Bibliometric study with statistical patterns of industry 4.0 applied to process control

Industries are interested in offering their products or services to the consumer using high standards in process control. Industry 4.0 has emerged as a series of technological tools that can be incorporated into various processes. This research aims to perform a bibliometric analysis of the application of Industry 4.0 in process control in different sectors from 2013 to 2022 through the Scopus and Web of Science databases. The data studied were extracted from the bibliographic information of citations, abstracts, and keywords published by the articles collected. These data were processed in RStudio. As a result, it was found that the most cited articles are deep and automatic learning. Both technologies aim to reduce anomalies, increasing product efficiency, reliability, and quality. The contribution of physics in this work is shown in data mining tools, such as Bibliometrix, whose foundation is given by mathematical and statistical models, to extract data useful for future scientific studies.


Introduction
Process control research has been increasing its value and quantity due to its importance in the application in different sectors, increasing economy, productivity, and improving quality, among other aspects.There are various applications of Industry 4.0 and other technologies in process control and machine learning, which can find anomalies and then aim to stop and provide real-time solutions, offering a breakthrough in designing and implementing anomaly detection [1].This detection is a challenging task, which must consider the restrictive characteristics of the transmission data [2].Machine learning designs can find specific characteristics of what is to be elaborated, monitoring the process, which is why they have been widely used, being able to detect the quality, for example, of parts in manufacturing [3].
The Internet of Things (IoT) performs in different areas to perform activities more quickly and efficiently, one of them being a hydraulic system that helps to recognize anomalies; to achieve this, IoT sensor data, characteristics, and classification are collected through automatic and deep learning [4].In addition, IoT works in monitoring, predicting, and controlling, among other aspects, by simulating virtual models; it is applied in various stages of the cycle, in design, manufacturing, product maintenance, and in different fields [5].On the other hand, Big Data occupies a dominant position when making decisions in analysis and risk management, helping to investigate the different levels and dimensions of a process, preventing the system from being out of control, and helping managers to exclude the operational problem [6].
Bibliometric analysis has been increasing in process control studies related to science, big data, and artificial intelligence, among other aspects related to Industry 4.0.Bibliometrics has been used to identify Industry 4.0 as an important technique in the field of process control in different sectors.Limited attention has been given to bibliometric analysis based on the period of 2013-2022, which is the contribution of this article.Using the biblioshiny from the Bibliometrix library RStudio [7], it was possible to explore information extracted from Scopus [8] and Web of Science (WoS) [9] databases around the application of Industry 4.0 in recently published papers linked with process control in the last ten years.The following research questions were addressed: (i) Q1: How many research articles were published annually between 2013 and 2022 on Industry 4.0 that were applied to process control?(ii) Q2: What are the most cited articles on industry 4.0 applied in process control?(iii) Q3: What are the tools from Industry 4.0 currently in vigor in process control?
This bibliometric analysis provides us with an updated context of the applications of Industry 4.0 technologies in process control, the latest research related to the topic, and points to work on for possible future development.

Methodology
The methodology used is presented below.

Study design
Bibliometric analysis was applied as a tool frequently recognized in various academic areas to map research.This area is also called scientometric, which employs mathematics and statistics to uncover scientific interaction and relevance in an established period [10].

Data source
The Scopus [8] and WoS [9] databases were chosen for their recognition of appearing in high-quality journals and research papers.Institutional access was requested to download and confirm the content of the research files [11].

Bibliometric analysis
Graphs and tables were downloaded in BibTeX from the Scopus [8] and WoS [9] databases.Rstudio's Biblioshiny [7] was used to establish a mapping analysis of the field, offering data to understand the progress of knowledge in each period [12].It combines techniques that offer an analysis of citations, collaborations, charts, country data, research fields, keywords, h-index, etc. [13].

Limitations
The bibliometric study in the focused databases could be more perfectly aligned with each subject, given that most information is not from open access.The quantitative analysis is usually conditioned to the number of documents that are available to interested parties [14].

Search strategy
We placed a broad list of keywords in the two databases, using topics referring to Industry 4.0 (see Figure 1).On behalf of Industry 4.0, the words used were the following: data science, artificial intelligence, data analytics, data mining, Industry 4.0, data processing, big data, the internet of things, machine learning deep learning, and 3D scanning.In the representation of process control, the following keywords were included: quality control, process control, parameter control, activity control, process inspection, process supervision, analysis method, resource organization, industrial processes, process operation, operation control, quality improvement, yield improvement, and anomaly detection.These words were obtained by a cyclical process: from the articles found in the databases, more words were added.The timeframe selected considers data from 2013 to 2022 while the search was narrowed down to title and keyword filters to increase the accuracy of the search process to obtain articles from the target fields.The last day of the search was April 21st, 2023.

Results and discussion
The arguments on the stated objectives are shown in the next section, with data taken from the Scopus [8] and WoS [9] databases, it is observed that the Scopus [8] database is one of the most used databases to publish research articles related to industry 4.0 in process control, largely doubling the number of WoS [9] documents (see Figure 1).

Trends in the annual production of original articles
The number of papers indexed in Scopus [8] was higher than in WoS [9] from 2013 -2022.The overall average number of citations is usually lower than in WoS [9] compared to Scopus [8], obtaining 6.89 citations per year, while Scopus [8] has 5.71 (see Figure 2.).The research productivity scale from 2020 to 2021 in Scopus [8] and from 2020 to 2022 in WoS [9]; the trend was exponential, analyzing the graph, it is evident that there is a very rapid increase in production, although it is observed that it is from 2019 -2022.Although it is noted that it began its rise around the time of the COVID-19 pandemic, it is said that due to this, research on different topics within the application of Industry 4.0 in process control has increased, It was analyzed that due to the pandemic, different fields had to reduce the workforce due to the contagion that could occur, the machines did not produce in the same way, which is why the research for process control was increasing to ensure quality, safety, sustainability, among other aspects.

Most cited research articles
As can be seen and shown in Table 1, the most cited articles were three by Erfani S M, et al. [17], Hasan M, et al. [18], and Kwon D, et al. [19].The number of citations of the most cited paper counts with 624 average citations considering both databases.Deep learning and machine learning are topics of major interest, as seen in Table 1.Deep learning is used in the detection of anomalies in process control [17].Machine learning is used to accurately predict attacks and anomalies in different systems; some of these algorithms are logistic regression, support vector machine, decision tree, random forest, and artificial neural network.Their application in different industries aims to identify and solve optimization problems [18].Kwon D, et al. [19] and collaborators mentioned that today's industry is accustomed to the use of networks in process control, however, there are anomalies, where security is needed.Thus, deep learning appears in several sectors with applications such as natural language processing, machine vision, and so forth.In addition, Bao Y, et al. [20] declared that anomalies can occur in big data, and generate false alarms in the data, and structural performance.Using computer vision, data-cleaning algorithms, and deep learning, these issues may be diagnosed.Munir M, et al. [21] commented in their article that stopping anomalies is an important area in research, where different problems are referred to in different contexts, so diverse companies invest resources to obtain data and explore patterns of anomalies to prevent said difficulties.

Author's keywords
Keywords are a sample of research papers in scientific articles, where their frequent use reflects important points in the field of a particular study.The word cloud visualization of the Scopus [8] and WoS [9] databases in Biblioshiny [7], allowed us to obtain the most used and important keywords by the authors in studies on the application of Industry 4.0 in process control.As the results showed, the key terms for process control are anomalies and quality control, while for Industry 4.0 they are deep learning and machine learning (see Table 2).Deep learning is applied in different contexts for robust anomaly detection by combining unsupervised features and an anomaly detector.Also, deep belief networks (DBN) are claimed as a promising technique in anomaly detection, offering comparable performance with a deep automatic encoder [27].In general, machine learning tools are essential in anomaly detection; these algorithms, analyze failures, and cyberattacks, among other problems that affect the service offered [28].
Detection of anomalies is usually inconvenient for safety, industrial efficiency, etc.For instance, in the performance of industrial machines, it is necessary to know the conditions, problems, and feasible failures while analyzing and eliminating these anomalies to guarantee reliability and safety in the processes [29].Quality control improves production efficiency and increases customer satisfaction, apart from considering the application of Industry 4.0, which leads to productivity, minimizing waste, and raising production.Process monitoring involves identifying, isolating, and stopping anomalies while restoring the affected process.Fault diagnosis relies on existing parameters and considers the operational and maintenance history.Figure 3 shows that from 2016 to 2020, manufacturing processes and manufacturing were highly significant and extensively discussed in scientific literature.However, from 2020 to 2022, the most frequently mentioned topics shifted to anomaly detection, deep learning, and machine learning.This shift can be attributed to the recent advancements in new disruptive technologies.

Conclusions
This bibliographic review shows that journals and researchers are increasingly interested in studying the application of Industry 4.0 in process control.The main conclusions obtained by answering the research question mentioned in the introduction section are the following: (i) the production of original articles in the established field is currently growing exponentially; (ii) the most cited articles in process control are related to machine and deep learning; (iii) deep learning and machine learning are the most used technologies in process control.From the point of view of physics, this research contributes to the application of data mining strategies as a valuable tool to explore big data currently generated by the scientific community.

Figure 2 .
Annual trend of total WoS and Scopus publications and citations from 2013 to 2022[16].

Figure 3 .
Figure 3. Evolution of the subject matter of publications on Industry 4.0 in process control, considering only the Scopus database (2013 -2022) [7].

Table 1 .
Most cited articles on industry 4.0 in process control.
*TC: total of citations.