1. Introduction
Rapid urbanization in metropolitan areas presents a multifaceted set of challenges, including increasing traffic congestion, the exacerbation of environmental degradation, and an increasing scarcity of available land resources. These challenges are intrinsically linked to the increased demand for engineering endeavors, with a particular emphasis on civil engineering projects
[1],
[2]. During the processes of urbanization and industrialization, the emission of greenhouse gases contributes to global warming, which leads to climate change
[3],
[4]. In this regard, the construction engineering industry (CEI) plays an important role in promoting sustainable development, which is crucial for realizing the goal of carbon emission reduction and carbon neutrality in China. As per the findings delineated in a report on China’s building energy consumption, the CEI sector produced a total carbon emission volume of approximately 4.93 billion tonnes in 2020. Notably, this substantial figure represents a significant proportion of the total carbon emissions in China, amounting to 51.3%. Further examination of these emissions reveals distinct contributions from various segments within the CEI sector: The production of building materials constituted 55.2% of the total carbon emissions from this sector, while building construction activities contributed 2.0%. A substantial portion—approximately 42.8%—was attributed to the operational phase of buildings, underscoring the noteworthy role of these components in the carbon emissions landscape
[5].
Infrastructure construction plays a significant role in driving the growth of economic and national development in the long term. In urban cities, the construction of infrastructure encounters various social, environmental, and economic challenges, which arise before and during the construction process, presenting significant difficulties for construction engineering and management (CEM). CEM is a pivotal domain within the broader architecture, engineering, and construction industry, encompassing a spectrum of construction-related activities and their attendant management processes. These multifaceted activities are conducted in collaboration with various stakeholders, resulting in intricate networks of interactions and interdependencies.
The characteristic of CEM involves dealing with the risks and occurrence of unforeseen factors associated with engineering projects
[6],
[7]. A series of methods have been proposed to address challenges in CEM for different sizes of engineering projects, such as multi-criteria decision-making (MCDM) techniques
[8],
[9],
[10],
[11] and machine learning methods
[12],
[13],
[14],
[15]. These methods can be categorized into MCDM and intelligent techniques (collectively termed “smart techniques”). In the context of large-scale engineering projects, there is a notable emphasis on the utilization of numerical simulations and machine learning methods. These analytical tools primarily hinge upon the acquisition of precise and comprehensive datasets through various measurement techniques. However, it is important to emphasize that the intricate influence of human factors may be insufficiently acknowledged within this rigorous analytical framework. These human factors can encompass a broad spectrum of elements, ranging from the decision-making process of project managers to the operational behaviors of onsite workers. Thus, there is a discernible gap in comprehensively integrating human factors into analytical frameworks, as the latter are heavily reliant on empirical data and computational models. In addition, limited measured equipment is utilized in smaller engineering projects in order to reduce construction costs, which hinders access to adequate data. Moreover, owners are often reluctant to share the data from their engineering projects, making the data for construction projects limited and incomplete. Under these circumstances, existing approaches face considerable difficulties in handling the limited, incomplete, and multi-source data from engineering projects.
In regard to CEM, with its limited data, MCDM approaches have been utilized to address the issues associated with engineering projects in past decades. MCDM is a potential technique for addressing complex decision issues; it considers the various criteria involved in order to research a problem and then provides solutions involving diversified indexes under fuzzy and uncertain environments. MCDM techniques follow three main steps: ① Relative criteria, attributes, and alternatives are defined and utilized to build a decision hierarchy; ② numerical values or linguistic terms provided by decision-makers are utilized to assess the relative importance of items; and ③ fuzzy techniques are applied to process the data and rank the considered alternatives.
However, there are several issues related to the application of MCDM techniques in CEM. Firstly, different types of fuzzy sets have been employed to assess the relative significance of criteria on a qualitative scale. The reliance on subjective judgments from decision-makers introduces variability into the ranking of alternatives. The execution of engineering projects is heavily dependent on the expertise and experiences of professionals, which presents significant challenges in the practical application of MCDM techniques in CEM
[16],
[17]. Secondly, the number of participants in MCDM will make results in some kind of deviations. Most importantly, it is difficult to use fuzzy MCDM techniques to deal with the large volume of data from an engineering project. Thirdly, during decision analysis, fuzzy techniques are merely applied without the integration of other methods such as an analytical hierarchical process
[18],
[19] or fuzzy analytical hierarchical process
[6],
[20]. In practice, the data generated from projects comes from different sources, such as safety construction logs and measured devices. Large volumes of data are generated through these advanced technologies, which cannot be dealt with via simple MCDM approaches. Thus, intelligent techniques have been implemented, along with data science, to deal with the large volumes of data.
MCDM and intelligent methodologies can be effectively combined to address the multifaceted challenges associated with engineering projects. Their synergy offers a robust framework for tackling the diverse issues and considerations that arise in such projects, given the projects’ varying attributes and complexities. The research gaps in this area can be summarized as follows: ① The emphasis on MCDM methods in the literature is a limitation, as certain studies have exclusively concentrated on MCDM methods, without duly recognizing the potential contributions of intelligent techniques in CEM; ② there is a neglect of expert judgments in intelligent technique reviews; and ③ there is an overemphasis on the specific project stage, such as the particular stages within engineering projects. It is crucial to note that CEM encompasses a broad spectrum of activities, including pre-design, planning, construction, operation, and maintenance. Addressing these research gaps will contribute to a more comprehensive understanding and effective application of intelligent methodologies in CEM.
Thus, we have undertaken a comprehensive review that encompasses the integration of subjective expert judgments with objectively measured data processed through smart techniques for engineering projects. The novelty of this review lies in its emphasis on the broad-scope application of smart techniques across the life cycle of an engineering project. This analysis underscores the holistic integration of smart techniques under diverse scenarios, ranging from scenarios with limited data to those with copious data. It delineates a seamless transition from MCDM techniques to intelligent techniques, highlighting the synergistic role of these techniques in advancing CEM practices. Its comprehensive approach positions this review as an important exploration of smart techniques in the dynamic realm of CEM.
The remainder of this study is structured as follows: Section 2 offers a succinct analysis of CEM, elucidating its conceptual framework and characteristics. Section 3 delineates the outcomes derived from a bibliometric analysis. Section 4 rigorously examines recent advancements in smart techniques within the domain of CEM. Section 5 provides the anticipated future directions for CEM. The study concludes with final remarks in the last section.
2. A brief analysis of CEM
In this section, a concise examination of CEM provides valuable insights into the utilization of smart technologies within the construction sector. It also outlines the attributes that define CEM within the context of construction practices.
2.1. A conceptual framework of the application of smart techniques in CEM
CEM is undergoing a profound digital transformation, driven by the integration of smart techniques. This review describes a comprehensive framework that encapsulates the application of smart technologies across the life cycle of engineering projects. This comprehensive framework takes a holistic approach, embracing various stages, scenarios, and applications. It emphasizes how smart techniques, encompassing MCDM and artificial intelligence (AI), are reshaping the CEM landscape.
Data acquisition is a fundamental aspect of CEM that spans engineering projects. A wide range of data types are harnessed from diverse sources, encompassing geological, structural, and maintenance data, among others. The intrinsic characteristics of CEM also pose considerable challenges (
Fig. 1(a)), which will be discussed in Section 2.2. Moreover, the data collected in CEM exhibits a diverse set of properties (
Fig. 1(b)), further complicating the data-processing landscape. This data includes both structured and unstructured data and varies in format and organization. Multidimensional data adds another layer of complexity, as it involves information from different dimensions. Time-series data captures the temporal dimension of construction projects, enabling the analysis of trends and patterns over time. Data interdependencies highlight the intricate relationships between various data points, influencing decision-making processes. Geospatial information adds a spatial context to the data, enabling precise location-based analysis. The sources of CEM data are heterogeneous, spanning textual, image, and video data categories and thereby reflecting the multidimensional nature of construction information. Such diversity necessitates advanced approaches to data management and analysis.
The life cycle of CEM unfolds in three pivotal stages (
Fig. 1(c)): planning and design, engineering construction, and operation and maintenance. The planning and design stage represents the project’s blueprint, establishing a foundation for success. Here, the use of MCDM techniques, which are deeply rooted in mathematical principles, facilitates decision-making by evaluating and ranking design alternatives. The robustness of MCDM methodologies becomes evident in situations characterized by data scarcity. MCDM methods are invaluable in CEM, as they are based on the input of expert judgments and facilitate the decision-making process.
As a project progresses to the engineering construction stage, the volume of data burgeons, necessitating a transition from MCDM to data-centric approaches. Construction sites become veritable data generators, producing real-time information on project progress and quality control. AI-driven systems excel in processing these extensive and diverse datasets, uncovering patterns and trends that underpin timely and informed decision-making. The dynamic characteristics of construction sites necessitate responsive, data-driven approaches to address challenges, and intelligent techniques excel in fulfilling this requirement.
The operation and maintenance stage signifies a transition, as the focus shifts from construction to infrastructure functionality and sustainability. Data continues to be collected, especially in the form of maintenance and inspection data. Here, the integration of smart techniques with the expertise of construction professionals becomes indispensable. Predictive maintenance models utilize historical data to anticipate maintenance needs, thereby mitigating the risk of costly equipment breakdowns and optimizing the allocation of resources. Within this intricate landscape, smart techniques are the cornerstone for effective data processing in CEM (
Fig. 1(d)). Smart techniques encompass both MCDM techniques and intelligent techniques, with Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and ViseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR) serving as key approaches within MCDM techniques. These techniques offer a robust means of tackling the data generated by engineering projects. In essence, smart techniques empower decision-makers with the tools needed to navigate the multifaceted world of construction, ensuring that their choices are well-informed and responsive to the dynamic challenges posed by construction projects.
The application of MCDM and intelligent technologies in CEM holds tremendous promise (
Fig. 1(e)). These innovations aim to realize ambitious goals, including carbon neutrality and sustainability, reshaping not only the ways in which construction projects are executed but also the long-term environmental and societal impact. This review also provides guidance on how to realize intelligent construction in an engineering project. Intelligent construction is a novel construction model that is deeply integrated with next-generation information technology and is designed to cover the life cycle of an engineering project. This approach facilitates the digitization of the design process, the industrialization of construction, and the modernization of management, serving as a leading path to accelerate the transformation and upgrading of the construction industry and bring about fundamental changes in production methods.
The framework for applying smart techniques in CEM encompasses the life cycle of engineering projects, addressing unique challenges posed by data dynamics, complexity, uncertainty, and diversity. By assisting in the navigation of the planning and design, engineering construction, and operation and maintenance stages, smart techniques facilitate data-driven decision-making, enhance efficiency, and drive sustainability.
2.2. Attributes of CEM
CEM is a collaborative activity with a specific construction period and under specific requirements. Although the sizes of engineering projects vary widely, these projects have common characteristics, which are summarized below.
2.2.1. Dynamics
CEM is a field characterized by its dynamic and ever-evolving nature. In the realm of CEM, projects unfold within a dynamic environment shaped by technological advancements, regulatory changes, and shifting economic landscapes. The management of construction and engineering projects requires a keen ability to adapt to these dynamic factors, ensuring successful project completion. Project managers in CEM must navigate complex challenges, such as tight deadlines, budget constraints, and unforeseen obstacles, all while staying abreast of the latest innovations in construction technology and sustainable practices. The dynamic nature of CEM demands a proactive approach to problem-solving, effective communication, and strategic decision-making to address the multifaceted challenges inherent in the industry.
2.2.2. Uniqueness
Each construction project is unique, as evidenced by varying project sizes, construction environments, budgets, and contractor requirements, which place distinct demands on project managers and heighten their challenges. Moreover, the uniqueness of construction products stems from their different functional requirements, forms, and designs. Different kinds of construction methods and machines must be utilized according to the characteristics of construction production. Moreover, as construction projects change, construction procedures and labor resources must be rescheduled and rearranged. Construction organization design is usually based on the specific attributes of a project; in other words, the construction schedule, arrangement of labor resources, and finances of one engineering project cannot simply be applied to other projects without modifications. Instead, lessons from previous engineering projects—including successful experiences and failures—provide guidance for project managers’ decision-making.
2.2.3. Complexity
Tasks and participants are the main sources of complexity in a construction project. In practice, various other sources (e.g., labor, machines, materials, technologies, and weather) also increase the complexity of a project. There are three main types of complexity in a construction project: organizational, decision-making, and control complexity. Firstly, organizational complexity is defined by the stakeholders involved in the project and arises from the differences between departments. Here, the “organization” includes the owners, contractors, supervision unit, construction unit, and stakeholders. Thus, a greater number of established departments increases the management complexity. In addition, participants at different levels in the organizational structure shoulder different responsibilities according to their job positions, resulting in organizational complexity. Secondly, the execution of construction activities can be viewed as a form of decision-making by the participants. In this regard, differences and interactions between participants create decision-making complexity. Thirdly, a project is intended to achieve various objects. Thus, it is necessary to control the procedures in order to achieve these goals. Accurate coordination of resources is a source of complexity here, as is the use of advanced technologies and construction materials.
2.2.4. Uncertainty and fuzziness
Uncertainty and fuzziness are obvious attributes of CEM, especially in the construction of underground infrastructure. Uncertainty in engineering projects arises from incomplete knowledge about a system, process, or outcome, often due to factors such as insufficient information, variability in input data, and unpredictable environmental conditions. This uncertainty, influenced by a lack of foresight, unrealistic expectations, and misunderstanding of objectives, leads to numerous possible outcomes for the research objective. When the likelihood of achieving the objective becomes unpredictable, uncertainty transitions into risk. Additionally, fuzziness, characterized by vagueness or ambiguity in the decision-making criteria, further emphasizes the imprecision in the available data. In general, the uncertainty and fuzziness are higher in more complex construction projects. For example, geological parameters have high fuzziness, since the soil is a three-phase system composed of solids, liquids, and gases
[21],
[22]. The values of geological parameters are obtained by drilling boreholes
[23],
[24]. Other factors that inevitably increase uncertainty and fuzziness in CEM include the weather conditions, reliability of the suppliers of construction materials
[25], management quality, and supervision of workers at the construction site.
2.2.5. Incompleteness
Construction projects, planning processes, or management processes may lack essential, comprehensive, or detailed information, rendering decision-making and analysis challenging due to the inherently incomplete state. The attribute of incompleteness highlights the presence of gaps, uncertainties, or missing data within various aspects of CEM, including project documentation, stakeholder communication, design specifications, or project schedules. Data is mainly obtained via measured devices, safety logs and others. However, some types of data are difficult to access, due to confidentiality principles in the construction unit. In addition, construction units usually adopt a limited number of measurement devices and decrease the number of measured items in order to reduce construction costs. In practice, experts are usually invited to perform a risk analysis based on their engineering experience; however, data from experts has a certain degree of subjectivity under some circumstances.
2.2.6. Systems
CEM is a highly technical and complicated activity that involves various aspects, such as project organization, design, and construction. In CEM, a system is a structured and interconnected set of elements, components, or processes that work together to achieve specific objectives or functions within the construction project or management context. The system is designed to facilitate the efficient planning, execution, monitoring, and control of construction projects or management activities. It encompasses various dimensions, including project-management systems, decision-support systems, and communication systems, all harmoniously integrated to ensure the successful realization of construction projects. In a system, the elements are interrelated and mutually restricted. In essence, a CEM system encapsulates the multifaceted and organized approach employed in the construction industry to address the diverse challenges and tasks encountered during the life cycle of construction projects.
3. Bibliometric analysis
A bibliometric analysis is an effective way to quantitatively deal with the large volume of datasets in research publications; it can provide statistical insights into the publications and knowledge structure in a research field. Thus, we used a bibliometric analysis to examine the state of the art of smart techniques in CEM. The research methodology, which is provided in Section S1 in the Appendix A, involved four aspects: data collection, an analysis of publications, an examination of the journals publishing the relevant papers, and keywords co-occurrence.
3.1. Data collection
Smart technologies can be categorized as MCDM or intelligent techniques. Thus, the bibliometric analysis was divided into two sections. Publications on the use of MCDM and intelligent techniques in CEM were retrieved in order to perform a review. A Scopus database search was utilized. The keywords used were divided into two groups: ① keywords related to the application of MCDM techniques in CEM; and ② keywords related to the application of intelligent techniques in CEM. For the first group, a database search was conducted according to the following rules: (“multi-criteria decision-making” OR “fuzzy set theory”) AND (“civil engineering” OR “construction engineering” OR “construction management”). The search was set to the article title, abstract, and keywords in order to cover all publications with the chosen keywords in related research fields. The subject area was confined to “engineering,” and the search period was set from 2000 to 2022. Three restrictions were placed on the scope of the studies and research:
Type of publication: articles and reviews;
Language: English;
Time period of publication: 2000–2022;
The search yielded a total of 636 publications.
A similar procedure was then applied to the second group—the use of intelligent techniques in CEM. The search rule was defined as follows: (“construction engineering” OR “construction management” OR “construction industry” OR “civil engineering”) AND (“intelligent techniques” OR “artificial intelligence” OR “machine learning” OR “deep learning”). The search yielded a total of 601 studies.
3.2. Analysis of the publications
Annual publications on the use of MCDM techniques in CEM exhibited a steady upward trend during 2000–2022, especially after 2020, when the annual number of publications grew to more than 80. As shown in
Fig. 2(a), annual publications on this topic increased steadily over the period of 22 years. More than 85% of the articles were published after 2010, indicating that research interest in the use of MCDM-based technologies in CEM increased greatly after that year. The ExpGrow1 function was adopted to fit the data of annual publications better, with adjusted R-square (Adj.R) values of 0.961, as displayed with a red line with a 95% confidence interval in
Fig. 2(a). Over 100 publications were estimated for 2022, based on the fitted function.
Annual publications on the use of intelligent techniques in CEM exhibited an increasing upward trend after 2017; the distribution of the 601 articles is displayed in
Fig. 2(b). Research interest in the application of intelligent techniques to CEM began in the 2000s, when 47 articles were released, accounting for nearly 8% of the considered publications. A steady increase in publications during 2000–2017 then occurred. From 2017 onwards, a considerable number of articles were published: nearly 480 articles, accounting for almost 80%. This steep growth in publication indicates that the era of intelligent techniques has arrived. A fitting curve was set using the Adj.R values (0.960) via the ExpGrow1 function, which estimated the publication of over 200 articles in 2022, clearly demonstrating the mainstream development of CEM.
3.3. Journal analysis
The 636 publications on the application of MCDM techniques in CEM were mainly distributed among 36 journals.
Fig. 3(a) displays the journals with the greatest number of considered articles and citations. The impact factors (IFs) of 12 of these journals are higher than 3.00, while the CiteScores (CS) are higher than 5.00, except for the
Journal of Intelligent & Fuzzy Systems (
JIFS). Moreover, each of several domain journals had over 10 publications on the relevant topics, namely,
Expert Systems with Applications (
ESWA),
JIFS,
Information Sciences (
IS),
Journal of Cleaner Production (
JCP),
Computers & Industrial Engineering (
CIE), and
Journal of Civil Engineering and Management (
JCEAM). Except for
CIE and
JCEAM, these journals also had over 3000 relevant citations each. Among these journals,
ESWA had the greatest number of relevant publications (49 papers) and citations (nearly 5700).
The 601 articles on the application of intelligent techniques in CEM were found to be distributed among 45 journals. Most of these publications were in specific journals, as presented in
Fig. 3(b). Except for
Advances in Civil Engineering (
AICE) and
JIFS, the IFs and CSs of the listed journals are generally over 4.00 and 6.00, respectively
. Among the journals displayed in
Fig. 3(b),
Automation in Construction (
AIC) is ranked in the first position, with 131 publications and nearly 3800 citations. In addition, the three journals
AIC,
Journal of Construction Engineering and Management (
JCEM), and
Journal of Computing in Civil Engineering (
JCCE) have over 30 relevant publications each. The citation counts for
AIC and
JCEM each exceeded 1000. It should be noted that the data are based on those from 2022.
3.4. Co-occurrence analysis of keywords
In this study, the co-occurrence analysis of keywords was manifested using a knowledge map via VOSviewer
[26]. VOSviewer is a freely available scientific visualization computer software based on Java that focuses on the processing of literature data. Its strong graphic visualization of scientific knowledge and its capacity to deal with large volumes of datasets from different kinds of databases are the main advantages of this software in comparison with other bibliometric software
[26].
A total of 4577 keywords related to the application of MCDM techniques for CEM were first collected from 636 publications. A further refinement process retained 83 keywords, applying a minimum occurrence threshold of 15.
Fig. 4(a) displays the co-occurrence network of these keywords. Notably, keywords such as “decision making,” “multicriteria decision making,” “fuzzy set theory,” “construction industry,” and “fuzzy sets” feature prominently in terms of occurrence frequency. “Decision making” emerges as a central keyword, with 519 occurrences and a significant total link strength (TLS) of 2184, indicating that it is the focus of widespread attention, particularly in conjunction with fuzzy sets in the context of MCDM techniques in CEM. Dense links between “decision making” and other keywords, such as “construction industry” (TLS = 61) and “sustainable development” (TLS = 67), reflect comprehensive research into decision-making techniques for sustainable construction. The quantitative co-occurrence analysis is summarized in Table S1 in Appendix A, offering insights into keyword strength. A clustering approach was used to categorize nodes with closely correlated relationships into four clusters, each denoted by distinct colors in
Fig. 4(a).
A total of 6031 keywords related to the application of intelligent techniques in CEM were obtained from 601 publications. Finally, only 92 keywords remained, after a further refinement process that applied a minimum occurrence threshold of 15. As shown in
Fig. 4(b), the keywords with high occurrence frequency are “machine learning,” “construction industry,” “deep learning,” and “project management,” whose occurrences are 198, 173, 156, and 92, respectively, and whose TLSs are 1017, 1064, 742, and 538. The keyword “machine learning” is located in the center of the network and has the largest size compared with the other nodes. The links between “construction industry” and “risk assessment” and “occupational risks” are thicker in cluster 3, with TLSs of 247 and 276, respectively. The cluster analysis results are presented in Table S2 in Appendix A. The five clusters in
Fig. 4(b) can be viewed as different hot research topics in recent years. For example, intelligent techniques such as machine learning and data mining have been frequently utilized in construction projects
[27],
[28],
[29],
[30],
[31],
[32],
[33].
4. Smart techniques in CEM
This section elucidates MCDM methodologies, research areas pertaining to intelligent techniques, and the application of advanced methodologies in underground space exploitation (USE).
4.1. MCDM techniques
MCDM techniques can be utilized to address various problems in CEM related to fuzzy and uncertain environments. Fuzzy set theory was originally developed to resolve and simulate the intrinsic vagueness of the cognitive procedures of human beings
[34],
[35],
[36]. Membership functions and corresponding fuzzy numbers are the basic elements for decision-making. Different kinds of fuzzy sets provide significant and innovative ways to extend various MCDM approaches
[16],
[17]. In addition, MCDM is regarded as a smart technique in CEM due to its ability to handle complex and multidimensional decision problems effectively. Using MCDM methods, decision-makers can systematically evaluate and prioritize different alternatives based on the determined criteria. This approach allows decision-makers to make informed choices that optimize multiple objectives simultaneously. Thus, MCDM serves as a smart decision-making method in CEM by providing a structured framework for navigating complex decision scenarios and achieving optimal solutions. Typical MCDM methods are illustrated in Table S3 in Appendix A.
4.2. Research topics related to intelligent techniques
The primary objective of intelligent techniques is to grant machines perception, learning, and cognition abilities. More specifically,
perception refers to giving machines the ability to perceive and process information from external stimulus.
Learning refers to giving machines the ability to imitate the learning ability of human beings, such as how to interact with the surrounding environment and how to discover the hidden patterns in a large volume of datasets in order to realize data-driven estimations of research tasks.
Cognition refers to giving machines cognitive ability, which mainly includes knowledge representation, understanding, and reasoning (KRUR). There are six main research topics based on these three aspects of intelligent techniques, as summarized in
Fig. 5. Machine learning is a fundamental component of intelligent techniques that allows systems to automatically learn from data and make data-driven decisions without explicit programming. It underpins many advanced techniques by providing the computational power and algorithms necessary to analyze data, recognize patterns, and make predictions.
4.2.1. Big data technology
Big data technology encompasses the tools and methodologies for handling, processing, and analyzing large and complex datasets, including structured, unstructured, and semi-structured data
[37]. It includes data-storage solutions, data-processing frameworks, and data-management and analytics tools. Big data provides the vast and diverse datasets necessary for training machine learning models, while scalable and real-time data-processing capabilities enable efficient model training and evaluation. This infrastructure allows machine learning algorithms to access, process, and learn from extensive data, enhancing their accuracy and robustness.
The main characteristics of big data can be summarized as follows
[38],
[39]: ① It is generated and updated at high speed; ② it is generated in large volumes and high dimensions; ③ it is diverse (i.e., the data is generated from different sources); and ④ its high value stems from its high veracity, significant value, dynamic variability, and effective visualization. In this regard, the issues of how to deal with big data and how to investigate potential information are significant and challenging in CEM. Big data analysis can meet the requirements for dynamic measurement and analysis in engineering projects. It can contribute to the development of CEM in the following ways: ① Big data analysis can improve the effectiveness and quality of engineering project management by helping managers to adopt diversified approaches to cope with complicated issues in CEM; and ② it can comprehensively assist in the risk control of projects.
Big data technology can be utilized to explore the hidden useful information in an overall project, thereby assisting managers in identifying potential risks. In addition, big data technology has been successful in dealing with data in the form of numerical values, images, and videos. For example, big data technologies combined with digital twins have been considered in the development of smart strategies for sustainable manufacturing, for which they can act as a critical tool for data acquisition, estimation, and mining in uncertain and complex environments
[40]. Big data technology has been adopted to explore transportation planning and retail optimization issues within a supply chain network under various carbon policies, by using uncertain theory to manage data uncertainty and characterize variables
[41]. In this way, hidden useful information and relationships from raw data can be explored and extracted via big data technology. Big data technology enhances decision-making and project efficiency by analyzing enormous datasets for insights into project performance, resource allocation, and risk management, resulting in improved project planning, cost savings, and better predictive maintenance strategies.
4.2.2. Computer vision
Computer vision empowers machines to process, analyze, and understand visual data from images and videos, allowing them to extract meaningful information and make informed decisions
[42]. It involves techniques for acquiring, processing, analyzing, and understanding visual information to extract meaningful insights. This field encompasses tasks such as image classification, object detection, facial recognition, and image segmentation
[43],
[44],
[45],
[46],
[47],
[48]. Machine learning is integral to computer vision, as it provides the models that enable these visual tasks. Machine learning techniques have revolutionized computer vision by significantly improving the accuracy and efficiency of visual data analysis. Machine learning models learn patterns and features from large datasets, which allows the models to generalize and perform complex visual tasks. The synergy between computer vision and machine learning enables machines to perceive and understand the visual world with increasing precision and capability. As large quantities of visual data are generated in practice, several tasks related to CEM that involve visual measurement with extensive human participation can be automatically implemented through computer vision instead. For example, motion analysis techniques were developed and used along with data collected by advanced depth sensors to detect unsafe behavior among construction workers (i.e., safety management)
[49]. Defects related to cracks and leakage in a metro tunnel were detected using a fully connected network (i.e., defects detection)
[50],
[51]. Point cloud technology was utilized to evaluate construction quality (the over-excavation or under-excavation situation) of an underground tunnel (i.e., quality inspections)
[31]. These CEM-related works show how computer vision can benefit workers by enabling effective inspections of engineering structures and ensuring safe engineering construction, thereby improving the efficiency of construction activities.
In the field of computer vision, convolution neural networks (CNNs) have wide applications. CNNs were expressly crafted for the purpose of handling data structured in multiple arrays
[52]. Such data exhibits a variety of array dimensions: one-dimensional (1D) arrays for sequences and signals, which are common in linguistic data; two-dimensional (2D) arrays for representations such as audio spectrograms or images; and three-dimensional (3D) arrays for more complex data, such as volumetric images or videos. There are four critical points in CNN: local connections, weights’ sharing, pooling, and the utilization of numerous layers
[52]. These grant CNN more capacity to process images holistically. Each layer in a CNN can extract the features from the pixels of images. However, CNN-related machine learning approaches are utilized to detect objects or classify images based on the features of objects, which consumes a great deal of computational resources.
Table 1 [45],
[46],
[47],
[48],
[49],
[53],
[54],
[55],
[56],
[57],
[58],
[59],
[60],
[61],
[62],
[63],
[64],
[65],
[66],
[67],
[68],
[69],
[70],
[71],
[72],
[73],
[74],
[75],
[76],
[77],
[78],
[79] lists critical computer vision techniques and their applications in construction activities. To summarize, computer vision improves project efficiency and safety by enabling the automated monitoring and analysis of construction sites through visual data, which enhances progress tracking, quality control, and hazard detection, leading to more effective management and reduced risks.
4.2.3. Speech recognition
Speech recognition is a technology that enables machines to convert spoken language into text by processing and interpreting audio signals
[80]. It involves several steps, including capturing audio, pre-processing to filter noise, feature extraction to identify the relevant characteristics of the speech signal, and decoding to translate these features into text. Machine learning is essential in speech recognition, as it provides the necessary algorithms and models that drive these processes. Techniques such as long short-term memory network are effective for handling the sequential nature of audio data, allowing models to learn temporal patterns and dependencies within speech. The integration of machine learning in speech recognition has led to significant advancements, making it possible to develop highly accurate and efficient systems that can be used in applications. Automatic speech recognition (ASR) is a subset of speech recognition that involves machines autonomously recognizing and transcribing spoken language without human intervention
[81]. Hidden patterns in identity, emotion, language, and textual transcription toward the spoken utterance of speakers in speech can be detected and identified. Traditional ASR has three main parts—namely, a language model, a pronunciation dictionary, and an acoustic model—while modern ASR is fully end-to-end
[81].
Retrieval systems based on voice information are frequently used and applied in search engines with ASR-based virtual assistants, such as the Google virtual assistant. These search engines rely not only on brief commands from a human voice but also on an understanding of natural language from a human.
ASR is typically applied in CEM through integration with building information modeling (BIM) and permits convenient communication between the BIM and end-users. In BIM-based engineering projects, semantic and geometric data for the components of the buildings can be provided by the BIM model to all system users. The integration of ASR and BIM in CEM can be helpful in decision-making and in detecting defects in the infrastructure system. ASR enables interoperability in related devices and provides effective information-retrieval solutions based on users’ voices. For example, interaction is performed using mobile devices to access the BIM model in a designated room, which allows users to automatically deal with information.
In the literature on the integration of ASR with BIM, spoken dialogue systems have been adopted to compile building operation knowledge and have been used to capture data related to the operation of building systems
[82]. A natural voice-driven navigation system was proposed to enable a collaborative response to fire emergencies by providing pathfinding assistance location-sharing and indoor positions for participants in a dynamic way
[83]. A sound measurement system with a framework described as a data collection-extracting feature-training classified-measurement scheme was developed to prevent damage to underground pipelines caused by construction activities
[84]. ASR makes it possible to integrate concepts into BIM in order to minimize human interference and reduce the errors in the overall project. Platforms that integrate ASR and BIM hold great potential for BIM-based engineering projects. In summary, speech recognition enhances communication and documentation efficiency by enabling hands-free data entry and real-time reporting, thus improving safety and productivity on construction sites.
4.2.4. Natural language processing
Natural language processing is intended to equip machines with the capability to comprehend, interpret, and generate human language through the in-depth analysis of textual data
[85]. Usually, there are two kinds of data: structured and unstructured data. Structured data is mainly obtained from measured devices or sensors and is subsequently stored in tables, whereas unstructured data mainly comprises images, videos, or text. Thus, different methods are needed to deal with structured and unstructured data. In practice, considerably more unstructured data is produced than structured data. Natural language processing encompasses translating languages, analyzing sentiment, summarizing texts, and answering questions. Machine learning plays a fundamental role in natural language processing by providing models that enhance natural language processing capabilities. These models are trained on enormous datasets of text to learn the complex structures and nuances of human language. By leveraging techniques such as word embeddings and attention mechanisms, machine learning models can understand context, disambiguate meanings, and generate coherent and contextually appropriate responses. By continuously learning from new text data, these models can adapt and improve over time, handling a wide range of linguistic variations and nuances.
The main tasks performed by natural language processing are natural language understanding (NLU) and natural language generation (NLG). Unstructured text is displayed or encoded through NLU, which allows machines to comprehend human language. Structured data is decoded into text to enable machines to produce natural language that can be understood by humans.
Natural language processing can be applied in CEM in three main scenarios: ① automatic management of the knowledge and information related to a project; ② the analysis of reports and accidents in order to increase the safety of engineering construction and improve risk estimation and management; and ③ intelligent compliance checking. In the first scenario, project management includes data, knowledge, and information (DKI) extraction and retrieval for specific purposes. Knowledge extraction is performed to analyze the documents from engineering projects, which helps in exploring valuable information in order to facilitate project management. DKI retrieval is realized via similarity comparisons between two represented items.
For example, a natural language processing-based approach was developed to learn and explore the knowledge of project scheduling from collected records; it was subsequently utilized to check the logical dependencies and schedule construction activity based on this learned knowledge
[86]. A question-answering system is another form of DKI retrieval in which answers are generated from relevant texts or documents that are retrieved
[87]. For example, an effective end-to-end method integrated with natural language processing techniques was developed to accurately and quickly provide answers to questions from users based on building regulations
[88].
Natural language processing is also useful in accident-risk analyses and accident management, as critical information can be extracted from numerous noisy documents to determine the causes of accidents. Automatic compliance checking can also be achieved through straightforward comparisons. For example, semantic natural language processing techniques integrated with data-based methods were adopted to transform and extract design and regulatory information in order to automatically achieve compliance inference
[89]. Safety information was also checked or localized in a project plan by means of natural language processing
[90]. A four-part framework for automated rule checking was developed based on natural language processing
[91]. Rule interpretation in texts and the accurate alignments of concepts are required in automated rule checking
[91]. To summarize, natural language processing enhances project management by enabling the automated extraction and analysis of information from large volumes of text data, such as contracts, reports, and communication logs, leading to improved decision-making, efficient document management, and better risk-assessment and mitigation strategies.
4.2.5. Machine learning
Machine learning, which allows computers to learn from a given number of datasets
[92], is mainly categorized into three groups: ① supervised learning; ② unsupervised learning; and ③ reinforcement learning, as displayed in
Fig. 6. Machine learning serves as a backbone for various techniques, driving advancements in computer vision, speech recognition, natural language processing, and KRUR, by leveraging big data technology to improve the accuracy, efficiency, and robustness of these applications.
In supervised learning, datasets labeled with inputs and targets or outputs are offered to a system, which attempts to establish correlation relationships between the inputs and estimated objects, as shown in
Fig. 6(a). The primary goal is to deduce and approximate the values associated with one or more target variables, predicated upon specific sets of input variables. This paradigm hinges on the model’s capacity to discern patterns and relationships within labeled datasets, and thereby effectively facilitate the prediction or classification of unseen or future data points based on the acquired knowledge from the training dataset. The values of estimated variables may be continuous (for regression issues) or discrete (for classification issues).
In unsupervised learning, relevant machine learning algorithms are utilized to perform inference based on input variables without labeled responses in datasets, as shown in
Fig. 6(b). The main object is to investigate any possible inherent patterns or relationships in the datasets.
Table 2 [22],
[93],
[94],
[95],
[96],
[97],
[98],
[99],
[100],
[101],
[102],
[103],
[104],
[105],
[106],
[107],
[108],
[109],
[110],
[111],
[112],
[113],
[114],
[115],
[116],
[117],
[118],
[119],
[120],
[121],
[122],
[123],
[124] lists some studies on supervised and unsupervised learning in CEM.
Reinforcement learning was specifically proposed to deal with decision-making and control issues in sequence
[125]. As illustrated in
Fig. 6(c) for reinforcement learning, the agent interacts with the environment by taking an action
At based on the current state
St. The environment responds by providing a reward
Rt and transitioning to a new state
St+1. In such problems, agents iteratively interact with a fuzzy dynamical system environment, learning to maximize cumulative rewards over time. The main steps for reinforcement learning are as follows: ① The agent performs actions according to the experience learned through dynamic interactions with the environment; ② the state of the environment is observed; and ③ rewards are received according to the state and action. The objective of the agent in reinforcement learning is to propose the optimal path for a specific issue by maximizing the accumulated rewards and balancing exploitation.
In CEM, planning work for an engineering project at a construction site involves various types of information from diverse sources, so it is regarded as a constrained optimization issue. Here, reinforcement learning provides an alternative solution. For example, reinforcement learning integrated with a linked-data constraint was utilized to generate a schedule that tackled problems related to manual planning and assisted in efficiently arranging construction activities
[126]. In relation to design activity, a multi-agent reinforcement learning system integrated with BIM was developed to automatically deal with rebar designs for reinforced concrete joints without clashes
[119]. Moreover, a
Q-learning method was adopted to extend previous work in order to achieve a realistic path design. A reinforcement learning-based framework was developed to optimize the strategy for conventional tunneling in engineering construction. In a virtual environment, a deep
Q-network was used as an agent’s architecture, in which the objective for the tunnel breakthrough was economically and safely achieved by receiving feedback from the designed reward system
[127]. Examples related to the application of reinforcement learning in CEM are displayed in
Table 2.
Deep learning (DL) allows computers to learn from datasets and previous experience
[128]. A DL model is composed of numerous hidden layers, which are adopted for transformation, feature extraction, and pattern analysis. In comparison with time-consuming missions that require manual effort, the automation and reliability of DL to perform different kinds of tasks is revolutionizing the construction industry. Training tasks with large-volume datasets benefit from DL techniques. However, training datasets in the construction industry are limited, which constrains the utilization of DL techniques under some circumstances, as DL techniques have rigorous requirements regarding the number and quality of datasets. The skills and knowledge for learning tasks can be implemented via transfer learning, which overcomes the absence of data or limited data from engineering projects.
4.2.6. Knowledge representation, understanding, and reasoning
KRUR utilizes domain knowledge in symbolic representation and predefined manipulations to construct knowledge-based systems. This enables computers to understand the available knowledge, manage the stored knowledge via reasoning, and present valid conclusions. Machine learning augments KRUR by providing sophisticated models and algorithms that facilitate data-driven learning, pattern recognition, and the continual enhancement of comprehension and reasoning abilities. Machine learning models can continuously learn and refine their knowledge base by integrating new data, which improves their ability to make accurate predictions and decisions. Fault tree analysis (FTA), Bayesian network (BN), fuzzy cognitive map (FCM), and graph neural networks (GNNs) are the most frequently used methods for KRUR.
FTA is a deductive technique that simulates the dependability of a system based on logical graphical correlations between faults and their causes
[129]. The analysis results reveal how a combination of components can lead to the failure of the system under certain conditions in a qualitative or quantitative way. FTA has been widely utilized for risk analysis to improve safety in engineering construction. For example, an FTA-based probabilistic decision-making method was proposed for the safety risk evaluation of a metro project under complex construction conditions
[130]. Since FTA and BN have similar inference mechanisms and network structures, they are usually used together in probabilistic analysis and risk failure assessment, such as in tunnel collapse analysis
[131] and navigation accidents analysis
[132].
BN is also an effective graphical model for simulating the causal correlation relationships between variables and a research object
[133]. Typically, a BN is constructed based on the knowledge and experiences of experts or massive historical data. A BN can be viewed as a potential tool for performing inference, which can be done in two ways: ① The occurrence probability of a risk event under the combined effects of influential factors is obtained through forward propagation inference; and ② the posterior probability distribution of influential factors can be calculated via back propagation under the specific conditions for a risk event. In this way, construction events that are fraught with high-risk factors can be systematically discerned, allowing for the estimation of their associated failure probabilities. This critical endeavor furnishes project managers with invaluable insights, empowering them to take timely and efficacious countermeasures. To increase the capacity of BN, a multi-status BN was developed
[32]. For example, a BN-based decision-support approach was developed to dynamically realize safety control for tunnel construction in complicated environments
[134]. The approach provided a decision tool allowing project managers to control the overall safety risk during the construction procedures of the project (i.e., pre-accident, construction, and post-accident control).
FCM is an effective tool in complex system modeling that is based on the attributes of fuzzy logic and neural networks
[135]. FCM was developed based on the concept of a cognitive map. The main aim of FCM is to identify the structure of a model by constructing a causal diagram, which displays the studied system and the causal correlation relationships among the interacting components at the macro level
[135]. The causal and functional interactions of the components in the system can be captured based on the decision-makers’ knowledge. The graph structure of FCM makes it possible to capture system dynamics and causal inference, allowing the analysis of complex component interactions through dynamic reasoning to identify key drivers within the studied system. This characteristic of FCM makes it a useful decision-making tool in CEM. For example, an FCM-based approach that enabled a root cause analysis was proposed to evaluate the construction performance of a tunnel-boring machine for a tunnel
[136]. In construction, FCM has been utilized to analyze change orders and their causes in engineering projects, by considering the causal interactions among the causes of the changing orders
[137]. FCM integrated with a structural equation model (SEM) was developed to analyze the risk of a public–private partnership project; the SEM was applied to study the causal correlation relationships for risk factors based on collected data from the project
[138]. In engineering and project management, FCM is mainly utilized in risk analysis, as illustrated in
Table 3 [52],
[130],
[131],
[134],
[136],
[137],
[138],
[139],
[140],
[141],
[142],
[143],
[144],
[145],
[146].
A GNN is a method used to explore structural correlation relationships among objects, based on given data
[147]. GNNs can be classified into two categories: spatial-based GNNs and spectral-based GNNs. In a spatial-based GNN, such as a graph attention network, previous neighbors and nodes are considered in order to generate embedding
[148], which has been utilized to transform attributes into lower dimensions and continuous vectors
[149]. Spectral-based GNNs, which operate in the frequency domain of graph data, include graph convolutional networks
[150]. Spectral graph theory, which studies the properties of a graph in relation to the eigenvalues and eigenvectors of its Laplacian matrix, has been applied for efficient graph signal processing and management. Data related to model topologies, geometries, and the values of properties should be allowed to be exchanged between related software and platforms within a BIM environment. In this context, a graph presentation is considered a viable solution to address interoperability issues, as project information can be effectively stored using a graph data structure. Moreover, the information can be effectively extracted for other applications via retrieval. All of these provide an opportunity to explore the advantages of GNN in practice.
Table 3 presents the latest applications of GNN-based approaches for KRUR in CEM. In brief, KRUR enhances decision-making and problem-solving by enabling intelligent systems to organize, interpret, and make decisions in relation to complex project information and processes.
4.3. Typical applications of smart techniques in CEM
The typical applications of smart techniques in USE are exemplified across the planning and design stage, the construction stage, and the operation and maintenance stage.
4.3.1. Planning and design stage
In USE, the ideal construction technology not only has little impact on the surrounding environment but also reduces the construction cost. As shown in
Fig. 7, these aims can be achieved using MCDM techniques
[151], following three steps: ① comprehensive decision hierarchy construction; ② data collection and processing; and ③ selecting the optimal construction technology under a fuzzy and uncertain environment. In the first step, the research object, criteria, and sub-criteria are determined according to the construction situation of the engineering project. In addition, critical aspects such as the geological conditions and construction requirements are considered as criteria, denoted as
C1,
C2, …,
C5. Each criterion also consists of various sub-criteria. For instance, within criterion
C1, there are three sub-criteria designated as
Q11,
Q12, and
Q13. Next, the determined criteria and the corresponding sub-criteria are collectively employed to construct a decision hierarchy. First, alternative construction technologies denoted as (
A1,
A2,...,
A6) are tentatively identified, based on the attributes of the engineering project. Then, the decision hierarchy—in conjunction with the array of available alternatives—is taken as the foundation for constructing a comprehensive decision hierarchy. In the second step, experts are requested to assess this comprehensive decision hierarchy. In this step, the experts adopt different kinds of fuzzy sets with various scales to judge the evaluated items. It is difficult to determine the exact values of some items in a comprehensive decision hierarchy, such as the availability of working space. Thus, simulated results can be used as a valuable reference for project managers’ decision-making. For example, based on collected data from experts, the optimal construction technology for a metro line was determined via MCDM modeling in step 3.
The design process for an underground structure is illustrated in
Fig. 8. A traditional structural analysis is generally performed via analytical approaches developed from the codes of numerical simulations (e.g., finite-element analysis)
[152]. However, intelligent techniques can aid in the design process for an underground structure once the building materials are determined. There are three steps in the application of intelligent techniques within this context: ① dataset preparations and training (performed by an interpreter); ② performance evaluations of the developed intelligent model (performed by a designer); and ③ application of the developed model (performed by a modeler). In the first step, previous designs (e.g., engineering design drawings for excavation engineering) are learned by a developed intelligent model, associating the structural design with the training datasets. Relevant designs that satisfy the specifications are evaluated by designers and engineers. After that, a semantic process is performed by extracting the structural and architectural elements from the design drawings and coding them using color patterns. In this way, the significant design elements and corresponding information of the structural layout are retained. Training performance can be greatly enhanced by the incorporation of semantic designs. At this point, a performance evaluation of the intelligent model is conducted by experienced engineers. Finally, the developed model is used to design a preliminary scheme of the underground structure by entering the semantic engineering design drawing into the developed model.
4.3.2. Construction stage
Fig. 9 Ref.
[153] illustrates the application of intelligent techniques in the construction stage of an underground space. The intelligent techniques are utilized to estimate the operational parameters of the earth pressure balance (EPB) shield machine, which helps to control and regulate significant parameters for tunnel construction. As shown in
Fig. 9, there are three steps: ① data preparation and collection; ② deep-neural-network-based intelligent model development; and ③ application of the developed model for tunnel construction. For an underground project, the data mainly consists of operational datasets from the automatic monitoring system in the EPB shield machine and geological datasets from the geological investigation report. Prior to tunnel construction, site investigation is performed using borehole technology. As these boreholes are limited, the geological information is sparse and discontinuous. Under these circumstances, technologies such as Bayesian compressive sampling
[154], Kriging methods
[155], random fields
[156],
[157],
[158], and Markov chain methods
[159],
[160],
[161] are often utilized to predict the geo-data and evaluate the corresponding statistical uncertainty. The geological conditions along the tunnel line can be roughly revealed using the estimated geological data. After the data collection, the data must be pre-processed before being inputted to the deep-neural-network-based model. The datasets obtained from tunnel construction are in a time-series, so they are usually coped with using a long short-term memory neural network
[28],
[106] or a gated recurrent unit deep neural network
[30],
[162]. Simple empirical methods are utilized to guide the tunnel construction, as insufficient data is available in the early construction stage. The volume of the measured datasets increases once the project enters the normal construction stage. As the construction continues, new operational data is continuously generated, enlarging the existing datasets and thereby improving the predictive performance of the deep-neural-network-based model. The performance of the intelligent model can be better trained and improved by the use of sufficient datasets. Meanwhile, the most critical factors are identified via a sensitivity analysis, and the estimated parameters that are in an acceptable range are recommended for tunnel construction, thereby reducing the impact to the surrounding environment.
4.3.3. Operation and maintenance stage
Once the construction activity of the tunnel is finished, the project enters the operation and maintenance stage. The principal objective of the operation and maintenance stage for a metro system is to ensure that the infrastructural assets consistently and reliably meet the requisite minimum standards for normal operational functionality throughout their entire operational lifespan. This enduring commitment to preserving and enhancing the operational integrity of critical transportation infrastructure underscores the significance of effective operation and maintenance practices. Construction quality evaluation and defect inspection, which are significant tasks, are typically conducted with the assistance of intelligent techniques, as illustrated in
Fig. 10. In this case, 3D laser scanning was used to evaluate the construction quality of the tunnel
[31]. The measured points and elevation of the tunnel were recorded in a point cloud. Differences between the measured and designed data were then utilized to generate an unrolled map, which revealed any over or under-excavation situations in the tunnel.
Fig. 10(a) displays the situation, which indicates that the over-excavation of the tunnel is mainly distributed in the inner wall along the central axis. Cavities in the tunnel surface are reflected by segmented areas of under and over-excavation. Voids will be generated during the construction of the backfill layer. All aspects of the construction quality of the tunnel can be detected and verified via 3D laser scanning and ground-penetrating radar.
A traditional visual inspection that is manually performed cannot meet the basic requirements nowadays, as it has a low efficiency in recognizing defects under environmental and operational interferences. Instead, DL techniques are utilized to recognize the semantic segmentation that indicates defects
[50]. This procedure is illustrated in
Figs. 10(b–e). Images related to crack and leakage defects are captured in datasets by automatized image-collection equipment
[163], as shown in the working scene and sketch map displayed in
Figs. 10(b) and
(c). All the collected images must first be processed. A moisture section on the tunnel surface is denoted along the tunnel boundaries with a polygon. Once the drawing of the polygons is complete, a label with dialogue is added, providing an annotation with information (e.g., the width and length of the moisture section).
Fig. 10(e) presents some annotated images. Based on the initial processed images, a deep neural network model is established and applied in engineering practice. In general, there are three steps in developing a model to identify tunnel moisture: feature extraction; the production of a regional proposal; and the identification of the leakage section. The results for the moisture identification in this case are illustrated in
Fig. 10(f); they provide a reference allowing the project manager to address the leakage in the tunnel surface in time.
5. Future directions
The trajectory of technological advancement is yielding an ever-expanding array of cutting-edge techniques stemming from smart technologies. The pervasive integration of such technologies underscores the dynamic evolution of engineering practices, as modern methodologies continue to harness the full potential of smart technologies to enhance project efficiency, sustainability, and performance across the complete project life cycle. As illustrated in
Fig. 11, interactions among environments, participants, and sensors play a significant role in CEM. Based on these interactions, we summarize six significant future directions for CEM-related issues in practice: the metaverse, the Internet of Things (IoT), blockchain, digital twins, BIM, and smart robots. These directions can be efficiently linked to the interactions among the built environment, participants, and sensors of an engineering project.
Finally, we believe that it is possible to achieve the objectives of sustainability, carbon neutrality, and Industry 5.0 in the construction industry. Within CEM, Industry 5.0 is characterized by its core objective of synergizing the innovative capacities of human experts with efficient, intelligent, and precision-driven machinery. Importantly, Industry 5.0 is uniquely tailored to address the specific intricacies and challenges of construction projects, emphasizing increased efficiency and advanced problem-solving capabilities. The specifics regarding the future directions of CEM are summarized in
Fig. 12.
5.1. The metaverse
The metaverse encompasses a diverse set of technologies that converge to create immersive digital experiences. It includes virtual reality (VR) for complete immersion in virtual environments, augmented reality (AR) to blend digital elements with the physical world, 3D modeling and simulation, edge and cloud computing for scalability, the IoT for real-world data integration, and advanced user interfaces for intuitive engagement. Together, these technologies form the foundation of the metaverse, enabling the formation of interconnected virtual spaces where users can work, socialize, and play in unprecedented ways. A metaverse is a virtual space that is shared by multiple users, allowing them to engage with a computer-generated environment in real time
[164],
[165],
[166].
In the construction industry, the metaverse can create immersive digital environments that facilitate collaboration among stakeholders
[165], allowing for the seamless exchange of data and insights for ongoing sustainability improvements. This virtual realm enables stakeholders to visualize and simulate construction processes, allowing for more efficient planning, design, and resource allocation. This reduces the material waste, energy consumption, and carbon emissions associated with traditional construction methods. In addition, remote collaboration platforms in the metaverse can facilitate communication among project teams, reducing the need for physical travel and further minimizing the project’s carbon footprint. Moreover, the metaverse can facilitate enhanced project visualization and coordination, leading to more precise construction. It can also integrate with IoT sensors, collecting data from construction sites, machinery, and even the wearable devices of workers. This data then feeds into analytics platforms powered by machine learning methods, and these algorithms identify patterns and anomalies, allowing construction managers to optimize workflows, reduce energy consumption, and enhance safety. Consequently, metaverse technology can enhance the overall efficiency of the construction process. Insights gained from the metaverse can enable construction companies to align their practices with sustainability goals, effectively minimizing their carbon footprint during the construction stage. Finally, the metaverse can serve as a platform for virtual training and education in sustainable construction practices, ensuring that workers are equipped with the necessary knowledge and skills to implement eco-friendly solutions on-site.
The future of the metaverse in CEM is expected to include the following aspects: ① Interactive design and prototyping will advance, providing engineers and architects with highly detailed virtual spaces for immersive and collaborative design experiences, and thereby reducing the need for revisions during construction; ② metaverse simulations will integrate AI and data analytics, enabling accurate project outcome predictions and real-time data integration for continuous monitoring, optimizing efficiency, and safety; ③ operational potentials will expand as virtual replicas of physical structures enhance user experience and enable precise maintenance and repair tasks; and ④ metaverse platforms will evolve into comprehensive project-management hubs, facilitating global collaboration, real-time tracking, resource allocation, and risk analysis.
5.2. The IoT
The IoT encompasses the interconnection of physical devices and engineering objects equipped with sensors, software, and advanced technologies, enabling them to collect and exchange data through Internet connectivity
[167]. These devices, ranging from sensors and cameras to communication networks, create a seamlessly interconnected world in which data is shared among devices. This integration of physical and digital infrastructure empowers real-time data acquisition for enhanced control and informed decision-making processes. IoT devices are instrumental in gathering data across a wide spectrum of parameters. By leveraging advanced analytics tools such as machine learning algorithms, this data can be subjected to analysis, facilitating the predictive maintenance and optimization of various infrastructure systems. In CEM, the IoT can be applied in complex engineering projects in the following ways: ① IoT sensors can be used to monitor the health of bridges, buildings, and other structures in real time, providing data (e.g., stress, strain, and vibration) that can be used to evaluate potential structural problems early on and to optimize maintenance and repair schedules
[168],
[169]; ② IoT devices can be applied to monitor traffic patterns, weather conditions, and road conditions, which can optimize traffic flow and reduce congestion
[170],
[171]; ③ IoT sensors can be deployed at construction sites to monitor resource usage, energy consumption, and emissions in real time, with the help of machine learning algorithms to analyze the data for improvement in construction processes, where the data from these sensors empowers construction managers to optimize workflows, ensuring minimal waste and emissions while maintaining productivity
[172],
[173],
[174]; and ④ IoT-enabled smart devices can automate processes (e.g., lighting, heating, and ventilation) to optimize energy efficiency and reduce carbon emissions.
The IoT is poised for significant future advancements in the following key areas: ① The role of IoT in data-driven decision-making will intensify, driven by the proliferation of sensors and devices within construction sites, and the surge in data volume will necessitate the evolution of sophisticated data analytics methodologies, incorporating machine learning for actionable insights; ② the integration of IoT with BIM will further enhance construction processes, building performance, and facility management by providing real-time data; ③ the IoT will play a pivotal role in improving safety on construction sites through wearable devices and real-time safety alerts; and ④ the IoT will continue to influence sustainability initiatives in CEM by enhancing energy efficiency, minimizing waste, and encouraging environmentally sustainable construction practices.
5.3. Blockchain
Blockchain technology is a decentralized and distributed digital ledger system meticulously designed for the purpose of documenting transactions across a multitude of computers. Its intrinsic architecture is engineered to imbue the recorded data with a trifecta of paramount attributes: security, transparency, and immutability. These transactions, coalesced into coherent clusters recognized as “blocks,” are interconnected in a sequential and chronological manner, thereby constituting an unbroken chain of interconnected blocks, aptly termed a “blockchain”
[175]. To bolster its integrity and safeguard sensitive information, blockchain technology employs intricate cryptographic mechanisms. Moreover, it eschews any dependence on central authorities or intermediaries by orchestrating its operations within a peer-to-peer network paradigm, ultimately cultivating an environment characterized by decentralization and trust. For CEM in a built environment, blockchain automatically enables convenient collaborative design and code compiling examination with natural language processing in the initial design stage of an engineering project. In this process, collaborative design increases the interactions and trust among contractors, designers, and clients within the collaborative platform (e.g., BIM). With the wider application of BIM in CEM, various issues with BIM collaborative platforms (e.g., cyber security) can be addressed via integration with blockchain technology, whose transparency and immutability permits any information related to the modification of BIM models to be tracked or recorded
[176],
[177]. In this way, legal responsibility and intellectual attributes from BIM models can be determined.
Larger volumes of data are produced in the operation and maintenance stage (e.g., data related to the operational status of underground tunnels and buildings), for which blockchain facilitates data and document management
[178]. Significantly, the use of blockchain in CEM offers several avenues for improving transparency, accountability, and efficiency in sustainable practices. Blockchain’s transparency makes it possible to track the energy consumption of buildings or underground facilities, as well as their emissions throughout their operational life. Data recorded on blockchain provides a comprehensive view of the environmental performance of buildings or underground facilities. In addition, blockchain can enhance supply chain transparency by providing a secure and immutable ledger that tracks the origin, production methods, and transportation of construction materials. This enables stakeholders to verify the sustainability credentials of materials, such as their environmental impact and carbon footprint. Smart contracts powered by blockchain can automate the enforcement of sustainability standards and certifications, ensuring that contractors and suppliers adhere to agreed-upon sustainability criteria. Furthermore, blockchain technology facilitates the sharing of data and insights among project stakeholders, enabling collaborative decision-making and the optimization of sustainability initiatives. Thus, the use of blockchain technology in CEM offers innovative solutions for promoting transparency, accountability, and efficiency in sustainable practices.
Blockchain holds significant promise for future application in the following areas: ① revolutionizing engineering management transparency; ② utilizing smart contracts for automated project management; ③ decentralizing project documentation for improved accessibility and security; ④ streamlining the payment process to address persistent delays and disputes; and ⑤ integrating with IoT and sensors to ensure the integrity of data generated by these devices. Blockchain provides a centralized source of truth for project documentation, expediting payment processes, and ensuring the security and reliability of data from IoT devices and sensors.
5.4. Digital twins
Digital twins are intricate virtual counterparts of tangible entities, including physical objects, complex systems, or intricate processes
[179],
[180]. These digital incarnations are uniquely empowered by the seamless integration of real-time data streams and advanced simulations. Through the harmonious interplay of these components, a comprehensive and dynamic digital model is meticulously constructed. This model faithfully encapsulates not only the physical attributes of the original entity but also its intricate behavioral dynamics. Such digital representations are formidable tools in the realm of technology, enabling a multifaceted array of applications and insights across various domains and industries
[179]. In the context of CEM, digital twins are utilized for modeling and simulating construction projects and building operations. The intrinsic capability of digital twins to support ongoing real-time monitoring and data-driven decision-making empowers engineers and management professionals to make judicious and well-informed decisions and to skillfully deal with constantly changing and complex situations in work environments. A digital twin provides a holistic perspective of the entire life cycle of a construction project, from its design and construction through to its operation and maintenance, and thus plays a crucial role in CEM. The integration of digital twins into engineering management practices holds promise to enhance productivity, minimize downtime, and heighten safety measures. Consequently, digital twins have been widely adopted for optimizing production processes, monitoring equipment health, and simulating intricate construction endeavors.
For example, a digital twin can be used to generate a virtual representation of physical assets, such as a building or infrastructure, by incorporating real-time data from sensors, IoT devices, and various other sources. A digital twin enables designers to run simulations and analyses, making it possible to optimize energy consumption from the earliest design phases. By virtually mirroring the performance of a building or underground structure using a digital twin, any shortcomings in sustainability can be identified and addressed before construction even begins. MCDM also facilitates the evaluation of various design alternatives by considering multiple criteria simultaneously. Finally, digital twins provide engineers with a platform to simulate diverse scenarios and experiment with various solutions for engineering construction, thus contributing to more efficient and effective CEM practices. Conducting predictive analytics with a digital twin makes it possible to identify opportunities to improve sustainability, including enhancing resource efficiency.
Future transformative possibilities for the use of digital twins in CEM are as follows: ① The integration of digital twins with advanced technologies is poised to reduce information dynamics update costs, enabling real-time monitoring and automated decision-making; ② the incorporation of multi-physics modeling with VR technology through digital twins enhances human–machine interaction efficiency and operational feasibility, contributing to reduced costs and improving real-time capabilities; ③ introducing virtual scenario modes in a digital twin allows for the construction of basic infrastructure in virtual spaces by combining AR and VR for intuitive project management; and ④ the broadening influence of digital twins across industries and disciplines signifies a paradigm shift in collaboration and innovation, offering extensive potential for digitizing traditional sectors and fostering interdisciplinary cooperation.
5.5. Building information modeling
BIM is the digital representation of the functional and physical attributes of infrastructure; it serves as a database with information on the infrastructures throughout the project. The attributes of BIM, such as visualization, cooperativity, and optimization, grant it wide application in the construction industry
[181]. BIM is implemented based on a series of object-oriented software comprising parametric objects for the components of infrastructures. In a BIM platform, model information can be updated and analyzed, providing reliable references for decision-making
[181]. In CEM, BIM is viewed as a tool to facilitate the activities of engineering projects, which is realized in the following ways: ① It permits drawings of 2D infrastructures and corresponded 3D geometry models with detailed information (e.g., the properties of materials) to be visualized; ② it allows problems related to CEM to be quantitatively performed (e.g., virtual construction)
[181]; ③ a BIM model with detailed information on a construction project’s components (e.g., non-geometric and geometric information) and the relationships among components is both helpful and convenient for the optimization of engineering projects. In practice, the implementation of BIM models in certain projects is described in terms of functionalities, interoperability, and techniques. BIM has multiple potential applications in CEM, over the entire life cycle of buildings and infrastructures. It is a collaborative work that involves participants working with the architecture, structure, electricity, and fabrication, based on the requirements of the engineering project and stakeholders. In this process, BIM data is used to support and simulate specific requirements from the project via flowcharts with the logical flow of information. The functionalities of a BIM model facilitate information storage, conversion, and exchange, increasing the interoperability of a manual information-delivery framework. BIM models for new and existing buildings or infrastructures differ due to the varying qualities of project information, data availability, and specific requirements. In this regard, techniques related to data capture, processing, and object recognition are significant for BIM model creation, since such techniques affect the quality of data. Techniques for data capture mainly consist of non-contact methods (i.e., image-based methods or various methods combined) and contact methods (e.g., manual techniques)
[51],
[53]. Data processing is then executed, which enables the identification of functionality-related BIM objects based on previously captured data. Object recognition includes the identification of objects and the extraction of semantic information
[43],
[44].
The role of BIM can play in achieving carbon neutrality and sustainability relates to its ability to create a virtual representation of a construction project. During the engineering construction stage, BIM allows for resource allocation and scheduling. By simulating various scenarios and optimizing designs, it minimizes waste and energy consumption. In addition, BIM aids in efficient procurement by ensuring the selection of sustainable materials and processes. Real-time monitoring through BIM during construction helps to promptly identify inefficiencies, allowing for timely adjustments that can significantly reduce carbon emissions. The utility of BIM even extends into the operational phase, as it provides comprehensive data for the ongoing management and maintenance of structures, thereby optimizing their long-term sustainability. Furthermore, BIM facilitates the simulation of building performance, enabling stakeholders to assess energy usage, daylighting, thermal comfort, and indoor air quality, and to optimize a building’s operations for sustainability.
The following future trajectories are expected for BIM in CEM: ① Dynamic data-driven transformation through BIM platforms is set to empower professionals with critical functionalities for detailed analysis, intelligent links to measured data, and seamless integration with related systems, optimizing resource allocation and project-management efficiency; ② interoperability and data standardization are pivotal for streamlined collaboration, fostering innovation and competition among software providers, leading to advanced solutions for construction projects; ③ the integration of cloud-based BIM systems is poised to revolutionize project management and collaboration, offering real-time access, enhanced communication, and data-driven decision support across all project phases; ④ the evolution of BIM into four- and five-dimensional (4D and 5D) modeling allows a comprehensive understanding of project schedules and budgets, improving project management and reducing risks; and ⑤ mobile BIM, leveraging smartphones and tablets, will play a crucial role in enhancing onsite collaboration and project monitoring, and in overlaying digital BIM data onto physical structures through AR applications, further advancing efficiency and communication in the construction industry.
5.6. Smart robots
Traditionally known for its labor-intensive processes, the construction industry is undergoing a significant transformation, driven by technological advancements. One of the most remarkable developments in this transformation is the increasing integration of smart robots into CEM practices. The application of smart robots in CEM presents opportunities to improve efficiency and enhance safety in construction processes while minimizing environmental impact. Smart robots, equipped with advanced sensing, automation capabilities, and intelligent technologies
[182], are revolutionizing various aspects of construction, from design and planning to execution and maintenance. These technologies enable robots to sense, perceive, reason, and learn from the environment and through interactions with humans, allowing them to perform complex tasks and autonomously adapt to changing situations. Their ability to perform a wide range of tasks, collect data, and work in hazardous environments makes them invaluable assets in CEM. For example, smart robots can help minimize resource consumption and improve overall project efficiency by optimizing construction processes. In addition, smart robots can be programmed to operate in energy-efficient modes, thereby reducing energy consumption and the carbon emissions associated with construction activities. Moreover, by automating repetitive and hazardous tasks, smart robots can enhance worker safety and reduce the risk of accidents, thereby contributing to a safer and healthier work environment. All in all, the application of smart robots in CEM offers a promising approach to achieve more environmentally friendly and efficient construction practices.
Future directions for the application of smart robots in CEM include the following: ① Enhanced automation and autonomous construction processes, facilitated by smart robots equipped with advanced sensing technologies, are set to significantly reduce labor requirements and construction timelines; ② swarm robotics, in which multiple robots are coordinated to perform collaborative tasks such as bricklaying and concrete pouring, will optimize resource allocation and enhance project efficiency; ③ the presence of collaborative robots working alongside construction workers, assisted by intuitive interfaces, will streamline physically demanding and dangerous tasks; ④ advanced 3D printing technology will enhance construction by enabling faster, more sustainable building practices, and robots equipped with high-resolution cameras will conduct inspections and maintenance tasks to ensure structural integrity and adherence to safety standards; and ⑤ smart robots will contribute to sustainable construction practices by optimizing material usage and participating in deconstruction and recycling processes.
In summary, the fusion of smart techniques and advanced technologies with CEM is a significant step toward achieving carbon neutrality. When integrated synergistically across an engineering project, these technologies optimize resource utilization, reduce emissions, and enhance energy efficiency. The construction industry is thus poised to align itself with global sustainability goals, embracing innovation and technological progress as the cornerstones of a carbon-neutral future and Industry 5.0.
6. Conclusions
This study presented a comprehensive review of the applications of smart techniques, including MCDM and intelligent techniques, in CEM. MCDM techniques have been frequently used to solve construction problems with fuzziness and multiple uncertainties. For certain engineering projects that heavily rely on construction experience or expert judgment, especially engineering projects involving multiple indices in complicated construction environments, the incorporation of MCDM techniques with fuzzy sets in different scales is an effective solution. The availability of a large volume of data from different sources and the increasing size of engineering projects provide an opportunity to explore the potential of applying intelligent techniques in CEM. Intelligent techniques offer many benefits, such as high efficiency, high reliability, and automation.
This review summarized essential attributes of CEM, including its dynamics, uniqueness, complexity, uncertainty, fuzziness, incompleteness, and system. Moreover, this review comprehensively explored and discussed the state-of-the-art smart techniques applied in CEM from 2000 to 2022. It also illustrated the applications of smart techniques via several construction scenarios based on the life cycle of engineering projects. More importantly, this review outlined future research directions that aim to sustainably realize smart construction and to achieve the goal of carbon neutrality for the CEM.
Acknowledgments
This research work was funded by the project of Guangdong Provincial Basic and Applied Basic Research Fund Committee (2022A1515240073) and the Pearl River Talent Recruitment Program (2019CX01G338), Guangdong Province.
Compliance with ethics guidelines
Song-Shun Lin, Shui-Long Shen, Annan Zhou, and Xiang-Sheng Chen declare that they have no conflict of interest or financial conflicts to disclose.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2024.08.023.
PDF
(6231KB)
