An engineering front refers to the key direction that is forwardlooking, leading, and exploratory. It has a major influence  and a leading role in the future development of engineering  science and technology and serves as an important guide  for cultivating the capabilities for innovation in the field of  engineering science and technology. The front is focused  on the theoretical research or application development of  engineering science and technology. Engineering fronts are  divided into engineering research fronts and engineering  development fronts. In this research, the engineering front  identification is based on public data and expert research;  therefore, it does not involve nonpublic domains.  

Underpinned by experts’ evaluation and data, the 2021  Global Engineering Fronts project has adopted multiround  interactions between experts and data for iterative research  and analysis, realizing a deep integration of judgments of  experts and data analyses. In 2021, 93 global engineering  research fronts and 93 global engineering development fronts  are selected, with 28 engineering research fronts and 28  engineering development fronts being listed as the key focus  of interpretation. The distribution of engineering research  and engineering development fronts among the nine fields is  shown in Table 1.1.  

The research of fronts consists of three stages: data  preparation, data analysis, and expert review. In the data  preparation stage, domain, library, and information experts  revise the initial literature and patent data to clarify the scope  of data mining. In the data analysis stage, the co-citation  clustering method is used to obtain clustered literature topics  and ThemeScape patent maps. In the expert review stage, the  fronts are gradually selected and determined through patent  map interpretation, expert panel discussions, questionnaire  surveys, and other methods. Then the list of the Top 10 fronts  is modified, and the front naming is improved based on  the performance of the front in literature or patent data. To  address the problem of lacking of novelty due to algorithm  limitations or lags in data mining, experts from different fields  were encouraged to check the results of the data analysis to  fill in the gaps and nominate engineering fronts. A flowchart  of the operation procedure of the Global Engineering Fronts  project is illustrated in Figure 1.1, in which the green, purple,  and red boxes indicate the data analysis, expert research, and  multiround iterative interactions between experts and data,  respectively.  

《1  Identification of engineering research  fronts》

1  Identification of engineering research  fronts

 The identification of the engineering research fronts is  performed in two steps. The first step involves determining  the clustered literature topics through the clustering method  of co-citation according to the SCI journal papers and data of  

《Table 1.1》

Table 1.1 Distribution of engineering research and engineering development fronts among the nine fields  

《Figure 1.1》

Figure 1.1 Operation procedure of the Global Engineering Fronts project

conference proceedings collected from the Web of Science  Core Collection of Clarivate. The second step is defining the  engineering research fronts through expert nomination.  Alternative engineering research fronts that were identified  through expert argumentation and refinement went through  questionnaire surveys and multiple rounds of expert  discussions, yielding 93 engineering research fronts in the  nine fields.  

《1.1 Acquisition and preprocessing of paper  data》

1.1 Acquisition and preprocessing of paper  data

 Clarivate mapped the fields of Web of Science and the nine  academic division fields of the CAE and obtained a list of  journals and conferences in each field. After the correction  and supplementation by domain experts, the sources for  data analysis in the nine fields were determined to be 12 215  journals and 44 153 conferences. In addition, for articles from  72 multidisciplinary sciences journals such as Nature and  Science, the field of each article was reassigned to the most  relevant subject area according to the subjects cited in its  references. Accordingly, the articles and conference papers  published between 2015 and 2020 were retrieved (the cut-off  date of the citations was January 2021).  

For each field, Clarivate comprehensively considered the  differences between journals and conferences, the publication  year, and so on. Next, the list of aforementioned papers  was retrieved and extracted. By processing journals and  conference proceedings separately, papers with high impact  that are ranked among the top 10% of the citations were  selected as the original dataset for the analysis of the research  hotspots, as shown in Table 1.1.1.  

《1.2 Mining of clustered literature topics》

1.2 Mining of clustered literature topics

 Through the co-citation clustering analysis of the top 10%  highly cited papers in the aforementioned nine data datasets,  all the clustered literature topics in the nine fields were  obtained. The topics of papers published during 2019–2020  were selected according to the number of core papers, total  number of citations, and proportion of consistently cited  papers. Thereafter, 25 different literature topics were obtained.  The topics of the papers published before 2019 were selected  according to the mean publication year of core publications  and the proportion of consistent citations. Consequently, 35  diverse literature topics were extracted. Overlapping topics  were replaced by topics that did not intersect with other fields.  In addition, subjects that were not covered by clustering topics  were extracted separately by keywords. Finally, 775 clustered  literature topics in the nine fields were obtained (Table 1.2.1).  

《1.3 Determination and interpretation of research fronts  》

1.3 Determination and interpretation of research fronts  

While processing and mining the paper data, domain experts  put forward research front issues by a comprehensive analysis  of data pertaining to science and technology news and  national strategic layouts of different countries, and integrated  them into each stage of front determination.  

In the data preparation stage, the library and information  experts transform the front research questions raised by the  domain experts into search formulas, which are an important  part of the initial data source. In the data analysis stage, for  subjects that are not covered by clustered literature topics,  the domain experts provide keywords, representative  papers, or representative journals to support Clarivate for  customized search and mining. In the expert review stage,  the domain experts will check for omissions based on the  clustered literature results provided by Clarivate and conduct  the second round of nomination for fronts that do not exist in  the data mining results but are considered important. Library  and information experts provide data support. Finally, the  domain experts merge, revise, and refine the engineering  research front topics obtained through data mining and  expert nomination. Subsequent to questionnaire surveys and  multiple rounds of conference discussions, approximately 10  engineering research fronts were selected for each field.

 In each field, three key research fronts were selected  according to the development prospects and the significance.

《Table 1.1.1》

Table 1.1.1 Number of journals and conferences in each field and number of top 10% highly cited papers

《Table 1.2.1 》

Table 1.2.1 Statistics of co-citation clustering results in each field

Authoritative experts in the front direction were invited to  interpret the fronts in detail from the perspectives of national  and institutional layout, cooperation networks, development  trends, and R&D priorities.  

《2  Identification of engineering  development fronts  》

2  Identification of engineering  development fronts  

The identification of the engineering development fronts is  primarily performed using two methods. First, based on the  Derwent Innovation patent database of Clarivate, the top 10  000 patent families of 53 subjects in the nine fields with high  citations were clustered, and 53 ThemeScape maps were  obtained. The domain experts interpreted the alternative  engineering development fronts from these maps. The  second approach involves nomination by expert or patent  analysis by small peer group. The alternative development  fronts obtained through these two methods went through  questionnaire sur veys and several special seminars.  Consequently, approximately 10 engineering development  fronts were identified in each field.  

《2.1 Acquisition and preparation of the  ThemeScape maps》

2.1 Acquisition and preparation of the  ThemeScape maps

 In the data preparation stage, based on the Derwent  Innovation patent database, Clarivate developed the initial  patent data retrieval scope and search strategies for the 53  disciplines of the nine fields using the Derwent World Patents  Index (DWPI) Manual Codes, International Patent Classification  numbers, United States Patent Classification numbers , and  other patent classification numbers, and specific technical  keywords. Domain experts deleted, supplemented, and  improved the DWPI Manual Codes to determine the patent  retrieval criteria; further, the nominated alternative front  topics were selected, which were then transformed into patent  search formulas by library and information experts. Clarivate  integrated the above two parts of the search formulas,  determined the patent search formulas of the 53 disciplines,  searched in the “enhanced patent data - DWPI and DPCI  (Derwent Patent Citation Index)” collection, and obtained  the patent literature of the corresponding disciplines. The  retrieved patents were published between 2015 and 2020; the  cut-off date of the citations was January 2021.  

To further concentrate patent literature, the millions of patent  documents were screened according to the annual average  number of citations and technical coverage width indicators,  thereby obtaining the top 10 000 patent families in each  discipline.  

《2.2 Mining of patent topics  》

2.2 Mining of patent topics  

Semantic similarity analysis of patent texts were conducted  for the top 10 000 highly cited patents on 53 disciplines in  the nine fields. Based on literature topic clustering using  DWPI titles and abstracts, 53 ThemeScape patent maps were  obtained, which effectively display the distribution of the  engineering development techniques and show the overall  technical information of the collected patents in the form of  keywords.  

Experts from various fields, with the assistance of library and  information experts, selected the engineering development  fronts from ThemeScape maps, merged similar fronts, and  determined the final development fronts. Finally, they  selected the alternative engineering development fronts  of each specialty group. To avoid missing emerging fronts,  domain experts interpreted the data from patents with few  citations and poor correlation in the ThemeScape maps.  

《2.3 Determination and interpretation of  development fronts  》

2.3 Determination and interpretation of  development fronts  

While processing and mining the patent data, domain  experts identified issues on development front based on a  comprehensive analysis of other data, such as science and  technology news and national strategic layouts of different  countries, and integrated them into each stage of front  determination.  In the data preparation stage, the library and information  experts transformed the key front issues raised by the domain  experts into patent search formulas as an important part  of the basic dataset.

In the data analysis stage, the domain  experts conducted the second round of front nomination to supplement the emerging technology points that are  significant, but have been submerged in data mining with few  patents. In the expert review stage, the domain experts studied  highly cited patents, and the library and information experts  assisted them in interpreting patent maps from multiple  perspectives, such as “peaks” and “blue oceans”. Finally, the  domain experts merged, revised, and refined the interpreted  results of the patent maps and fronts nominated by experts to  obtain candidate engineering development fronts, and then  selected approximately 10 engineering development fronts in  each field through questionnaire surveys or multiple rounds  of seminars.

 In each field, three key development fronts were selected  according to the development prospects and the significance.  Authoritative experts in the front direction were invited to  interpret the fronts in detail from the perspectives of national  and institutional layout, cooperation networks, development  trends, and R&D priorities.  

《3  Terminologies  》

3  Terminologies  

Publications/Papers: This includes peer-reviewed and  published journal articles, reviews, and conference papers  retrieved from Web of Science.  

High-impact papers: Papers that are in the top 10% in terms  of citation frequency are considered to be of high impact,  taking into account the year of publication and journal subject  category.  

Clustered literature topic: A combination of topics and  keywords obtained through a co-citation clustering analysis of  high-impact papers.  

Core papers: Depending on how the research front is  obtained, core papers have two meanings. If the paper  originates from a front revised by data mining experts, then  the core paper is considered as a high-impact paper. If it  comes from a front nominated by domain experts, the core  paper is included in the top 10% of papers in terms of citation  frequency obtained using the corresponding search formula.  

Percentage of core papers: The proportion of core papers  in which a country or institution participates among the  total number of core papers produced by all countries or  institutions.

Citing papers: Collection of papers that have cited core  papers.  

Citation number: The number of times the paper has been  cited by the Web of Science Core Collection of Clarivate .  Mean publication year: Average publication years for all  papers among the clustered literature topics.  

Citation velocity: An indicator used to measure the growth  rate of the cumulative number of citations for a certain period.  In this study, the citation velocity of each paper begins with  the month of publication, and the cumulative number of  citations per month was recorded.  

Consistently cited papers: Papers included in the top 10% in  terms of citation velocity.  

Highly cited patents: The around top 10 000 DWPI families  ranked by the average annual DPCI citations.  

Core patents: According to the different ways of obtaining  the development front, core patents have two meanings. If it  comes from the front of the patent map, the core patent refers  to the highly cited patent; if it is from the front nominated by  domain experts, the core patent refers to all patent obtained  by topic search.  

Percentage of published patents: The proportion of  published patents in which a country (priority country) or  institution participates among the total number of published  patents produced by all countries or institutions.  

ThemeScape map: A themed landscape representing the  overall outlook of a specific industry or technical field. It is a  visual presentation in the form of a map obtained by analyzing  the semantic similarity of value added DWPI information of  patents to gather the patents of related technologies.  

Technical coverage width: It is measured by the number of  DWPI Classes to which each DWPI patent family covers. This  indicator can reflect the breadth of the technology coverage of  each patent.  

Specialty division criteria system of the academic divisions  of the CAE: This includes 53 specialized fields covered by nine  academic divisions of engineering science and technology. It is  determined according to the Academic Divisions and Specialty  Division Criteria of the Chinese Academy of Engineering for the  Election of Academicians (for Trial Implementation).