish independently. This indicates that collaboration has become a common trend in the current research environment. This collaborative model not only helps to integrate the expertise and resources of different researchers, but also increases the efficiency and impact of research [60].

3.3. Literature Clustering Analysis

Literature clustering analysis is an important text-mining and information-retrieval technique and one of the current advanced literature analysis methods [61]. Its core idea is to group a large amount of literature data according to similarity or correlation between them, so that the literature within the same group has a high degree of similarity in terms of content, topics, research methods, etc., while the literature in different groups has differences [62,63]. Citespace stands out among the knowledge graph mapping tools due to its automatic clustering feature. The following analysis will focus on the 453 selected documents from the cited journals, examining the clustering keywords and a three-field map.

3.3.1. Cluster Analysis of Co-Cited Journals

Figure 8 shows the four main clusters produced by the cited journals in the 453 WOS studies. CiteSpace gives three algorithms (the LSI (Logarithmic Great Likelihood Ratio), LLR (Shallow Semantic Indexing), and USR (User Defined)) for extracting clustered subject terms from the clustered citation documents. The default automatic tagging of words is based on the LLR algorithm [64]. Expression (4) is given below:

$$LLR=log{\displaystyle \frac{p\left(C_i|V_{ij}\right)}{p\left({\overline{C}}_j|V_{ij}\right)}}$$

(4)

where LLR is the log-likelihood of the word Wi of category C_j, and $p\left(C_i|V_{ij}\right)$ and $p\left({\overline{C}}_j|V_{ij}\right)$ are density functions in categories C_j and ${\overline{C}}_j$ [56,57]. And the core of the LSI (Light Semantic Indexing) algorithm is based on Singular Value Decomposition (SVD) to downscale the document–word matrix so as to capture the potential semantic relationships between words. The expression of Singular Value Decomposition is as follows [65]:

$$mathbfA=U\sum V^T$$

(5)

where A is the document–vocabulary matrix, U is the left singular vector matrix (principal component in the vocabulary space), Σ is a diagonal matrix whose elements are singular values (in descending order), and V^T is the transpose of the right singular vector matrix (the principal component in the document space).

Compared with the LSI algorithm, the LLR algorithm has less influence on data noise and can resist the interference of noise to a certain extent, so as to extract more accurate clustered subject terms. Therefore, the LLR algorithm was chosen for the cluster analysis of the co-cited journals in this paper, which were finally divided into four themes: case study, rockfall hazard, permafrost region, and soil loss.

Case studies are an important part of roadbed slope instability research, which can reveal the specific causes, processes, and influencing factors of slope instability [9,66]. Although numerical simulation can simulate complex physical processes and boundary conditions, the results often depend on the accuracy of the model and the reasonableness of the parameters. The lack of validation by case studies may lead to a large deviation between the simulation results and the actual situation, making it difficult for theoretical research to effectively guide practice [67,68]. On the one hand, in the process of scientific research, case analysis combined with numerical simulation can make the research results more accurate and more in line with reality; on the other hand, through the analysis of actual cases, effective prevention, management, and emergency measures are extracted to directly guide engineering practice, reduce the risk of slope instability, and ensure road safety [69,70]. However, the current researchers utilizing cases for analysis mostly focus on “after-action analysis” and lack the prospective design of early warning mechanisms. For example, although the review of the Hangrui high-speed landslide accident verified the drainage design flaws, it did not propose a dynamic risk assessment framework, which led to the recurrence of similar accidents.

The analysis of the 453 papers reveals that the journals referenced in relation to rockfall hazards and soil losses indicate that roadbed slopes consist of both rocky and soil components. The journals dedicated to “rockfall hazard” and “soil losses” concentrate on research pertaining to rockfall and soil erosion, addressing various topics such as mechanism exploration, risk assessment, and the design and execution of protective measures [71,72]. These studies are of great significance to ensure the safety and sustainable development of human society.

Roadbed slope instability is a complex problem involving many disciplines, such as geotechnical engineering, geological engineering, and environmental engineering. As a special geological environment, the permafrost zone has soil properties, hydrological conditions, and climatic characteristics that are closely related to the stability of roadbed slopes [73,74]. Therefore, the study of roadbed slope instability within the permafrost zone can help us to deeply understand the mechanism of geological hazards in the region and also provide an important theoretical basis and technical support for the design, construction, and maintenance of roadbed engineering. With respect to this cross-discipline and its relevance, research papers in the field of roadbed slope instability are naturally noticed and cited by journals in the field of “permafrost zone” [75].

3.3.2. Keyword Clustering Analysis

Keyword cluster analysis can bring together similar keywords and themes, thus helping us to identify the intrinsic connections and logical order between the different parts of a review. This helps to construct a well-organized and logical framework for the review. Figure 9 shows a keyword clustering timeline graph of 453 documents, and the clustering algorithm used is the LSI algorithm. Since the LSI algorithm can reveal the implicit semantic relationships between texts, it can more accurately identify the keyword clusters with similar topics or research foci during the clustering process [64]. Compared with the traditional clustering algorithms, the LSI algorithm shows higher efficiency and accuracy when dealing with high-dimensional and sparse text data [76]. In order to improve the readability of the knowledge graph, we choose the MST pruning algorithm, and the strategy is to choose “pruning the merged networks”, which cuts out the relatively unimportant nodes and connections to make the focus more prominent.

This research was categorized into the following main themes by examining and summarizing the keywords from the clusters:

(1)

Instability mechanisms and risk assessment and kinetic analyses (different research objectives)

Among the 453 papers analyzed, they can be classified based on this study’s objectives into three categories: studies on destabilization mechanisms, risk assessment, and dynamics analysis. The destabilization mechanism examines the circumstances that cause slopes to become unstable and the different factors that contribute to this destabilization [77]. This typically includes various factors like the soil’s physical and mechanical characteristics, the geological structures, the water conditions, and the weather patterns. Risk assessment combines the likelihood of slope instability with the potential outcomes of a failure, employing a risk level to gauge the severity of the risk. The risk degree obtained from assessment is compared with the safety level specified in the standard to determine the risk level of the slope [78,79]. However, dynamic analysis studies the response and stability of slopes under dynamic actions (e.g., earthquakes, wind loads, and wave impacts). Attention is given to dynamic processes such as deformation, stress distribution, and the energy transfer of slopes under dynamic action and how these processes affect the stability of slopes [80]. The study of the instability mechanism of slopes, the risk evaluation of slopes, and the dynamic analysis of slopes each have their own unique concerns and research methods in the field of slope engineering, which complement and support each other, and together provide a scientific basis and technical support for the design, construction, management, and maintenance of slope engineering [81].

(2)

Embankments and road riffles and artificial and natural slopes (different study objects)

The stability of embankment slopes is closely related to the slope gradient, and an excessively steep slope will increase the risk of landslides. Therefore, the reasonable design of a slope is an important measure to ensure slope stability. In addition, the design of a drainage system is also extremely critical. Rainfall is one of the main factors leading to the destabilization of embankment slopes, and the erosion and infiltration of rainwater on slopes can be reduced by constructing effective drainage systems, such as interceptor ditches and drainage pipes. Embankment slopes can be reinforced with supporting structures, such as anti-slip piles, soil nail walls, and anchor cables.

The stability of a road graben slope is greatly affected by the geological structure. Especially when there are faults, weak interlayers, and other geological defects, the stability of the slope will be significantly reduced. Therefore, in the design stage, it is necessary to carry out the detailed geological investigation of the slope, construct a risk assessment index system, use the hierarchical analysis method and other methods to classify the risk level of the slope, and formulate corresponding risk prevention and control countermeasures for high-risk road graben slopes. They should be installed with a wireless automatic slope monitoring and warning system for the real-time monitoring of the deformation of the slope and issue warning signals in time to protect the safety of transportation vehicles [82,83].

Artificial slopes are usually composed of fill, and their stability is greatly affected by the mechanical properties of the fill and the construction quality. In stability evaluation, the limit equilibrium method and the strength reduction method can be used to analyze and combined with numerical simulation software for comparison and verification. For artificial slopes with insufficient stability, reinforcement measures such as micro steel pipe pile top support technology can be used to prevent the slope destabilization caused by rainwater erosion [84]. The stability of natural slopes is significantly affected by groundwater and rainfall. Research shows that the transport law of groundwater is closely related to the stability of a slope, and the water content and resistivity of the soil body of the slope can be monitored using the technology of the high-density electric method, so as to analyze the distribution characteristics and migration law of groundwater. In addition, the geological structure of natural slopes is complex, with joints, fissures, and other structural surfaces, which are potential sliding surfaces for slope destabilization [85].

(3)

Freeze–thaw and rainfall and climate change (different study conditions)

There are many causes of slope instability, such as rainfall, earthquakes, blasting vibrations, water level changes, permafrost thawing, excavation, plant roots, and external loads [86], but in the literature studied in this paper, the three conditions with the highest frequency are freezing and thawing, rainfall, and climate change. Both freeze–thaw and rainfall affect the permeability, water content, and soil shear strength of slopes, which can lead to slope instability [87,88]. And climate change contains freeze–thaw cycles and wet–dry cycles. Since the climatic conditions of different regions are different, the research focuses are different; for example, in China, southeast China has more precipitation and does not freeze in winter, so the effects of rainfall and dry–wet cycles on slopes have been studied more, while northwest China has less precipitation and a long freezing period, so the effects of the freezing-and-thawing and freezing-and-thawing cycles on slopes have been studied more [89].

(4)

Numerical simulations and physical experiments and theoretical deduction (different research methods)

Numerical simulation is one of the most commonly used methods in slope stability studies and is able to cope with various types and complexities of slope stability problems. By adjusting the model parameters and the input data, numerical simulation can be adapted to different geological and soil conditions for analysis and prediction, which improves the versatility and flexibility of the method. The common numerical simulation methods include finite element analysis (FEA), Boundary Element Analysis (BEA), and Discrete Element Analysis (DEM). FEA is able to discretize the slope body into a finite number of small cells to account for the nonlinear properties of the soil body and the effects of complex boundary conditions. BEA represents the problem using boundary integral equations, which are suitable for solving unbounded or semi-unbounded slope problems. DEM is a specialized numerical method used to simulate the mechanical behavior of granular materials or discontinuous media, which is particularly suitable for complex granular systems in slope stability analysis. FEA can provide accurate stress distributions under complex geologic conditions, but its dependence on meshing may lead to computational inefficiencies, while DEM can simulate the interaction of particles, but the difficulty of parameter calibration restricts its engineering applicability [90,91]. However, the results of numerical simulations depend on accurate parameter inputs, the acquisition of which often needs to be verified by experiments or field monitoring. Therefore, numerical simulation results have a certain degree of uncertainty and need to be combined with other methods for comprehensive assessment. For example, in slope stability analysis under complex geological conditions, numerical simulation can provide detailed stress and deformation distributions, but the results need to be verified by physical experiments or theoretical derivations.

Physical experiments are one of the most important tools for slope stability research, especially for cases where the experimental conditions are limited, or field experiments cannot be conducted. Physical experiments can help researchers better understand the mechanical behavior of slopes by constructing models or conducting indoor tests to simulate the actual working conditions of slopes [92,93]. Theoretical derivation is the foundation of slope stability studies and helps to understand the nature of slope stability at the theoretical level. The classical theoretical analysis methods include limit equilibrium methods, such as Bishop’s simplified method, Spencer’s method, and the Morgenstern–Price method. These methods calculate the factor of the safety of slopes by simplifying the stress–strain relationship of the soil body, which is applicable to different sliding surface morphologies and geological conditions [94].

In summary, the research methods used by researchers in the field of the stability study of roadbed slopes tend to be diversified, but there are shortcomings; few studies can be combined with field experiments, and there is a single set of conditions, and there are insufficient studies conducted under coupled conditions, which may become a research trend in the future.

3.3.3. Three-Field Plot Analysis

In three-field analysis, these instructions were followed: select the authors on the left, the keywords in the middle, and the country of the authors on the right; adjust the number of fields displayed to 10; and then click run to generate a three-field graph. Figure 10 clearly shows the correlation between the three fields, and the authors of several papers are centered around the fields “slope stability”, “numerical simulation”, “kinematic analysis”, and so on, revealing the main research methods and analysis perspectives of the scholars. China is a vast country with long roads, and it has the most publications in this field. From Figure 10, it can be seen that China and the United States do not focus on a certain keyword or theme, and the proportion of keywords is relatively average, while India represents a prominent proportion of “kinematic analysis”. This is due to the fact that India is located in the South Asian subcontinent, with complex geological conditions, including mountains, plains, hills, and other terrains [95,96]. At the same time, India also faces variable climatic conditions, especially heavy rainfall during the rainy season, and such special geological and climatic conditions may prompt India to pay more attention to the kinetic analysis of roadbed slope instability in order to more accurately predict and respond to the risk of slope instability [97].

3.4. Analysis of Research Hotspots and Current Situation

Keyword emergence analysis is able to detect a sudden increase in keywords during a certain period of time, which often represent the hot topics or emerging trends within the current research field [98]. Using CiteSpace’s visualization, researchers can identify the keywords that gain focus over time, helping pinpoint the research trends. In this study, we used CiteSpace to analyze the keyword emergence in the literature, with N set to five at N%. The burst detection model uses α1/α0 to control sensitivity; decreasing α1/α0 increases sensitivity and detects the more prominent items. ‘αi/αi-1′ relates to the duration and intensity of bursts, with the higher values capturing the longer-term trends. The γ parameter determines the time window size for burst detection, influencing how long the algorithm searches for burst terms. By adjusting the parameters appropriately, the final burst graphs of 12 keywords were obtained, as shown in Figure 11. Strength indicates the strength of a keyword’s cited bursts, i.e., the extent to which a keyword’s citation frequency significantly increases within a certain period of time. Begin and end indicate the start and end years of the keyword bursts. The light blue areas on the right side, covering the years from 2014 to 2023, represent the period during which the keywords did not appear, while the blue zones mark the time period when the keyword started to appear, and the red lines specifically point out the time period when the keyword reached its peak or explosive growth.

The results of this study show that the study of roadbed slope stability is multifaceted, and the keywords emerge with little regularity. From 2014 to 2017, the keywords “slope stability”, “cut slope”, “slopes”, “plateau”, and other keywords appear, which may be related to focused research on the basic theory of slope stability and geological conditions at that time. In 2020–2021, “settlement” “cut slopes’, and “lesser himalaya" appear prominently, reflecting the focus on specific regions (e.g., the Himalayan region) and issues such as slope settlement during this period.

It is noteworthy that there is no significant keyword burst after 2021, but when analyzed in conjunction with Figure 3, 2021–2023 are the three years with the highest number of published articles. This phenomenon suggests that the research focus in this research area has become more decentralized despite the absence of significant hot keywords. This may be due to the fact that research on roadbed slope stability gradually covers more subdivided areas, such as slope stability analysis under different geological conditions, the effect of rainfall on slope stability, and slope stabilization techniques. This dispersed research focus reflects the diversification and depth of the research in this field and also shows that researchers are exploring the stability of roadbed slopes from multiple perspectives in order to cope with the complex and changing practical needs of engineering [99].

Figure 12 shows a graph of the hot topic trends. The chosen field is the title, and the N-gram chosen is bigram, which refers to two consecutive occurrences of a character or word. The horizontal coordinate is the year, the vertical coordinate is the keywords appearing in the title, and the size of the circle indicates the frequency of occurrence. This figure illustrates the prominent topics in recent years, showing a significant frequency of “slope stability” and “stability analysis”. This high occurrence is primarily because the literature we examined is sourced from the WOS core collection, which focuses on this keyword. From Figure 12 below, it can also be seen that the frequency of the use of railway roadbed is more than that of highway, which is due to the fact that the road condition of a railway is often more complicated, the load of a train is often larger, and it is more difficult for a train to avoid danger in time in the case of a landslide or collapse, so it has been studied more [100,101]. In recent years, numerical simulation techniques have gained popularity in slope stability research, driven by advancements in science and technology, as well as the inherent benefits and drawbacks of the methods themselves. These techniques offer advantages such as low cost, high flexibility, and excellent repeatability, making them suitable for a variety of complex geological scenarios and slope types. The ongoing advancements in computer technology, particularly the development of high-performance computing platforms, have significantly enhanced the efficiency and scale of numerical simulations. For instance, contemporary numerical simulation methods are evolving towards multi-scale and real-time dynamic simulations. These technological improvements allow for the more precise representation of the mechanical behavior of slopes under various conditions, serving as a valuable resource for analyzing slope stability [102]. However, numerical simulation methods also have certain limitations. Their results are highly dependent on the accuracy of the input parameters, the acquisition of which often needs to be verified by experiments or on-site monitoring. In addition, the establishment and validation of numerical simulation models require specialized knowledge and experience, which may otherwise lead to deviations between the models and the actual engineering problems.

4. Discussion

The research field of roadbed slope stability has received a lot of attention from scholars, but there are few review studies in this field. Therefore, this paper employs bibliometric analyses to select all the relevant literature within a specific time period to quantitatively explore the knowledge structure, the research trends, and new insights in this specific scientific field [103,104,105,106].

4.1. Systematic Review

The bibliometric approach utilized in this paper aids researchers in quickly and thoroughly grasping the current state and key trends in roadbed slope research, aiming to offer theoretical backing for future studies and investigations. The practical value of this bibliometric method includes the following: (1) Guiding engineering practices by identifying well-established reinforcement techniques and management strategies in the existing literature on roadbed slope instability. (2) Refining research directions by highlighting the current research hotspots and gaps. For instance, this analysis indicates that while there has been a surge in studies focusing on rainfall-induced instability, there is a notable lack of research on instability caused by other external factors such as earthquakes. This finding suggests that researchers should consider redirecting their efforts towards analyzing slope stability in the context of earthquakes to address this research gap and enhance theoretical support for designing and reinforcing roadbed slopes in areas prone to seismic activity. (3) Additionally, it can aid in policy development; for example, in cities undertaking extensive slope-cutting and excavation projects, the findings from this bibliometric analysis could lead to the implementation of a new management protocol that mandates stability assessments and detailed reinforcement plans for all slope-related projects.

The following is a comprehensive review and summary of the results of this study:

First, we analyzed the literature over the years 2014–2023 and understood the research categories of the retrieved literature. The graph of the percentage of major countries reveals that China represents a high percentage of 46.58% of the published articles, and it is obvious that in countries with more mountainous areas and more railroad and highway construction, there will be more research on this field. From 2019 onwards, there is a significant growth in research output, and the number of studies published in the last five years is more than twice as high as that of the previous five years. The book “Outline of the Construction of a Strong Transportation State” was published by the People’s Publishing House in 2019, an initiative that to a certain extent promotes and stimulates the development of the research aspect of roadbed slope stability; moreover, it can be seen from the keyword burst map in Section 3.4 that around 2019, numerical simulation methods, such as the finite element method (FEM), the discrete element method (DEM), and the finite difference method (FDM), are widely used in slope stability analysis, so 2019 is an important time point [107].

Then, we used the Bioloimetrix R package to derive the top five and ten studies with the highest total and local citations, respectively, and recorded their TCL and NCL values. We found that some of the literature had high standardized citation parameters, but not high citation counts, which was due to the year of publication of the literature and that the value of this type of literature should not be underestimated. When evaluating the authors of these studies, it was found that SINGH TN was the core of the team and contributed the highest number of articles to the field, up to 29. In addition, we also plotted a three-field analysis graph and a hot topic trend graph and a hot topic map. The hot topic map demonstrated the changes in the research hot topics in this field over the years, and it was found that the scholars used more and more methods with numerical simulation in this field, which was mainly due to the development of technology, the low cost of this method, and the increasing maturity of various simulation software. Next, we analyzed the covariograms of the literature with regard to the country, the institution, the keywords, and the authors and the year. In the context of its national background and policy environment, China attaches great importance to research and development in the field of roadbed slope instability, and it was found that China cooperates closely with the United States, the United Kingdom, and Canada. The two institutions contributing the most papers to this field are the Chinese Academy of Sciences and the University of Chinese Academy of Sciences; both from China, and these two institutions collaborate the most frequently, a phenomenon attributed to China’s rapid development in road construction and the Chinese government’s strong support for scientific research. This phenomenon is attributed to China’s rapid development in road construction and the Chinese government’s strong support for scientific research, especially in the fields of geotechnical engineering and geological engineering [108].

Finally, we conducted keyword emergence analysis using Citespace and discovered that the three years with the highest number of published articles did not show significant keyword emergence. This indicates a scattered focus and consistently used methodologies in research within this field. Additionally, we performed the analysis of the literature through co-cited journal and keyword clustering. Co-cited journal clustering analysis reveals the types of journals referenced in this area, while keyword clustering analysis organizes similar keywords into various topics or categories, making it easier for scholars to quickly identify the key themes and the potential connections in the literature. Through the cluster analysis of the co-cited journals and the keyword clusters, we categorized the research topics into the “mechanisms of roadbed slope instability,” the “risk assessment of roadbed slope instability,” and the “destructive characteristics of roadbed slopes.” Currently, the mechanisms behind roadbed slope instability are largely understood, and there has been notable progress in developing risk assessment models for roadbed slope instability. However, creating a widely applicable model to simulate the roadbed slope instability process remains challenging due to the complexity and variability of the issue [109,110].

4.2. Limitations and Challenges of Technology

Although significant progress has been made in the study of roadbed slope stability, there are still some limitations in practical application, including the numerical simulation techniques, the integration of the methodology and the actual working conditions, and the integration of economic and environmental dimensions.

4.2.1. Limitations of Numerical Simulation Techniques

Numerical simulation techniques, while advantageous in terms of cost and software, have several limitations. Their accuracy and reliability are often inadequate because the models and calculation methods depend on assumptions and approximations, which can deviate from reality and exacerbate errors in complex nonlinear problems. Determining the parameters can be challenging, as it requires extensive experimental data, and in practice, limited data and measurement errors can compromise accuracy. Additionally, high-precision large-scale simulations demand significant computational resources, necessitating powerful computers, ample storage, and efficient algorithms; otherwise, the computation times may become lengthy, or the problems may be unsolvable. The simplification and complication of boundary conditions restrict the ability to accurately represent real-world issues. Under intricate working conditions, the coupling of multiple physical fields, multi-scale interactions, and the complex behavior of materials complicate the establishment of precise and reasonable boundaries, further impacting result accuracy. Moreover, varying research methods and assumptions in complex scenarios can lead to significant discrepancies in the results, making it difficult to determine which is closest to reality. The capacity for multi-field coupling analysis is also limited, as practical engineering problems often involve the interaction of multiple physical fields, and numerical simulations face challenges in achieving efficient multi-field coupling. The differences in numerical methods and models across various physical fields add to the complexity and difficulty of coupling simulations.

4.2.2. Inadequacy of the Combination of Methodology and Actual Working Conditions

The summary of the literature also reveals that the current research methods are overly biased towards theoretical analysis, creating a disconnect between research and practical application. In the real world, the dynamic environment is complex and variable, with cyclical fluctuations in hydrological conditions, long-term changes in climatic factors, and frequent and diverse impacts of human activities, which are intertwined and have profound and unpredictable effects on slope stability. However, theoretical analysis cannot often comprehensively and dynamically capture these complex interactions, resulting in a large deviation between the theoretical prediction and actual slope stability in the dynamic environment, which greatly limits the direct applicability of theoretical prediction for the assessment of actual slope stability and makes it difficult to meet the demand for the accurate assessment of slope stability in engineering practice. Kempena Adolphe and colleagues found that the prediction accuracy of the traditional limit equilibrium method and the finite element method significantly decreases under complex geological conditions, especially in the presence of multistage sliding surfaces and non-homogeneous rock bodies. The simplifying assumptions of the model may lead to the overestimation or underestimation of the safety factor [111]. This study emphasizes the importance of combining field monitoring data with geologic models to improve the applicability of theoretical models, and further illustrates the drawbacks of studies that are overly biased toward theoretical analysis.

4.2.3. Integration Aspects of Economic and Environmental Dimensions

In slope engineering, it is important not only to ensure technical feasibility, but also to consider the economic costs and the environmental effects. However, current research often places too much emphasis on technical stability analysis, neglecting the economic and environmental aspects. There is a lack of a systematic engineering economic evaluation framework. This research imbalance hinders the thorough and precise assessment of the true value and sustainability of slope projects, making it challenging to properly balance the technological, economic, and environmental factors in project decision making. Consequently, this impacts the overall benefits and sustainable development of slope projects. For example, Yao and his colleagues reviewed the recent progress and application of Life Cycle Assessment (LCA) in slope engineering, which is an environmental management tool that can comprehensively assess the environmental impacts of a slope project throughout its life cycle, from design, construction, and operation to demolition [112]. This research provides a new idea for the sustainable development of slope engineering and a reference value for the field of roadbed slopes.

4.3. Future Research Directions

One of our goals in upcoming research is to take advantage of the advances in numerical simulation technology by using numerical simulation software to transform complex natural conditions into multi-physics field problems. We will prioritize the advancement of highly accurate numerical models for thermal–hydrological–mechanical–chemical (THMC) multi-field coupling. Taking the embankment slope in the roadbed slope as an example, Figure 13 shows the framework of the established finite element model of the embankment slope under multi-field coupling conditions, in which Figure 13a demonstrates the setting of the model boundary conditions, and Figure 13b shows the simple flow of numerical simulation. Using ABAQUS finite element analysis software, we can effectively utilize its related subroutine functions to efficiently create a multi-field coupled model. First, we will create the original ground stress field and the seepage field to simulate the water migration characteristics. Next, we will create static and dynamic loading stress fields by writing DLOAD and VDLOAD subroutines, respectively. In addition, we will create a creep field using the CREEP subroutine since slopes typically creep under long-term loading. These fields will be integrated into a single model, resulting in a complex model with multiple coupled fields. While the model may not be universally applicable to all roadbed slope conditions, it will more accurately reflect the real-world conditions and significantly reduce the errors in slope stability analysis due to simplifying assumptions, thus providing a more reliable basis for slope design and safety assessment.

The establishment and use of the above multi-field coupled numerical model improves the accuracy of numerical simulation, alleviates the problem of insufficient simulation accuracy in current research, and enhances the accuracy of slope stability analysis in order to provide a solid guarantee for the sustainable development of infrastructure. In addition, future research should also focus on the dynamization and intelligence of roadbed slope stability prediction, integrating artificial intelligence and big data technology, combining multiple research methods to promote interdisciplinary exchange, and constructing a multi-criteria decision-making (MCDM) framework that balances the technological, economic, and environmental factors, as well as providing new paths for the sustainable development of roadbed slopes with the help of tools such as LCA, which will help infrastructure construction professionals to meet sustainable development goals.

5. Conclusions

This study offers the bibliometric analysis of 453 research articles concerning roadbed slope stability sourced from the WOS core database. The aim is to enhance our understanding of the research domain and deliver a thorough summary of the key literature on the topic. The detailed examination of the current research landscape is conducted, focusing on the authors, the journals, the countries, and the keywords within the literature. The specific findings are as follows:

• This paper uses knowledge mapping to reveal the hotspots and trends in global roadbed slope stability research. The key topics include stability, numerical simulation, rainfall, and freeze–thaw cycles. China dominates the field, with 46.58% of publications and collaborates closely with the US and the UK. The study numbers surged after 2019, driven by China’s transportation development initiatives and advances in numerical simulation. Some influential works include Intrieri’s landslide prediction research and Zhang’s work on permafrost slope stability. The leading author is SINGH TN, who has published 29 papers on slope destabilization mechanisms and risk assessment. These results form a comprehensive knowledge framework for the field.
• Although the existing numerical simulation techniques have cost advantages and advanced software support, they face accuracy and reliability challenges due to the lack of sufficient practical validation, and the differences between the different research methods and the assumptions are more significant under complex working conditions. Future research should focus on multi-field coupled conditions, strictly calibrate the parameters through field measurements and experimental validation, and emphasize the development of highly accurate numerical models for THMC multi-field coupling.
• The current research methods usually favor theoretical analysis and ignore the interactions in dynamic environments, such as hydrology, the climate, and human activities. This disconnect limits the direct applicability of theoretical predictions to practical slope stability assessment. Future research should focus on the dynamization and intelligence of roadbed slope stability prediction, which needs to further integrate artificial intelligence and big data technology and requires a combination of multiple research methods to achieve interdisciplinary collaboration.
• The current research on roadbed slope stability is still mainly focused on technical feasibility, and the systematic assessment of economic cost and environmental impact is very limited. Future research should emphasize a multi-criteria decision-making framework that balances the technical, economic, and environmental factors to ensure that infrastructure projects are in line with the goals of sustainable development, and frameworks such as Deep Innovation MCDM and LCA provide a pathway to address the sustainability of roadbed slopes.

While this research was limited to English language literature, potentially causing incomplete analyses due to missed papers, it still underlines the utility and value of bibliometric methods for uncovering the global research trends in roadbed slope stability. This work helps fill the gap in comprehensive studies on this topic. Looking ahead, other researchers should prioritize exploring the instability mechanisms under diverse coupled conditions, conduct more field research, and create innovative techniques and methods for studying roadbed slopes.