Chinese international process safety research: Collaborations, research trends, and intellectual basis

This article presents a bibliometric analysis and mapping of the Chinese process safety research, focusing on the contributions made in core process safety journals and on the influences of international collaborations and knowledge sources on the developments of this research domain. Collaboration networks, term co-occurrence networks, and co-citation network were analyzed to identify trends, patterns, and the knowledge distribution of the Chinese research on process safety. Work to data has been clustered mainly on safety of chemical processes, fire and explosion, and risk management and accidents. Chinese research contributions are concentrated in only few journals, while the corresponding intellectual base draws on the wider literature focused on understanding and modeling phenomena, and on the broader risk research literature, although to a lesser extent. While various foreign authors are highly cited by Chinese authors, only very few direct collaborations with international scholars are identified. The results are used as a basis for a discussion on future research directions and developments for the community. Increased focus on uncertainty treatment and handling of black swan events, risk evaluation and economic aspects of safety decisions, interorganizational risk management, road and maritime transport of hazardous substances, risk perception and communication, and integrated safety and security assessment, are highlighted as fruitful directions for future scholarship. It is hoped that the insights obtained from this work can facilitate new and consolidated collaborations, as well as further invigorate the Chinese process safety domain, ultimately contributing to improved safety performance of process industries in China and elsewhere.


Introduction
Process industries are an important sector in the global economic landscape and provide the backbone for economic and social development in many areas where industrial facilities are located. Nevertheless, process plants are hazardous facilities, which can lead to significant negative safety and environmental impacts and are thus of high societal concern. In support of its economic development, the scale, production output, and revenue of process industries has grown very fast in China (Yang et al., 2019a). Simultaneously, due to the above-mentioned safety and environmental concerns, process safety has gained increasing attention in Chinese industrial, government, and academic contexts. This has led to a steady improvement in safety performance of the Chinese chemical industries . Nevertheless, accidents and serious high-consequence loss events still occur relatively frequently, e.g. the explosions in Dalian Port (2010-7-16), Qingdao (2013-11-22), Tianjin Port (2015-8-12), and Xiang-Shui (2019-3-21). Various authors have performed research focusing on these and other accidents (Chen and Wang, 2012;Chen et al., 2019;Zhang et al., 2019;Zhao, 2014), indicating that scientific research is important for improving process safety in China.
Several authors have highlighted aspects of process safety in China.
Some authors focus on describing the characteristics of these accidents. For instance (Duan et al., 2011), made an analysis of hazardous chemical accidents in China between in the period 2000-2006, finding a fatality rate between 220 and 1100 people per year (Zhang and Zheng, 2012). presented a statistical of characteristics of chemical process accidents in China, providing insights in the associated time volatility, location, accident type, injuries and deaths, assigned causes, and industry type (Zhao et al., 2018). present an updated analysis of the chemical accidents in China in the period 2006 to 2017, providing insights in the time evolution, accident types, and assigned causes of these accidents. Other authors provide insights in aspects of process safety in China, with for instance several authors describing challenges for process safety for SMEs in China (Zhao et al., 2013), lessons learned for process safety management in China , or describe core elements and control measures for process safety management in China (Zhou et al., 2017). (Besserman and Mentzer, 2017) reviewed global process regulations, comparing regulation in China with that of the European Union, India, the United Kingdom, and the United States (Wang et al., 2018a). provided an overview of work safety in China's thirteenth five-year plan for the period 2016-2020, highlighting various challenges and tasks related to chemical process safety (Wang et al., 2018b). described problems, challenges and opportunities for process safety in China, while  provides a recent overview of the status of the chemical industry in China, safety-related problems, and pathways to sustainable development. Reviews and retrospective analyses are important catalysts to focus attention on the developments in a research domain. Several descriptive accounts of the history of process safety or one of its subdomains have been presented. For instance , provide an overview of methods and models in process safety and risk management (Swuste et al., 2016). review the literature on process safety indicators (Mkpat et al., 2018), on process safety education, and (Gao et al., 2020) and (Park et al., 2020) address inherent safety in chemical processes and during the conceptual design process, respectively.
Given the very large body of research on process safety which has accumulated over the years, narrative reviews are of limited use to provide insights in the overall developments of the research domain. Narrative reviews are very useful and important for gaining understanding of specific aspects of a research topic, but become cumbersome and unwieldy for obtaining insights in larger domains of research (Grant and Booth, 2009). For such purposes, bibliometric analyses are more suited as a review methodology, as these techniques can provide quantitative insights in the contents, structure, patterns, and developments of a research domain (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. Various such scientometric analyses have been presented for process safety research . made a bibliometric analysis focusing on domino effects in the process industry, describing temporal evolutions, collaborations between countries, organizations and authors, and co-citation clusters providing insights in important clusters of research contributions (Gobbo et al., 2018). performed a bibliometric mapping analysis linking process safety to environmental protection and industry 4.0 (Amin et al., 2019). present a scientometric analysis of all process safety related research, focusing inter alia on active countries, key areas, leading authors, publication sources, and focus topics (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. present a scientometric overview of the research contributions by Dr. Sam Mannan, one of the process safety pioneers who was active and highly impactful in several of its research subdomains. A comparative analysis of the publication trends, focus topic areas, collaboration networks and core co-citation clusters of three core process safety related journals (Journal of Loss Prevention in the Process Industries (JLPPI), Process Safety and Environmental Protection (PSEP), and Process Safety Progress (PSP)), is presented by Li et al. (2020a). Yang et al. (2020) present a bibliometric analysis of process safety in China from a sustainability perspective, highlighting provinces of high research activity, collaboration networks, hot topics, influential contributions, and international collaborations.
In the current article, compared to Yang et al. (2020), a complementary and more detailed analysis is made of the Chinese process safety research, focusing especially on the international collaboration of Chinese scholars, and on the influence of research originating from outside China on the Chinese process safety research. In this analysis, the focus is furthermore on research of Chinese origin published in the three core process safety journals analyzed in Li et al. (2020a): JLPPI, PSEP, and PSP. To provide further insights in the developments of process safety research in China, answers to following research questions are sought: RQ1. In what international geographic collaboration networks are Chinese process safety researchers and institutions active?
RQ2. What are the dominant narrative clusters and focus topics in Chinese process safety research, and what trends can be identified in these clusters?
RQ3. For what topics does the international collaboration drive the progress in Chinese process safety research?
RQ4. What research from which international (non-Chinese) authors has had a significant impact on the development of Chinese scholarship on process safety?
The aim of the current research is thus to provide further insights in the status and development of Chinese process safety research. A subsequent discussion will further contextualize the outputs of the Chinese process safety research community, focusing on selected future research directions considering recent trends in the wider process safety research community.
The remainder of this article is organized as follows. In Section 2, the data collection process is described and the applied scientometric methods briefly outlined. Section 3 presents the analysis results, providing answers to the above stated questions. In Section 4, a discussion on the results is given, while Section 5 concludes.

Data source and construction of datasets for analysis
The article data sample of Chinese international process safety (CIPS) research were searched in Web of Science (WoS) and downloaded on February 6, 2020. The Web of Science Core Collections (WoSCC) (SCI/ SSCI) from Clarivate Analytics was selected as the source database, as this is known to comprehensively cover most academic disciplines with a high data quality (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. The time span of the search strategy was set from 1900 to 2019. To obtain the Chinese international process safety papers, the steps shown in Fig. 1 were taken in the search process. First, articles originating from China in the core process safety journals (JLPPI, PSEP, and PSP) are downloaded as dataset #1. These journals are selected because earlier work has indicated that these are the 'core' process safety journals, see Li et al. (2020a). By construction, all papers in this dataset will have at least one Chinese researcher as author. This dataset contains 1, 148 records. Second, a title-based search strategy is used, with the search term "process safety" and country of origin of the document set as China. This search leads to dataset #2, containing 97 publications. Third, these two datasets (dataset #1 and dataset #2) are combined to dataset #3 using the "OR"-logic operator. This dataset contains 1,214 records, indicating there is an overlap of 31 documents between the two datasets. In Table 1, some characteristics of this dataset are shown. For further use in the remainder of this article, this dataset is labeled 'Dataset TP' (for 'Total Publications') and contains all the articles focusing on process safety originating from China.
To distinguish the domestic Chinese articles (i.e. articles originating from China without international collaboration) from internationally co-authored articles originating from China, an additional operation is performed using the WoS search platform. As shown in Table 1, in 'Dataset DP' (for 'Domestic Publications'), 931 records are retained by excluding articles which are recorded in the WoSCC database as originating from the countries or regions indicated in Fig. 1. Conversely, the 'Dataset ICP' (for 'Internationally Co-authored Publications') contains 283 documents, and is obtained by including the countries with which Chinese authors have collaborated in the search strategy (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. 1 Table 1 lists selected characteristics of this dataset. Fig. 2 and Table 1 show several bibliometric indicators and summary statistics of the finally obtained datasets DP, ICP, and TP. Fig. 2(a) shows that 283 (23.31%) of the papers have at least one co-author from other countries/regions than China, while 931 (76.69%) are domestically Chinese articles. With an overall international collaboration rate of just 0.23, there appears significant room for increased internationalization of Chinese process safety research. Apart from the benefit of international collaboration to address complex challenges (Sonnenwald, 2007), this can increase productivity (Beaver, 2001) and may lead to increased acceptance of research contributions, especially when co-authorships are established with high-visibility scholars (Chinchilla-Rodríguez et al., 2018). Fig. 2(b) shows that the main sources of Chinese international process safety research were published in Journal of Loss Prevention in The Process Industries (571, 47.0%), Process Safety and Environmental Protection (489,40.3%) and Process Safety Progress (88, 7.3%), with these top three journals accounting for 94.6% of the total papers in the TP dataset. This confirms that indeed, these three journals can be regarded as 'core' process safety journals, which supports the data selection strategy outlined above. The results of Li et al. (2020a) indicate that JLPPI, PSEP, and PSP have in the period 1999-2019 together published 5295 articles, where these respectively account for 45.4%, 35.7%, and 18.9% of this total. This indicates that PSP is less frequently chosen as a publication outlet by Chinese scholars than JLPPI or PSEP, when compared with the overall publication trends in these journals.
The annual trend of each dataset is shown in Fig. 2(c). It is seen that Chinese international process safety research started in 1995, with the first internationally co-authored article appearing soon after in 1996.  Table 1.   Table 1. 1 The countries listed in the search restrictions for set #4 and #5 in Fig. 1 are determined by a pre-analysis of the collaborating countries represented in set #3. As explained in Li et al. (2020aLi et al. ( , 2020bLi et al. ( , 2020cLi et al. ( , 2021, an initial analysis of the obtained dataset(s) is commonly made to refine the finally used dataset(s) in the envisaged scientometric analysis.
Compared to the overall process safety research domain, which research efforts initiated in the late 1960s (Amin et al., 2019) and intensified in the 1970s and 1980s through the efforts of influential process safety pioneers such as Lawley (1974), Gibson (1976), and Kletz (2012), Chinese scholarly contributions appeared relatively late. From about the mid-2000s and especially since 2010, a very fast increasing publication output is however clearly apparent. It is seen that domestic publications are characterized by an exponential trend, while the trend of internationally co-authored publications shows a polynomial trend. In the past five years (2015-2019), the growth of domestic and internationally authored co-publications can both be well approximated with linear trends, where the rate of increase of the latter is higher than that of the former. Thus, while the growth speed of domestic Chinese process safety research is higher than that of internationally co-authored Chinese process safety research, in recent years there appears to be an increased focus on internationalization of the research efforts.

Bibliometric analysis and mapping methods and tools
Various bibliometric mapping analysis methods are applied in the current research to obtain answers for the research questions stated in Section 1. Bibliometrics is a scientific area and is defined as the application of the mathematics and statistics to analyze the journals, books and other media in a quantitative way (Pritchard, 1969). Bibliometric mapping refers to the combination of visualization techniques with the statistical methods of the bibliometric analysis, which are thus used to visually represent the results to facilitate interpretation (Small, 1999).
In recent years, bibliometrics research has seen a fast growth as a subdomain of information and science studies, and its methods and tools have been increasingly applied in diverse domains of science. Currently, a wide variety of tools has been developed, which has led to increased interest in performing scientometric analyses across scientific domains (Moral-Muñoz et al., 2020). Within the safety science community, bibliometric mapping has to date primarily applied to identify networks of author, organization, or country collaboration, to identify patterns in journals as knowledge sources, and to construct topic maps of frequently occurring terms (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. Examples of application of scientometric methods to obtain high-level insights in the development of safety science include the identification and knowledge flows between core safety related journals Hale, 2015, 2016), and the evolution of the Safety Science journal (Merigó et al., 2019) and of the three core process safety journals JLPPI, PSEP, and PSP (Li et al., 2020a). Furthermore, several aspects, topics or specific subdomains of safety research have been analyzed using bibliometric mapping, for instance safety culture (van Nunen et al., 2018), human reliability analysis (Tao et al., 2020), laboratory safety in universities (Yang et al., 2019b), construction safety (Akram et al., 2019;Jin et al., 2019), and maritime safety (Luo and Shin, 2019;Gil et al., 2020).
There are several methods for bibliometric mapping analysis, each designed for a specific research purpose. In this paper, three types of bibliometric maps are created to answer the research questions listed in Section 1: collaboration network, terms network and co-citation network. Further details of the applied scientometric mapping methods and tools are given in Table 2.
To answer research question RQ1, co-authorship mapping is used to detect the collaborations within Chinese international process safety research. Collaborations are identified and counted first at the document level, which is then further aggregated to the level of authors, institutions, cities, or countries/regions (Leydesdorff and Wagner, 2008). Co-authorship mapping is useful to represent the relative significance of contributors to the scientific domain and displays the social networks and the structure of collaboration in the Chinese international process safety research community. For this analysis, dataset TP of Table 1 is applied, with results shown in Section 3.1.
For answering research questions RQ2 and RQ3, terms co-occurrence maps are created. Terms and term phrases occurring in specific research domains provide insights in the narratives of those domains, where frequently co-occurring terms can be used to obtain insights in the important narrative patterns in the research area (Van Eck et al., 2010b). Frequently occurring terms are furthermore considered to indicate hot topics in the research domain. In this paper, noun phrases of the Chinese international process safety research are identified, extracted, selected, and visualized. Referring to Table 1, dataset TP is used for answering RQ2, while datasets DP and ICP are used to answer RQ3. Results are presented in Section 3.2.
For research RQ4, co-citation maps are constructed, focusing on journals co-citation, authors co-citation, and references co-citation of Chinese international process safety research. Co-citation analysis of references was first introduced by Small (1973), where the basic idea is that if two references are cited together (i.e. are co-cited), there is a certain relation between them. When references are frequently cited together, this indicates a high degree of similarity between them in some way, for instance as they address the same topic or build on the same ideas. Building on Small's original ideas, authors co-citation analysis (White and Griffith, 1981) and journals co-citation analysis (Ding et al., 2000;McCain, 1991) were also developed, using the same notion of similarity of the co-cited information units. For this research question, dataset TP of Table 1 is applied, with results shown in Section 3.3.
In the current research, the bibliometric maps of Chinese international process safety research are constructed using the VOSviewer software (van Eck and . VOSviewer is well known in bibliometric mapping analysis and has been widely used in scientific communities, also in safety related topics (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. The VOSviewer tool implements the visualization of similarities method, which determines the strength of relation between objects using mathematical formulations (van Eck and Waltman, 2007).

Note:
Colors in the bibliometric maps normally indicate the cluster to which an item is assigned.
Color schemes can also be used as an overlay to highlight e.g. the temporal evolution or scientific impact associated with specific items in the map.
There are three main techniques implemented in VOSviewer to construct the bibliometric map: co-occurrence matrix normalization van Eck et al., 2010a), VOS mapping technique , and VOS clustering technique (Waltman and van Eck, 2013;Waltman et al., 2010). For further technical details, the reader is referred to these publications.

Results
In this Section, the results of the various bibliometric analyses are shown, where the bibliometric maps created as described in Section 2.2 for the datasets described in Section 2.1 are used as central aspects of providing the answers to the research questions of Section 1. In Section 3.1, collaboration networks of the Chinese international process safety are shown, answering RQ1. Section 3.2 shows the results of the terms cooccurrence network analyses, answering RQ2 and RQ3. Finally, in Section 3.3 the results of the co-citation analyses are shown, answering RQ4. Fig. 3 shows the collaboration network of Chinese researchers detected in the Chinese international process safety research, using dataset TP of Table 1. Of the 1,214 records in this dataset, 93.29% were geocoded successfully, i.e. 1,133 records. The remaining 6.71% of the records (81 items) could not be geocoded and are hence not represented in the results.

Institutions and authors collaboration network
There are 699 institutions identified from the addresses provided in the publications. For the analysis, institutions with 10 or more publications are extracted from the cleaned dataset TP of Table 1, with in total 55 institutions meeting this threshold. Finally, the collaboration network of these 55 institutions in Chinese international process safety is constructed and shown in Fig. 4. As explained in Table 2, the node and label sizes indicate the number of publications of an institution, where the width of the links between each node pair represents the strength of research collaboration between the two institutions. The colors represent clusters of institutions with a comparatively high degree of joint research activities.
Given the adopted search strategy of Section 2.1, it is self-evident that almost all institutions are from China. China University of Mining and Technology from Jiangsu province has published 109 articles of the total of 1,214 (8.98%). It is followed by China University of Petroleum (86, 7.33%), Nanjing Tech University (83, 6.84%), Beijing Institute of Technology (63, 5.19%), University of Science and Technology of China (53, 4.37%), and Tsinghua University (51, 4.20%). Among these highly productive institutions, the average publication year (APY) is used to determine which institutions have become active in recent years. It is found that Shandong University of Science and Technology (APY

Narrative clusters and focus topics in Chinese international process safety research, and topics driven by international collaboration
Terms appearing in the title and abstract of scientific articles are important descriptors of their key content. In this Section, terms appearing in Chinese international process safety research are extracted from the titles and abstracts of the scientific publications of the TP dataset of Table 1. This is done using the automatic term identification method (Van Eck et al., 2010b), with the terms map was visualized by VOSviewer (van Eck and Waltman, 2010). Fig. 6 shows the terms co-occurrence map. Adopting a threshold so that only terms occurring at least in 10 different articles are shown, a total of 314 terms are identified and visualized. A simple descriptive statistical analysis indicates that the term occurrence frequency distribution is very unbalanced, with only a few terms occurring very frequently and many terms only comparatively seldom. For example, there are only 13 terms which occur at least 100 times in the considered dataset. These terms are 'concentration' (232 occurrences), 'accident ' (198), 'experimental study ' (191), 'pressure' (187), 'increase' (181), 'approach' (177), 'risk' (161), 'safety' (128), 'mixture' (120), 'China' (103), 'case study' (102), 'property' (102), and 'reaction' (101). While these terms are indeed clearly associated with the general topic of process safety research in China, they are very generic and provide little insight in the specific narratives and focus topics in the considered research domain.
To provide further insights in the narrative patterns and the focus topics in the research domain, a terms cluster analysis is conducted, which analyzes the terms in pairs and groups. As outlined in Section 2.2, two terms are considered to co-occur these appear together in the same paper. Co-occurrence relations give insights in the semantic links in the dataset, more easily allowing interpretation of the dominant narrative patterns in the research domain. A further clustering of the terms based on their co-occurrence strength provides insights in the main research areas. Using the VOS clustering technique (Waltman and van Eck, 2013;Waltman et al., 2010), three clusters are detected and shown in Fig. 6: Cluster #1 'Process safety risk management' (including 120 terms), Cluster #2 'Fire and explosion process safety' (105 terms) and Cluster #3 'Chemical process safety' (89 terms). It is apparent that these are the same main research areas as those associated with all research published in the core process safety journals JLPPI, PSEP, and PSP, see (Li et al., 2020a). The labels of these clusters are judgments by the authors, based on the frequently occurring terms in each cluster.
To obtain further insights in the evolution and influence of the research topics in Chinese international process safety research, visual overlays are applied to the term co-occurrence map of Fig. 6. The results are shown in Fig. 7, with (a) depicting the average publication year in which the terms appear, and (b) the average citations of the terms. The overall average publication years of each cluster are very close to one another: Cluster #2 is 2015.876, Cluster #1 is 2015.997, and Cluster #3 is 2016.109. However, the standard deviation of the publication years of the terms are rather different, with Cluster #3 having the highest deviation (0.972). This indicates that while chemical process safety is wellestablished as a research area within Chinese international process safety research, it also has some significant emerging research focus topics. In Cluster #1, the terms 'risk ', 'assessment', 'risk analysis', 'Bayesian network', 'uncertainty', and 'disaster' have the highest average publication year, indicating that these topics are more recent. Within Cluster #2, the recently appearing topics are 'dust', 'explosion pressure', and 'explosion parameter'. Finally, in Cluster #3 the terms 'coal', 'spontaneous combustion', and 'FTIR' (Fourier Transform Infrared) are more recent.
In the co-occurrence map with overlay of the average number of citations, shown in Fig. 7(b), it is seen that the terms with the highest impact are primarily located in Cluster #3. Within this cluster, the term 'removal' has the highest average number of citations. Other impactful terms include 'pollutant', 'aqueous solution', 'reaction temperature', 'coal oxidation', 'adsorption process', 'response surface methodology', and 'chemical oxygen demand'. In Cluster #2, terms related to 'dust', including e.g. also 'dust explosion', 'dust cloud', and 'particle' are comparatively more impactful than other terms in the cluster. Finally, in Cluster #1, terms related to quantitative risk analysis, such as 'fault tree', 'failure probability', as well as 'gas pipeline' and 'construction' are highly impactful. Fig. 8 shows another overlay analysis of the term co-occurrence map of Fig. 6, indicating the ratio of the terms appearing in respectively the domestic and internationally co-authored Chinese process safety research. It is seen that the domestic papers primarily appear in Cluster #2 'Fire and explosion process safety', whereas internationally coauthored topics are mostly associated with Cluster #1 'Process safety risk management'. As it is well understood that only scientific, technical, or engineering knowledge is insufficient to effectively manage risks and reduce the number of occurrences of process safety accidents (Meyer and Reniers, 2016), the comparative lack of specific knowledge of safety and risk management can be seen as a shortcoming of the Chinese process safety research community. Considering that international collaboration can improve the capacity to address complex challenges and facilitate knowledge exchange (Sonnenwald, 2007), it is apparent that various Chines researchers have established collaboration networks as indicated in Section 3.1.2, in particular to enhance China's capacity to work on safety and risk management related topics.

Cited journals and authors analysis
Highly cited journals are journals from which articles are frequently cited by Chinese process safety researchers. Hence, these journals can be regarded as influential knowledge sources in the Chinese process safety research community. Fig. 9 shows the journals co-citation network of Chinese international process safety articles. In this network, Journal of Loss Prevention in the Process Industries (3124), Journal of Hazardous Materials (1694), Process Safety and Environmental Protection (1102), Fuel (836), Combustion and Flame (737), Safety Science (582) and Reliability Engineering and System Safety (513) are highly cited journals, each having received more than 500 citations from the Chinese international papers in process safety. The structure of the intellectual basis at the journal level is clustered in five groups: Cluster #1 'Environmental technology' (red in Fig. 9), Cluster #2 'Process safety and risk' (green), Cluster #3 'Fuel and mining' (blue), Cluster #4 'Chemical science and engineering' (purple), and Cluster #5 'Energy' (yellow). The dominant cluster clearly is Cluster #2 'Process safety and risk', showing the central importance of knowledge about safety, reliability, and risk aspects in implementing process safety. It is noteworthy that although JLPPI, PSEP, and PSP are core knowledge carriers for process safety research (Amin et al., 2019;Li et al., 2020a), PSP contributes comparatively less to the knowledge base for Chinese process safety researchers than the other core process safety journals, and even less than more generic safety and risk related journals.
Cluster #2 also contains a subset of journals related to fire and combustion, which are related as well to the fuel-related knowledge carriers in Cluster #3. In terms of application domains of the Chinese process safety research, it is apparent that 'Environmental technology' and 'Fuel and mining' related knowledge is comparatively more relevant for the Chinese process safety research community, which confirms the findings of Section 3.2 that related topics provide important narratives in the research domain. Knowledge carriers related to 'Chemical science and engineering' and 'Energy' are also important, although comparatively Fig. 6. Clustered term co-occurrence map of Chinese international process safety research and selected summary statistics, visualized using VOSviewer . less than the above-mentioned application domains.
Highly cited authors by the Chinese process safety research community show that articles by these authors are influential to the development of the Chinese process safety research. In this paper, particular interest goes to the scholars from other countries/regions, who have been highly cited by Chinese researchers. A co-citation map of the authors of the cited journals, presented in Fig. 10, shows that Khan FI (129) These authors all have received more than 50 citations from the Chinese process safety research community.
As shown in Fig. 5, among these authors, Khan FI, Amyotte PR, Mebarki A, and Reniers G already have established collaborations with Chinese colleagues, which may be among the reasons that their work has been more frequently cited than that of other international process safety scholars. Notwithstanding the value of the knowledge dissemination through collaboration with these above-mentioned international scholars, it may be fruitful for the development of Chinese process safety research, and ultimately for enhancing the safety performance of the Chinese process industries, to further diversify its international collaborations.
In Fig. 10, the cited authors are furthermore clustered based on their co-citation strength. The clusters show that the intellectual basis of from other countries/regions on which the Chinese process safety research community relies is mainly focused on four aspects. In Cluster #1 'Explosion and pressure', the key authors are Di Sarli V, Razus D, and Lee JHS. Cozzani V, Mebarki A, and Baum MR are distinguished authors in  Cluster #2 'Domino effects', with Cozzani V having strong co-citation links with Khan FI and Reniers G, who are primarily active in Cluster #4 'Process safety and risk analysis', but also are important contributors to domino effects . In Cluster #3 'Explosions', the key authors are Amyotte PR, Eckhoff RK, and Cashdollar KL. Finally, in the Cluster #4, apart from the already mentioned Khan FI and Reniers G, Khakzad N is a key scholar.

Cited documents analysis
Highly cited references signify publications which have been frequently used as a basis for new research developments. Hence, such articles can be regarded as the immediate intellectual basis of the Chinese international process safety research domain. In total, 80 references with more than 10 citations are identified and extracted to construct the reference co-citation network. Fig. 11 shows the distribution of these publications, indicating that these originate from 22 different sources, in the period 1965-2018. Among these sources, Journal of Loss Prevention in the Process Industries accounts for 30 out of 80 publications, showing its key role in advancing the knowledge basis of Chinese process safety research. It is followed by Journal of Hazardous Materials (14 of 80 highly cited references), whereas Process Safety and Environmental Protection and other journals are significantly less influential. It is noteworthy that Process Safety Progress, although a key process safety journal, is not represented in the list. In terms of total  The clusters of highly cited references in Chinese international process safety research are analyzed through a documents co-citations network, which provides insights in the intellectual structure of Chinese process safety research. The resulting clustered co-citation map is shown in Fig. 12. It contains six clusters, with the top 3 most frequently cited publications listed in Table 3. The labels of these clusters are judgments by the authors, based on the titles and abstracts of the top-3 cited articles in each cluster. The research in cluster #1 is mostly related to gas explosions; cluster #2 focuses on the application of Bayesian network models for process safety and risk analysis; cluster #3 concerns domino effects; cluster #4 relates to dust explosion; cluster #5 focuses on spontaneous heating of coal; while cluster #6 focuses on pressure and explosions. While clusters #2, #3, and #5 appear more peripheral in the figure and are rather disjoint from other clusters in the co-citation network, the clusters #1, #4, and #6 are located more closely to each other, and show stronger linkages between them. This is understandable as they all relate to various aspects of explosion, which is one of the key research areas within Chinese international process safety research, as found also e.g. in Fig. 6.   Fig. 11. Year and source of the highly cited references in Chinese international process safety research.

Interpretation and future research directions
The analyses of the Chinese international process safety research show that the overall research community has made a fast-paced increase in research productivity especially during the last 15 years. Whereas compared to other industrialized regions such as western Europe and eastern United States, the developments in Chinese process safety research came with a delay, there currently is a large and active community advancing this knowledge domain in support of safer and more environmentally sustainable process industries.
Despite the significant number of countries and institutions with which Chinese scholars have contributed as seen in Fig. 3, there are relatively few internationally leading scholars who have established extensive collaborations with Chinese authors, as can be seen from Figs. 4 and 5. Wagner and Leydesdorff (2005) found that international collaborations are preferably sought with high-impact international scholars. Evidence suggests that international collaboration and researcher mobility is associated with increased productivity (Beaver, 2001), higher quality (Abramo et al., 2011), higher citation rates (Gazni et al., 2012), and increased acceptance of the research outputs (Chinchilla-Rodríguez et al., 2018). Further developing such collaborations, e. g. through international mobility efforts, can help to further advance the Chinese process safety community. Hence, both international leading experts in process safety as well as Chinese scholars can mutually benefit from establishing further collaborations. A further diversification of the internationalization of Chinese process safety research may moreover be beneficial to widen the intellectual basis of the research contributions. As seen in the co-citation analyses of Figs. 10 and 12, many of the most frequently cited authors by the Chinese process research community are also those with which collaborations have been established, see also Section 3.1.2. As there likely is a causal connection between these observations in that authors may preferentially adopt knowledge from their collaborators, this suggests that a further differentiation of the international collaboration may benefit the development of process safety research in China.

Table 3
Top 3 high cited references in each cluster of the reference's co-citation network.   Ferrara et al. (2006) Note: NC = number of citations | ACP = average number of citations per year.
From Fig. 11, it is noteworthy that many of the highly cited articles in the Chinese process safety research literature are published relatively recently, with especially the period after 2005 representing a significant impact. This may be related to finding that the Chinese process safety research community itself became more active from around approximately that time, see Fig. 2. Considering that even very impactful articles receive fewer citations over time due to memory of ageing effects (Ponomarev et al., 2012), it is likely that earlier published process safety research has not received much explicit scholarly attention in China, whereas those sources may still spark new ideas. Therefore, it may be worthwhile for Chinese process safety scholars to trace back other impactful process safety literature from before the Chinese research community itself became increasingly active. Retrospective analyses such as those by Amin et al. (2019), Li et al. (2020aLi et al. ( , 2020bLi et al. ( , 2020cLi et al. ( , 2021, and Khan et al. (2015), and viewpoints and analyses of the legacy of safety pioneers such as Dr. Trevor Kletz (Mannan, 2012;Pasman et al., 2012;Vaughen, 2012) can be instrumental in this regard. It is also noteworthy that the work Dr. Sam Mannan, even though he was one of the most influential scholars in process safety (Sanders, 2018), and despite being somewhat linked to the Chinese process safety research community as observed from Fig. 5, is apparently not very well known in this community. At least, his works are (to date) not as impactful as that of other internationally leading process safety scholars. In this regard, it may be instrumental for the Chinese process safety researchers to get more acquainted with his work, and that of his collaborators, for which a recent retrospective analysis of this publications may be helpful (Li et al., 2020a(Li et al., , 2020b(Li et al., , 2020c(Li et al., , 2021. Considering the scope of the datasets as described it Section 2.1, where the focus of the current work is on Chinese process safety research published in the three core process safety journals JLPPI, PSEP, and PSP, it is also useful to make comparisons with the overall publication trends in those journals. This can provide insights especially in topics which are important in the wider process safety research community compared to those addressed in China. Considering the research areas identified in the clustered terms co-occurrence map of Fig. 6 in light of the results obtained by Li et al. (2020a), 2 it is evident that the same three main areas are found. Compared to the terms in Cluster #1 'Process safety risk management', the wider process safety research community has comparatively more focus on occupational safety, safety management, learning from incidents, and resilience than the Chinese community. Such topics can therefore be fruitful areas for Chinese process safety scholarship, where knowledge can be obtained from international process safety scholars. Within Cluster #2 'Fire & explosion process safety', the topics of the Chinese and international research communities are largely similar, but in international research there is comparatively more focus on vapor cloud explosions, dispersion models, and bleve (boiling liquid expanding vapor explosion), and blast waves. Considering Cluster #3 'Chemical process safety', the Chinese research is comparatively more focused on coal than the international research, which focuses on various other substances such as methanol, hexane, H 2 O 2 (hydrogen peroxide), as well as biomass. On the other hand, recently emerging and impactful topics related to adsorption, aqueous solutions as removal and treatment approaches appear both in the international as in the Chinese process safety community, indicating that the Chinese community keeps abreast of and helps drive this line of research.
Based on the above insights, focusing on the fact that international collaboration appears to be most beneficial to drive the work in research area #1 'Process safety risk management', selected further directions for further research activity by Chinese scholars are given. Several future research directions are given also in the complementary work to this research, by Yang et al. (2020). The ideas presented there are not repeated here. Instead, the research directions are elaborated based on what the authors consider to be the main upcoming areas in the field of process safety research, based on the presented analysis. As quantification of risks and probabilistic analyses appear to be an important focus, future work may more widely focus on the treatment and communication of uncertainties in risk analysis, see e.g. Milazzo and Aven (2012) and Goerlandt and Reniers (2016) for insights on these topics. The related topics of risk evaluation, risk acceptability, and economic aspects of safety decision-making (Ale, 2005;Reniers and Van Erp, 2016) also present fruitful avenues for future research. Further issues related to risk assessment concerns how to identify and manage risks related to black swan events (Paté-Cornell, 2012), which are also an important aspect of Sam Mannan's safety triad for process safety (O'Connor et al., 2019). An issue needing more scientific attention is interorganizational accident risk management (Milch and Laumann, 2016), which has received little attention in the wider process safety research community and appears lacking in a Chinese context. This may for instance be relevant for risks related to the transport of hazardous materials (Torretta et al., 2017), or the storing and/or bunkering of LNG at ports, see e. g. (Aneziris et al., 2020). While some research has addressed public risk perception and communication related to process industries (Huang et al., 2013), this also appears to be an under-researched area. Furthermore, with increasing connectivity and digitalization, and with heightened awareness of terrorist threats, focus on the development of effective security risk management approaches is important, see e.g. Baybutt (2017). In this context, considering the similarities and the differences between safety and security concepts and on the different knowledge domains on which these build (Li et al., 2020b), is important for developing integrated safety and security risk management strategies. Finally, while resilience engineering is a topic of rising importance in the broader safety and risk communities, only scant work in this direction is found in our analysis. Leading researchers such as Sam Mannan have made contributions towards such novel conceptualizations of ensuring safety, see e.g. Li et al. (2020aLi et al. ( , 2020bLi et al. ( , 2020cLi et al. ( , 2021, but further focus on this emerging area appears warranted.

Limitations of the work
As in any study, it is important to be aware of the limitations of the work. A first issue concerns the search strategy. As described in Section 2.1, a focused search is made in the three core process safety research journals for Chinese contributions, augmented by Chinese contributions specifically addressing process safety as a topic. This approach differs from the search strategy by Yang et al. (2020). Focusing on core journals of a research domain, as in Li et al. (2020a), is a suitable approach to identify the activity in driving knowledge carriers of a research domain (Li and Hale, 2015), and allows identification of the core authors, institutions, narrative clusters and intellectual basis in a given domain. Nevertheless, with a somewhat different search strategy, or by using other citation databases such as Scopus or SciFinder somewhat different results may be found. It is stressed as well that our current focus is only on process safety, i.e. work related to unintentional incidents and accidents. While the security of process installations (i.e. intentional events) clearly is immensely important as well, and work in that area has significantly developed as well in recent years, we have excluded security from our current scope.
An important issue in scientometric analyses is the interpretation of quantitative metrics such as number of citations and number of documents, which lay at the basis of the co-authorship, co-occurrence, and co-citation analyses as presented in Section 3. In scientometric analyses and scientometric mapping, citation metrics are used as a kind of proxy for the impact and significance of research contributions. This has however been much debated, see e.g. Garfield (1979) and Lynch (2015). These metrics should not be understood as quality measures or endorsements of the research outputs, but rather as indicators providing high-level insights in the content, structure, and evolution of a research domain. Highly cited references indicate that these significantly contribute to the development of the research domain, but as such their high citation rate does not imply correctness or impact beyond the academic context. In this context, it is noted also that as our focus is on the contents and evolution of the Chinese-authored process research field, we have not analyzed journal-specific indicators such as the journal impact factors, as this is only tangentially related to the development of the field itself.
Finally, it should be borne in mind that due to the lifecycle of research articles, for which citations accumulate over time (Ponomarev et al., 2012), the focus on number of citations in scientometrics can lead to an underappreciation of significant emerging research topics or authors. This is because more recently published articles have not yet had the time to accumulate a high number of citations, which may bias the analyses somewhat in favor of research published a longer time ago.

Conclusions
In this paper, a bibliometric mapping analysis is performed of the Chinese international process safety literature, focusing on the contributions in three core process safety journals. It is found that the earliest international academic contributions from China appeared in the mid-1990s, with the research community becoming increasingly active since approximately the mid-2000s, with a fast-increasing publication output since then.
The results of the collaboration analyses indicate that while Chinese researchers and institutions have collaborated with many international scholars located mostly in western Europe, northeastern USA, southern Australia, and Japan, by far most of the collaborations occur within China. Only with a few leading international scholars have more enduring collaborations been established, whereas the terms cooccurrence analysis indicates that these international collaborations primarily contribute to advance the knowledge in aspects related to process safety risk management. In contrast, Chinese researchers are mostly working domestically on aspects related to fire and explosion process safety, and chemical process safety. Within those research areas, while there are some differences in focus topics compared to the complete international process safety research community, the Chinese research community is well established. For the process safety risk management related research, the focus in Chinese contributions is on quantitative risk analysis, probabilistic analyses, and process accidents. A clustered co-citation analysis further indicates that especially the use of Bayesian network models and is an important influence from international authors, as well as approaches to understand and analyze domino effects. In terms of journals focusing on safety and risk management, Chinese scholars rely on knowledge from Journal of Loss Prevention in the Process Industries, Journal of Hazardous Materials, Process Safety and Environmental Protection, Safety Science, and Reliability Engineering and System Safety. Interestingly, while being a core process safety journal, Chinese researchers rely comparatively less on knowledge disseminated in Process Safety Progress. Compared to the international process safety community, and the wider safety and risk scholarly community the terms co-occurrence and co-citation analyses indicate that the Chinese research community has paid less attention to issues such as occupational safety, safety management, learning from incidents, and resilience.
Based on the findings, several suggestions for future research directions are formulated. These include increased focus on the treatment and communication of uncertainties in risk analysis, economic aspects of safety decisions and safety economics, black swan events, resilience engineering and systems approaches, interorganizational accident risk management, road and maritime transport of hazardous substances, process safety risk perception and communication, and integrated process safety and security. While scientometric methods can only provide a high-level overview of the Chinese process safety research domain, it is also recommended that further narrative literature reviews of its research domains are conducted, to obtain more detailed insights in the developments and knowledge gaps. The trends, patterns, developments, gaps, and research directions identified using the presented analyses could serve as a fruitful basis for this.
It is hoped that the insights obtained from the presented analysis can be used by Chinese researchers and international scholars to identify possible areas of future collaboration. This can invigorate the research domain, bring in new ideas, concepts, and methods, and may lead to higher productivity, strengthen the quality and acceptance of solutions for the various complex challenges facing the research domain, and ultimately help to further improve the safety performance of the process industries in China and elsewhere.