Literature Review on the Structural Health Monitoring (SHM) of Sustainable Civil Infrastructure: An Analysis of Influencing Factors in the Implementation

Abstract

Structural health monitoring (SHM) of civil infrastructure is significant for sustainable development. This review aims to identify the factors influencing sustainable civil infrastructure SHM implementation and analyze the properties, paths, and conditions under which they have an impact. The method adopted is a coding study based on Grounded Theory. First, the SHM implementation process in the literature is categorized through open coding to obtain an implementation framework that contains phase and activity levels. Second, based on this framework, a synthesis is conducted to categorize the influencing factors in dimensions of contents and properties through open coding and axial coding. Finally, selective coding is used to extract the factors that interacted across activities to propose a scheme of influencing factor relationships. The main findings of the synthesis are shown as follows: (1) sensor work scheduling and data transmission are promising endeavors to balance economic and environmental sustainability, while social sustainability is mainly in terms of safety and user experience; (2) the success of sustainable civil infrastructure SHM requires a collaborative technical and organizational effort; (3) since the influencing factors at different phases may interact with each other, the implementation process should emphasize forward-looking and holistic thinking.

1. Introduction

Sustainable development is a mode that satisfies the present generation without compromising the needs of future generations. Sustainability involves three dimensions, namely, environmental, economic, and social dimensions [1]. Sustainability considerations are becoming an increasingly important role in economic development. However, as one of the main sectors driving economic development, the construction industry uses many resources and emits numerous pollutants and greenhouse gases while creating value, showing a great potential to contribute to sustainability [2]. Specifically, as construction projects that are generally large in scale and closely related to people’s well-being, civil infrastructures are particularly in need of sustainability considerations. Furthermore, structural health monitoring (SHM) is defined as the in situ, non-destructive measurement of the operating and loading environments and the critical response of a structure, from which damage-sensitive features are extracted and statistically analyzed [3], then to detect the presence, location, and severity of structural damage, and further to determine the current health condition, evaluate the remaining useful life (RUL), and provide guide to engineers and inspectors in making informed decisions on issues related to the maintenance, rehabilitation, or replacement of infrastructure [4]. In recent years, SHM has gained more attention from academic research to concrete practice. The implementation of SHM in sustainable civil infrastructure (for this review, according to ISO 14963:2003 [5], including road, rail, pedestrian bridges, viaducts, civil infrastructure buildings, and other types of works provided that their particular structure justifies its application) not only reflects the environmental, economic, and social objectives in the process, but also enables civil infrastructure projects to contribute to the cause of sustainable development through long-term reliable operations [6]. Because the implementation of sustainable development theory requires improving the minimum service life of existing buildings rather than constructing new ones [7,8]. For example, after a critical incident, the functioning of the transportation network is highly sensitive to potential damage to bridges and buildings, and a comprehensive evaluation of the consequences to structures and infrastructure can quantify and plan the most appropriate mitigation measures [9]. SHM may be an important part of assisting in the implementation of the evaluation.

Specifically, SHM improves the resilience of infrastructure, which in turn extends the life of the structures [10]. According to MCEER researchers [11,12,13], infrastructure resilience encompasses four dimensions: robustness of the analysis unit to withstand the effects of a given level of degradation events without suffering damages; redundancy that allows local failures through alternative paths and keeps the structure stable and safe after the failure of any individual component; rapidity representing intervention efficiency and recovery time; and resourcefulness representing the ability to identify, prioritize, and organize resources when a threat or degrading event exists. The framework has been used as a common tool and indicator for assessing measures to improve infrastructure resilience, such as consolidation systems [14]. Meanwhile, the degree of damage detection in SHM is categorized into four progressive levels: presence, localization, quantification, and prognosis of damage [15]. The four levels of SHM and the four dimensions of structural resilience accomplish each other’s contribution to sustainability. Mostly, SHM plays the role of identifying and recognizing problems, while structural resilience provides the ability to take steps quickly and effectively. When the information provided by the SHM is only concerning the existence of damage, this information is required to be correlated with the redundancy level of the structure to support appropriate countermeasures, such as evacuation and closure of critical facilities, among other emergency measures. The second level of damage detection relates to the rapidity dimension of resilience, when damage is localized, the SHM system can be used to facilitate decision making thus enabling intervention and recovery steps to be carried out immediately in emergency situations. The third level of damage detection identifies the degree and even the type of damage that can support the resourcefulness dimension of resilience in assessing and prioritizing the resources needed to appropriately assist in the organization of structural recovery. Prognosis is the fourth level of SHM and measures the ability of a structure to withstand a certain level of damage while maintaining a typical level of function by predicting its performance, i.e., the robustness dimension of structural resilience [10,16]. In summary, SHM provides a deeper identification of factors related to damage and its evolution than simple periodic inspections, thus addressing increasingly complex knowledge and is a major contributor to improving the resilience of structures and infrastructures [10]. Furthermore, SHM system can limit the rate of natural degradation of infrastructure structures, for example, by minimizing the impact of sudden shocks such as earthquakes by reducing the peak of response [10].

It cannot be denied that SHM techniques may be essential if the goals of sustainable development are to be fully realized in civil infrastructure [17]. On the one hand, SHM removes uncertainty in condition assessment to some extent by providing a more accurate illustration of the structural state and allows pushing the useful life closer to its limits without compromising structural safety [18]. On the other hand, the total unit cost of minor periodic maintenance is significantly lower than that of a one-time major maintenance scheme. This illustrates how long-term periodic performance evaluation and maintenance can generate sustainability value and result in significant savings for society [17]. From the perspective of SHM development, promotions and advances have often focused on its subsets [19]. For example, Lynch, Farrar, and Michaels [19] pointed out that SHM can usually be divided into three main phases: damage detection/characterization, prognosis, and risk assessment. However, most current studies in the field always focus on the first phase. Machine learning tools, especially deep learning algorithms, have provided innovative developments in the field and have become increasingly practical, particularly for the extensive use of vibration-based structural damage diagnosis [10,16,20]. For example, Rosso et al. [21] utilize the information contained in the raw vibration data in conjunction with the subspace-based damage indicators and use a machine learning artificial neural network model to perform the first level of the damage detection task. Melchiorre et al. [22] use two different deep learning models: a convolutional neural network called Faster R-CNN and a convolutional recurrent neural network (CRNN), based on acoustic emission techniques to automatically detect the onset of the acoustic signal for damage localization. Both networks achieve the desired goal, with CRNN being more advantageous due to its ability to recognize damage patterns at the same time.

Indeed, in addition to technological advances, the successful implementation of SHM relies more on enabling the industry and the workforce to manage and conduct SHM processes in an orderly and accountable manner [23]. Therefore, a holistic approach is now needed to optimize the implementation [24] and to fully consider the influencing factors in the process. The studies on the influencing factors and stakeholders of sustainable building have achieved rich and insightful results, but mostly focusing on the project level. For example, Hwang, Zhao, and Tan [25] identified the key factors affecting the schedule performance of new and retrofit green building projects through questionnaires and face-to-face interviews. Li et al. [26] conducted a comprehensive review of the critical success factors of green building projects. Darko, Zhang, and Chan [27] examined stakeholder drivers during green building project practices and proposed a framework for categorizing the drivers. Hwang and Tan [28] investigated common barriers to green building project management and proposed solutions accordingly. More deeply, some research results not only focus on the identification and importance ranking of factors, but also consider the correlation between the influencing factors. For example, Yang and Zou [29] stated that most of the risks of green building projects are interrelated, and social network analysis (SNA) is used to assess and analyze these associations. In the research of Huang et al. [30], influencing factors in green building are identified based on the project life cycle, and then their interactions as well as the network importance are investigated through SNA.

However, for SHM which is at a more specific level, little attention has been paid to the influencing factors for its successful implementation. Thus, it is necessary and urgent to propose a scientific and useful solution for this situation. SHM is a relatively new paradigm for civil construction project stakeholders [31]. A common view conveyed by the literature on SHM management practices is that the technique, as a new safety and management tool and an ideal complement to traditional methods such as visual inspections, requires a rigorous approach to design and implementation as well as thorough and careful consideration during implementation to fully realize its benefits [32]. In addition, the movement of SHM from academic research/education to successful practice requires that all stakeholder groups reach a consensus on multiple aspects of definitions, terminology, standards, expectations, etc., and contribute to the combination of SHM’s multidisciplinary skills and the enrichment of the experience base during practice [23,32,33]. To fill the abovementioned gap, this review aims to (a) establish a more complete framework for SHM implementation through a systematic literature review, (b) comprehensively and integrally review the influencing factors in SHM implementation in the existing literature based on the framework, and (c) propose a scheme showing the linkage of influencing factors between different activities in SHM implementation. The significance of this review can be summarized as follows: (1) filling the blank in the management aspect of the SHM research field; and (2) contributing to the implementation of SHM from academic research to practical application, so that SHM can produce a better performance in sustainable civil infrastructure projects, thus promoting the development of the cause of sustainable construction. This review will present the rationality in Section 2, methodology in Section 3, analyze the findings in Section 4, Section 5 and Section 6, discuss the findings in Section 7, and conclude in Section 8.

2. Rationality

The separation between the design and performance of the technical aspects of SHM and the organizational systems is considered to be probably the main obstacle to the effectiveness of most infrastructure [23]. Integral asset management, which is the main purpose of SHM, is based on the design and development of an integrated system, which is also a prerequisite for effective monitoring [23]. Previous studies have illustrated, at the information technology level, the necessity of obtaining complex data from integrated operational and monitoring equipment and have allowed the conceptualization of the attributes and characteristics of the data, including spatial, temporal, frequency, and modal characteristics of the data [23]. Most of the published research is also dedicated to the front and lower-level SHM categories or subsets such as sensor placement, abnormality detection, and model validation, as well as working on specific parts of the SHM problem, such as the design of certain sensors [34]. However, civil infrastructures are expected to meet safe and sustainable operational requirements during long-term service [35], and whether deficiencies in structural safety can be mitigated through health monitoring alone is a question that must be carefully evaluated [23]. Given that assessing the health status of structures through tests and measurements is a common practice, there have long been associated guidelines [4]. The goal of SHM is consistent with this approach, although it takes advantage of new technologies in sensing, instrumentation, communication, and modeling and integrates them into an intelligent system to present a newer paradigm [4]. The SHM field has provided a solid foundation for continuous online monitoring and has advanced the science of sensing and data processing; however, it is important to start thinking about SHM holistically to bridge the gap between instrumentation, sensing, data processing, visualization, and final system robustness [36]. In other words, improved management through validation load testing should not be expected, and research and experimental resources are recommended to be consumed in well-designed and well-coordinated endeavors [23].

SHM technology is probably going to cause the next big revolution in civil infrastructure design, assessment, and management. When monitoring technology enters resource-limited environments, adoptions in concert between stakeholders with potentially conflicting goals for sustainable civil infrastructure projects are necessary to promote synergistic and effective solutions [1]. However, the lack of standardization of SHM principles and best practices is a significant barrier to adoption in concert and the application of promising SHM research to infrastructure SHM implementation [35]. For example, while a great deal of academic and commercial effort has been invested in the development of sensor technology and the design of sensor networks, there is actually a lack of consideration of the details and the many challenges faced [36].

The purpose of this review can be summarized as integrating and distilling two aspects from the literature. On the one hand, it corresponds to objective (a) to make the process of sustainable civil infrastructure SHM implementation more procedural and standardized. Categorization in SHM system is usually at the phase and module level, and the first goal of this review is to consolidate the categorization criteria at that level and to place a more complete content in a framework that encompasses both the phase and the more specific activity levels. On the other hand, corresponding to objectives (b) and (c), important details of the implementation process are analyzed to facilitate the realization of multi-stakeholders’ adoptions in concert. In addition, the repeatability of SHM measurements and data depends on many parameters that are captured in these details. In other words, the structured normative framework, the analysis of the influencing factors, and the proposed influencing factor relationship scheme can enhance the repeatability and variability of SHM. It should be pointed out, however, that due to the limitations of the scope and objectives of the review, this work is not focus on the subcategories of SHM to review the specific steps and techniques involved, related to which a lot of valuable works have already been published, e.g., Rosso et al. [20] and so on. Also, due to space constraints, specific case studies are not covered in this review.

3. Methodology

In this review, qualitative research method is adopted to conduct a literature review that is as integrative as possible. The defining characteristic of qualitative method is that it uses non-comparable and descriptive observation [37]. Qualitative analysis, then, is based on bits and pieces of non-comparable observations of different aspects of problem-solving [37]. Furthermore, the purpose of a qualitative review is to narratively draw conclusions from a set of studies [38]. Even when the object of the review includes quantitative findings from empirical studies, the conclusions are not quantified and appear more subjective and multifaceted than those of quantitative reviews [37,38]. Unlike the notion of a continuum of whether a review is descriptive or integrative, whether a review is qualitative or quantitative is a categorical difference [38]. This review acknowledges the shortcomings of not being able to quantify the findings. However, the reason for the adoption of a qualitative review is that such a method can achieve better integration based on relatively smaller numbers of the reviewed literature compared to the quantitative review method [38]. Specifically, the qualitative review method can range from purely verbal descriptions of findings in the area under review to integrating variations in findings [39]. Meanwhile, this integrative evaluation approach provides theoretical contributions that descriptive reviews cannot, requires less literature than meta-analysis, and is more flexible than meta-analysis in terms of conceptual integration [40,41].

The adopted specific procedure in this review is shown in Figure 1 and refers to the steps proposed by Dwertmann and van Knippenberg [38]. The first step is to define the topics and search strategy of the review and conduct the search. The second step is to perform a coding study in a large table based on the initial dataset in the review domain and selectively present this table in the final article based on the focus of the review and integration. The formed table covers the different subfields of the reviewed literature, and the iterative search of the literature and coding is continued systematically based on the different subfields. The third step is to complement the coding in Step 2 with theories from relevant fields outside the main review topic, based on the table obtained in Step 2 and continues to integrate theories from subfields of the main review field and from outside the main review field through coding. The fourth step is to propose a scheme based on the results of the integration.

3.1. Search Methodology

The data collection focuses on defining the scope of the review, literature search, practical screening, and quality assessment processes [42]. The main review area in the review is sustainable civil infrastructure SHM. Complementary topics include SHM for construction types except civil infrastructure, sustainable construction project factors, and life-cycle management for the operation and maintenance phase of construction projects. It is worth noting that the scope is limited to SHM for the construction industry, while SHM in other fields like mechanical and aerospace are not included. In addition, in terms of the nature of the literature, both scientific and gray literature are within the scope. Gray literature is considered to be relevant non-academic, industry-driven research findings [43]. For this review, in addition to the full academic papers, which make up a major part of the analysis, a few international standard documents, conference papers, and book materials were included in the analysis.

After determination of the review scope, a multi-stage iterative approach is used to systematically collect literature data. Literature data for this review is obtained from Google Scholar and Scopus, as these two databases provide wider coverage of scientific articles than any database [43]. The practical methodology employed for searching is through keywords as well as keywords in titles. Selected literature is published in the range of 2000–2023. The literature search is conducted based on three phases. The first and second phases are the specific actions of the second step of the abovementioned literature review procedure, which both are conducted within the main areas defined. The first phase is conducted through ‘sustainable civil infrastructure SHM management’, ‘sustainable civil infrastructure SHM implementation’, ‘sustainable civil infrastructure SHM process’, and other search terms to find relevant articles. Search terms ‘sustainable civil infrastructure SHM’ with ‘management’, ‘implementation’ ‘process’ ‘matter’, and other words are used in combination. A total of 11 searches with different combinations of search terms are conducted to cover all relevant articles as far as possible. In this phase, there are 63 articles initially selected, their abstracts and article structures are reviewed to remove duplicates and articles outside the main area, producing a database of 35 articles. Then, the selected 20 articles are read in full text, focusing on the SHM framework, process, execution, steps, and management aspects, and integrating these to produce a more comprehensive framework for SHM implementation.

The second phase refined new search terms based on the framework proposed in this review using subsets of the SHM area as the review scope. The literature search in the second phase is a multi-round iterative process, where more in-depth information about the subset of SHM areas obtained from abstract and full-text reading of the literature identified in the previous round inspired the search terms for the next round of the literature search. A total of 26 rounds of searches were conducted, with search terms including ‘SHM’ with the combinations of ‘sensor’, ‘data process’, ‘data transmission’, ‘damage feature extraction’, ‘damage diagnosis’, ‘decision-making’, ‘life cycle’, etc. Finally, the outcome is a database containing 153 papers.

The third phase reflects the third step of the abovementioned literature review method, taking ‘sustainable building projects’ and ‘sustainable infrastructure projects’ with ‘influencing factors’, ‘Indicators’, ‘Drivers’, ‘Barriers’, ‘Stakeholders’, etc., in combination as search terms, then reading abstracts and screening these articles to finalize a database of 21 articles in related fields outside the main search area. Finally, the articles in this database were read in full text, and the information gained complemented and helped to integrate the theories of the articles in the main field.

3.2. Data Analysis

The data analysis is conducted based on the Grounded Theory and through the coding study method. The significance of identifying and analyzing influencing factors is to bring guidance to the implementation of humans, and Grounded Theory is aimed at studying and clarifying the actions and behavior of humans based on induction and reasoning [44]. And the coding study can well identify the important contents and mark them with suitable terms to promote conceptualization. Two steps are included in the data analysis, so that the influencing factors in the SHM implementation process can be reviewed under the formed framework. The first step aims to construct this more integrative and systematic SHM implementation framework which contains two levels of categorization by open coding. Level 1 of categorization divides the process into four themes (phases). Then, Level 2 continued to categorize each theme into specific activities.

The second step is to conduct a coding study of content in the literature related to influencing factors based on the phases and activities identified in the framework. The main procedures involved in the coding study include open coding, axial coding, and selective coding. Firstly, in the open coding process, the literature in the obtained databases is reviewed based on the established framework, which aims to conceptualize and categorize the influencing factors in the implementation of SHM for sustainable civil infrastructures. Then, the axial coding process continued to generalize the categories based on the results of the open coding from two dimensions (contents and properties). In the final process, the outcomes of the open coding and axial coding are integrated and condensed furtherly in the selective coding, so that a scheme can be summarized and formed.

4. The Framework of SHM Implementation

One of the main barriers to the successful implementation of SHM in infrastructure is the technical aspect isolated from social, organizational, and managerial aspects [23]. To address this issue, a comprehensive sustainable civil infrastructure SHM implementation framework is proposed.

There is no standard or terminology for the delineation of phases or subsets included in the overall SHM system design and implementation, but the basic steps mentioned in the existing literature seem to be structural responses acquisition, data analysis and damage diagnosis, and prediction of future changes in the structure’s RUL. The further process also includes pre-assessment and post-decision making. In this section, based on the review of the descriptions and explanations, these major steps or modules are grouped into four categories representing four phases in order to summarize as comprehensively as possible the entire process of SHM. The obtained four phases include pre-assessment, data processing (acquisition, transmission, storage), data evaluation and damage diagnosis, and prognosis and decision-making. The specific coding results are shown in Appendix A (

Table A1. Coding results of SHM implementation phases.

Literature Resource	The Step/Module in the SHM Implement	Category (Phase)
[45]	Operational evaluation	Pre-assessment
	Data acquisition, normalization and cleansing	Data processing (acquisition, transmission, storage)
	Feature selection and information condensation	Data evaluation and damage diagnosis
	statistical model development for feature discrimination	Data evaluation and damage diagnosis
[154]	Sensor system	Data processing (acquisition, transmission, storage)
	Data processing system: data acquisition, transmission, and storage	Data processing (acquisition, transmission, storage)
	Health assessment system: diagnosis algorithm and information management	Data evaluation and damage diagnosis
[155]	Detection: if damage exists or not	Data processing (acquisition, transmission, storage)
	Localization: where is damage	Data evaluation and damage diagnosis
	Diagnosis: how severity is the damage	Data evaluation and damage diagnosis
	Prognosis: how much safe life is left	Prognosis and decision-making
[53]	Sensor system (structural health monitoring process)	Data processing (acquisition, transmission, storage)
	Data acquisition and transmission system (structural health monitoring process)	Data processing (acquisition, transmission, storage)
	The data processing and control system (structural health monitoring process)	Data processing (acquisition, transmission, storage)
	Structural health evaluation system (structural health evaluation process)	Data evaluation and damage diagnosis
	Structural health dataset management system	Data evaluation and damage diagnosis
	Monitoring and maintenance system	Prognosis and decision-making
[54]	Sensors	Data processing (acquisition, transmission, storage)
	Data storage	Data processing (acquisition, transmission, storage)
	Data transmission	Data processing (acquisition, transmission, storage)
	Database management leading to feature extraction	Data evaluation and damage diagnosis
	Data mining for feature extraction	Data evaluation and damage diagnosis
	Load/effect model development from study of data	Data evaluation and damage diagnosis
	Decision-making based on identified features in combination with identified models	Prognosis and decision-making
[46]	Measured the structural features using the sensors and other measurement instrument	Data processing (acquisition, transmission, storage)
	Measurement data transfer out of the field	Data processing (acquisition, transmission, storage)
	Data storage	Data processing (acquisition, transmission, storage)
	Check and analyze the monitoring results, and assess the structural health condition	Data evaluation and damage diagnosis
[156]	Design of the sensing, data acquisition, and communications	Data processing (acquisition, transmission, storage)
	Field implementation and challenges	Data processing (acquisition, transmission, storage)
	Data analysis and information generation	Data evaluation and damage diagnosis
[32]	Determination of target structure: consider whether the structure would benefit from SHM	Pre-assessment
	Risk analysis: identify the risk, degradation, and the related probability corresponding to the specific structure	Pre-assessment
	Estimating and correlating structural responses to risk and degradation events	Pre-assessment
	Design of SHM system and sensors selection	Data processing (acquisition, transmission, storage)
	Installation and correction	Data processing (acquisition, transmission, storage)
	Data acquisition and management	Data processing (acquisition, transmission, storage)
	Data evaluation: confirm the structural damage, generate alerts, warnings and periodic reports, then promote further action	Data evaluation and damage diagnosis
[19]	Damage detection/characterization: structural response measurement data acquisition, data normalization and analysis, damage confirmation through extracted damage-sensitive features	Data processing (acquisition, transmission, storage), data evaluation and damage diagnosis
	Prognosis: estimate structural performance in the future	Prognosis and decision-making
	Risk assessment: assess the risks associated with a decision, weighing the probabilities of the predicted scenarios, the costs of dealing with them, and the consequences of not dealing with them	Prognosis and decision-making

).

Pre-assessment (P1) describes the whole assessment before the acquisition of structural response data, for example, determining the structure needed to be monitored, what and how the data will be collected, and identifying the natural and operational environment the structure is in. Farrar, Doebling, and Nix [45] proposed an operational assessment process that consists of the socio-economic reasons, the probable types and likelihood of structural damage, the operational conditions and the natural environment, and the potential SHM implementation constraints. Inaudi [32] mentioned the pre-assessment steps to identify structures to be monitored, predicting and analyzing degradation and events that may affect the health of the structure and the corresponding structural responses. However, there are subtle differences between the pre-assessment in the studies of Inaudi [32] and Farrar and Doebling and Nix [45]. Farrar, Doebling, and Nix [45] give more consideration to possible barriers, while Inaudi [32] deeply analyzes possible structural responses corresponding to the risk and degradation events, emphasizing the most comprehensive possible consideration of potential responses through the correlation between the responses.

The category of data processing (acquisition, transmission, storage) (P2) can be explained as the data processing activities that can promote the quality or value of the data but do not involve the extraction and selection of structural damage features. In particular, data storage may be an activity that is born from the first signal acquired by a sensor node into data and continues throughout the whole processes of SHM [46]. For the sake of simplicity, consider that it does not involve a change in the SHM-relevant properties of the data and that the logical stage which it emerges—data storage—is also categorized into the phase of data processing (acquisition, transmission, storage). The semantics of this code emphasize the object status of the raw data. So, even though no data have been generated yet during the sensor selection and placement, these activities are steps in the data acquisition process and still fall under the data processing category. Furthermore, because raw data preprocessing (including data cleaning and normalization, etc.) is usually conducted in sensor nodes, it is contained in data acquisition.

Data evaluation and damage diagnosis (P3) usually start from feature extraction and end in damage diagnosis. This process is the core of SHM because SHM is essentially a statistical pattern recognition problem [3,47]. After the normalization and cleaning, the high-quality structural response data will be used to fit the data-based models (time-domain) or physics-based models (frequency-domain) to illustrate the features that might provide a guide to damage diagnosis [48]. Then, in terms of damage diagnosis, the related methods also can be categorized as a physics-based approach (finite element model (FEM)) and a data-based approach (machine learning). The former is to update FEM to fit the damage feature to diagnose the damage [49], and the latter is using the statistical model based on the training dataset of experimental structural response as a diagnosis tool [50].

The phase of prognosis and decision-making (P4) consists of the prediction of critical future performance changes in the RUL and weighing the costs and outcomes of interventions [19]. This phase is mainly concerned with the life-cycle management of the operation and maintenance phase of sustainable civil infrastructure, including the design and investment decisions of maintenance and renovation [47]. To summarize, P4 is a key stage for the value realization of SHM [19].

Based on the existing literature and the abovementioned four phases, the open coding method is adopted to label the main activities, so that the framework for SHM design and implementation is shown in

Table 1. SHM implementation framework.

SHM Implementation Phases (Categories)	Activities (Subcategories)	Literature Resource
Pre-assessment (P1)	Identifying the structure needed to be monitored (P1A1)	[32]
	Risk analysis: identify the risk, degradation, and the related probability corresponding to the specific structure (P1A2)	[32]
	Considering responses of structure to risk and degradation events (P1A3)	[32]
	Reliability Analysis of SHM System (P1A4)	[45]
Data processing (acquisition, transmission, and storage) (P2)	Sensors network design and placement (P2A1)	[51]
	Sensor installment and correction (P2A2)	[32]
	Raw data pre-processing (P2A3)	[52]
	Data transmission (P2A4)	[53]
	Data storage (P2A5)	[46,54]
Data evaluation and damage diagnosis (P3)	Data feature extraction (P3A1)	[55,56]
	Damage diagnosis (P3A2)	[57,58,59]
Prognosis and decision-making (P4)	Evaluation on the remaining useful life (RUL)(P4A1)	[60,61,62]
	Structural maintenance and renovation decisions (P4A2)	[63,64,65]

. The influencing factors are identified based on this framework.

5. Sustainable Civil Infrastructure SHM Influencing Factors

This section integrates and analyzes the influencing factors based on axial coding on their two dimensions (contents and properties), coding the factors involved in each stage except for pre-assessment to obtain categories and sub-categories from the content dimension, and coding the factors involved in each stage from the properties dimension, the details of the properties are shown in

Table 2. The details of properties for influencing factors.

The Properties of Influencing Factors
Action oriented	Form/method factor
	Decision-making factor
	Important consideration factor
	Organizational factor
Result oriented	Contributing factor
	Critical factor
	Risk factor/challenge
Sustainability	Economic factor
	Social factor
	Environmental factor

. The complete and detailed coding results are presented in Appendix A.

5.1. Pre-Assessment

Giving adequate attention to the planning phase plays a crucial role in the successful implementation of SHM in sustainable infrastructure [66]. However, most research and inexperienced SHM system technicians and executors tend to ignore this process and proceed directly to the design of the SHM system [23]. Such a situation will cause some difficulties in the design and implementation, resulting in poor outcomes for the benefits of the SHM system [23]. Meanwhile, SHM implementation relies on financial support and human resources; therefore, it is important to consider the purpose, role, and feasibility of the proposed SHM system before the design and implementation [67].

It can be found that the pre-assessment can be divided into two main aspects. On the one hand, it includes the assessment on operational and natural environment that the monitored structure faces, represented by activities of P1A1\P1A2\P1A3. On the other hand, it is represented by activity P1A4, specifically involving the reliability assessment on the potential SHM system with higher-level considerations before the formal design and implementation.

In the process of the assessment on operational and natural environment, most of the influencing factors are categorized as decision-making factors, such as factors that affect the choice of targeted structure to be monitored, the choice of what type of data to be collected, and the decision-making about monitoring strategy. ISO 14963:2003 [5] states that SHM systems should be designed to suit the specific purpose and type of structure, and categorizes monitoring strategies in terms of the range of objectives and the technology employed, etc. ISO 16587:2004 [68], on the other hand, standardizes the description of the performance parameters for evaluating the condition of a structure and provides guidance on the selection of target parameters, the determination of thresholds, and the choice of the form of monitoring according to the type of loading the structure is experiencing. ISO 14963:2003 [5] recommends that the basic monitoring methods be pre-arranged according to the target information that is desired and divides the response measurement system into global measurements and localized measurements. For global response testing, there are different configurations for each key location in the global deployment [5]. These designs should be processed after the preliminary investigation of pre-assessment [5]. The monitoring strategy for testing the local response requires a theoretical analysis of the structural elements to plan the location of the instruments [5]. Specifically, a sustainable civil infrastructure project often contains a wide range of structures, and a contributing factor is to identify which structure will be more likely to benefit from the SHM system at the start of pre-assessment. Regarding this point, Inaudi [23] reported a few types of structures, including structures with innovations in terms of design/construction procedures/materials, structures facing unusual risks or uncertainties, structures that are critical at the network level, structures that are representative and likely to generate greater information value, and structures that have known defects or low visual ratings in structural safety, etc. Meanwhile, Imam, and Chryssanthopoulos [69] statistically analyzed failure cases of metal bridges in the past, which helps to pre-assess the potential risk of bridges, and also proposed a risk categorization procedure to anticipate the responses of bridge structures to different risks, preparing for rough quantification of expected response and pre-determined extent of monitoring. It is worth noting that the analysis of the risk should not only concern the comprehensive assessment on the environment around the structure but also emphasize reasonableness, such as discarding risks with a low impact or likelihood rather than emphasizing all possible risks [23]. And the extent of monitoring indicates the pre-assessment on the monitoring level of the different environments and structures before formal SHM system implementation, such as the function of measurement instrument and the noise robustness. Furthermore, ISO 14963:2003 [5] recommends theoretical modeling and/or numerical analysis to obtain the magnitudes of the values to be measured prior to formal monitoring. These factors unify and guide subsequent detailed sensor network design.

Compared to the pre-assessment on the structure, there is more literature focused on reliability analysis of SHM system, which is mostly related to the cost and economic benefits. However, ISO 16587:2004 [68] standardizes the aspects that should be considered for the feasibility of acquiring measurements, including ease of acquisition, complexity of the required data acquisition system, level of data processing required, safety requirements, cost, and the existence of a monitoring or control system that already measures the parameter of interest. These aspects should be evaluated before making monitoring strategy decisions. The study by Farrar, Doebling, and Nix [45] mentions that technical limitations in evaluation should be considered. It is worth noting that most of the risk factors and challenges involved in the pre-assessment are emerged in technical limitations and influenced by organizational factors and form/method factors. For example, in practice, the return on investment of soft benefits such as knowledge, safety, and reliability are difficult to be quantified by proven methods, research and development funding is controversial to include in the cost budget due to the specificity of the source, there are a lack of databases and consistent and reasonable assumptions during the pre-assessment, and SHM engineers may have limited knowledge in the cost–benefit methodology based on the economic criteria [70]. The abovementioned situations may directly lead to reduced accuracy and increased difficulty in estimating economic aspects. However, soft-benefit estimation methods and the lack of consistent and reasonable assumptions can be improved through the accumulation of statistical information and standardization efforts by industry associations, the issue of disputes related to research and development funding requires stakeholders to identify the source of the funding, and to reach a consensus on how much of that cost is to be counted as a SHM system cost [70]. Meanwhile, the knowledge of SHM engineers in the economic domain needs to be enhanced through organizational learning, or by reducing the negative influences through a rational division of labor, so that theories and technologies related to the economic domain can efficiently reflect their value in the SHM field [71]. In addition, structured approaches that facilitate formal discussions among stakeholders are gaining more attention. For example, Vardanega et al. [72] presented an SHM value assessment methodology that discusses the roles of different stakeholders in delivering value, and the proposed value assessment matrix facilitates efficient and effective communication among stakeholders. Omenzetter et al. [73] proposed a framework for quantifying the value of the information generated by an SHM system, including decision trees and risk models, which promote the valuation of the soft benefits of SHM systems and the accuracy and objectivity of the assessment. Nepomuceno et al. [74] introduced a methodology for assessing the value of an SHM system prior to its deployment. Here, the output is a numerical value that expresses the approximate likelihood of the system generating value for the owner, which improves the efficiency of the communication between stakeholders by enhancing the intuition of the information.

5.2. Data Processing (Acquisition, Transmission, Storage)

The data acquisition, transmission, and storage involve many details that affect the achievement of a sustainable SHM system. Based on an integrative review of the related literature, it is discerned that this theme can be classified into three categories and twelve sub-categories as presented in

Table 3. Influencing factor categories for data processing (acquisition, transmission, storage).

Categories	Sub-Categories	Activities	Factor Properties Involved
			Orientation	Sustainability
Sensor placement	The selection of monitoring strategy	P1A1; P1A2; P1A3; P2A1; P2A4	Organizational factor; technical factor	-
	The decision on the types of sensors and monitoring network	P2A1; P2A2; P2A4	Organizational factor	-
	Indicators for the selection of sensing technologies	P2A1; P2A3; P3A2	Decision-making factor; organizational factor	Economic factor; environmental factor; social factor
	Development on the algorithm for the optimization of sensors placement scheme	P2A1	Organizational factor; technical factor; risk factor/challenge	Social factor; economic factor
	Sensor network innovations and use of advanced technologies	P2A1; P2A2; P2A3; P2A4; P2A5	Technical factor; decision-making factor; risk factor/challenge; organizational factor	Economic factor; social factor
	Installation and calibration of equipment in the field and self-monitoring of sensors	P2A2; P3A1; P2A3; P2A4	Organizational factor; technical factor; important consideration factor; decision-making factor,	Social factor; environmental factor; economic factor
	Sensor Energy Strategies	P2A1; P2A3; P2A4; P2A5	Technical factor; organizational factor; form/method factor	economic factor; environmental factor
Raw data preprocessing	Data normalization	P2A1; P2A3; P3A1	Technical factor; organizational factor; risk factor/challenge	-
	Data cleaning	P2A1; P2A3	Organizational factor; technical factor	-
Data transmission and storage	Data communication	P2A1; P2A3; P2A4	Technical factor; decision-making factor; organizational factor;	Environmental factor; economic factor
	Data storage and database management	P2A5	Decision-making factor; important consideration factor; technical factor; organizational factor;	Social factor; economic factor
	Data security	P2A4; P2A5	Technical factor; organizational factor	Social factor

. The two categories ‘sensor placement’ and ‘raw data preprocessing’ are related to data acquisition. Data transmission and storage are determined as one category because some of the involved influencing factors are applicable to both activities.

Sensor placement is a critical factor because it affects the implementation of subsequent activities and system performance [75]. Decision-making factors of sensor placement mainly contain monitoring requirements determination based on knowledge of the infrastructure project under observation and the evaluation of potential sensor solutions’ attributes. These two aspects are interdependent and should be jointly contemplated [76,77], so that iterative decisions on sensor placement can be made, which satisfy monitoring needs without being excessively extravagant. Specifically, many studies highlight the advantages of wireless sensor networks over traditional, labor-intensive wired solutions in terms of convenience, flexibility, and cost-effectiveness [78,79], especially in large-scale structural monitoring [80]. However, despite the developing trend of sensors being wireless [78], the selection of sensor network types still depends on the properties of specific structures [75]. For instance, wireless communication may not be feasible for intricate steel structures [81]. In this regard, ISO 14963:2003 [5] provides the basis for the implementation of this factor by categorizing roads, bridges, and other construction structures according to material, static design, construction method, and function.

The issue of sensor placement refers to decisions on resource allocation (e.g., the number of each type of sensor and the required sensitivity) and the arrangement of sensor locations [82]. An important aim of the sensor configuration approach is to solve three difficulties that objectively exist in the inverse process of inferring structural parameters from measured responses in practical engineering—the under-posing problem, the sensitivity requirement, and the problem caused by differential operators in the forward governing equations [83]. There are many kinds of performance criteria for sensor configuration schemes, which are categorized into four main groups: modal parameter-based, response reconstruction-based, energy-based, and information-based [82]. Tan and Zhang [84] provide a more specific discussion of sensor configuration evaluation criteria, including the modal assurance criterion, which makes modal vectors easier to distinguish, the Fisher information matrix, which relates to the information stored in the measured response, the information entropy criterion that quantifies the uncertainty of such information, effective independence (EI) criterion that maximizes such information, and effective independence driving-point residue criteria that combines EI indices with associated driving-point residue weighting to improve the signal-to-noise ratio, and criteria based on measured structural energy. Based on the above evaluation criteria, a variety of optimization methods can be used to solve the sensor placement problem, mainly including evolutionary optimization methods, sequential sensor placement algorithms, deterministic optimization methods, probabilistic methods, and so on. Since sensor placement design for infrastructure usually involves numerous degrees of freedom [85], the assistance of optimization algorithms might be helpful. Hassani and Dackermann [86] reported optimization algorithms for SHM sensor placement and their features with suggestions in specific applications and illustrated that using optimization to find optimal solutions for sensor deployment in SHM systems is becoming more widespread. Capellari, Chatzi, and Mariani [87] proposed a sensor placement optimization approach based on cost–benefit considerations to maximize the information obtained from the measured data and optimize the information-to-cost ratio. Li et al. [88] investigated sensor placement issues from the perspective of civilian requirements, proposing an optimization methodology to enhance energy efficiency based on computer science. In particular, evolutionary algorithms, which are the most widely used and advantageous, use an evolutionary strategy to update the population based on a fitness function representing the evaluating parameter until a certain termination criterion is satisfied [84]. Genetic algorithm, monkey algorithm, particle swarm optimization, firefly algorithm, simulated annealing algorithm, and ant colony optimization are representative evolutionary algorithms [84]. Improvements are often made to these algorithms in implementation, and they share common improvement strategies, including decoding systems, new operators, improved movement schemes, and iterative strategies [84].

Not only does sensor placement benefit from advanced methods and intelligent algorithms, but the overall development of SHM in recent years has been transitioning from ‘automated monitoring’ to ‘intelligent monitoring’ [89]. For example, the practical application of real-time Global Navigation Satellite System high-precision technology in civil infrastructure SHM is investigated in the study by Cinque et al. [90]. The findings show that the system is most cost-effective in the absence of ground monitoring systems due to the difference with ground monitoring systems, and the lack of accuracy in real-world scenarios compared to laboratory scenarios. These points should be concerning when the technology has been applied to the implementation of SHM. Ozer, Feng, and Feng [36] reported an innovative smart SHM platform based on smartphone sensors and crowdsourcing concepts, in which machine learning techniques are introduced to solve data quality problems. Then, the contribution of Martín et al. [91] is to propose an Edge/Fog/Cloud architecture for civil infrastructure SHM, which makes the implementation of SHM more flexible and highly versatile. A valuable finding from the synthesis is that the integration-related factors of smart sensing technologies are paid considerable attention in academia [89]. For one thing, an integration strategy is a critical factor in exploiting the synergies between different types of sensing/diagnostic technologies and advanced concepts from other domains [89,92]. For another, it needs to be driven by both technical and organizational levels. For instance, Boyle et al. [93] proposed a multi-vendor, multi-sensor interface SHM sensor network capable of integrating a range of advanced technologies and integrating energy analysis of these sensors. And an important factor for the implementation of this solution is the synergy and close communication between end-user experts, sensor network system developers, and SHM system contractors [94]. In addition, smart technologies are more intensively used for self-monitoring in sensor operations, as the utilization of machine learning and artificial intelligence algorithms provides a solution to automatically detect and handle sensor failures during daily operations, which can be determined as a critical factor in regular operation of the SHM system [75].

Although many efforts and achievements in terms of smart technology development and specific practices have been contained in the SHM sensor network, the data collected by the sensor nodes are still subject to constant variability in the natural and operational environments [95]. Thus, the raw data should be processed, including data normalization and data cleaning, such processing is conducted before damage feature extraction [3]. Furthermore, the recognition and quantification of such sources of variability are one of the influencing factors in the development and selection of data deviation identification and normalization procedures, because the decision criterion may be dependent on the degree of recognition of the sources of variability [96,97]. According to ISO 16587:2004 [68], uncertainty includes calibration uncertainty, error due to sensor installation, instrument measurement fluctuation, and environmental influences on the measurement system. These errors can be minimized by proper application of standards such as ISO 5347 [98], ISO 5348 [99], ISO 7626 [100], and ISO 16063 [101]. Meanwhile, the quality of the training data used in the development of the data processing algorithms is also influenced by the variability of the sources [95]. Data cleaning is necessary when environmental influences are excessive or interference signals are contained in the measurement. In this situation, the knowledge and experience of the staff directly involved in data measurement and the utilization of filters are both important factors [102]. From both perspectives of sensor self-monitoring and data processing at the sensor network level, new finding from an integrative review indicates that identifying anomalous data types and refining the targeting of data processing procedures might pose a critical factor and challenge. Anomalous data caused by the environment, sensor failure, and structural damage may be confounded [103], and data features caused by structural damage may be masked by the data normalization process [96].

It is worth noting that sustainability factors are most concentrated in the phases of data acquisition, transmission, and storage. Promoting resource efficiency in sensor network placement and operation constitutes the main economic factor [75,90,104]. For the social category, structural safety [86,105], user experience [88,91,106], positive incentives [36], safety of personnel [107], data security [46,108,109], and energy-saving sensor network [86,93,110] hold significant potential to contribute to energy conservation and reduced emissions. An observation derived from the literature synthesis suggests that although environmental factors driven by energy conservation may not definitely be an economic factor (because the cost might not only be from energy consumption but also technological or hardware aspects potentially cause larger economic expenditures), enhancing energy utilization efficiency through work scheduling can achieve both environmental and economic sustainability, simultaneously. For sensor nodes, managing duty cycles (such as scheduling sensor sleep and awake times and other operational strategies) can facilitate energy conservation at the sensor network level [111]. In data communication, scheduling strategies can address numerous resource efficiency issues, including low data packet utilization, channel congestion, and idle listening [104].

5.3. Data Evaluation and Damage Diagnosis

Data evaluation and damage diagnosis are mainly to determine the existence, location, and severity of damage. The factors involved in this theme can be divided into two main categories (i.e., feature extraction and damage diagnosis) and six subcategories, as shown in

Table 4. Influencing factor categories for data evaluation and damage diagnosis.

Categories	Sub-Categories	Activities	Factor Properties Involved
			Orientation	Sustainability
Feature extraction	General matters	P1A2; P1A3; P2A1; P3A1	Technical factor; organizational factor	-
	Factors for data-based feature extraction (time domain)	P3A1	Technical factor; important consideration factor	-
	Factors for physics-based feature extraction (frequency domain)	P3A1	Important consideration factor; decision-making factor	-
Damage diagnosis	Physics-based damage detection methods	P3A2; P2A1; P3A1; P1A1; P1A3	Important consideration factor; risk factor/challenge technical factor	Economic factor
	Data-based damage detection methods	P3A2; P2A1; P3A1	Important consideration factor; contributing factor; risk factor/challenge factor; technical factor	Economic factor
	Integration of data-based and physics-based damage detection methods	P3A2	Contributing factor; important consideration factor; technical factor	Economic factor

.

5.3.1. Feature Extraction

It is a fundamental axiom of SHM that system response data do not directly indicate damage, and thus the effectiveness of feature extraction is a critical success factor for damage diagnosis and subsequent tasks. Factors in the category of feature extraction generally aim at the automatic identification of modal parameters and/or generation of time series models in the ways practically applicable [56,112,113], flexible [114], and well-fitted to actual response data [56,113]. The specific situation-based consideration in the process of feature extraction is an important factor in putting technical research into practical production. García-Macías and Ubertini [112] acknowledged that it is necessary to design the strategy of feature extraction according to evaluation criteria of the actual operation. Specifically, the design should consider the adopted sensing technology, the accessibility constraints in fieldwork, the characteristics of the type of structure to be monitored, and the properties of the engineering materials; furthermore, some items, as formed in the pre-assessment (P1), such as the risks and degradation prediction and how the corresponding structural response in the structure is going to be monitored, should also be included in the considerations [112]. Meanwhile, Amezquita-Sanchez and Adeli [115] indicated that robust signal processing techniques are critical in the feature extraction process because the techniques can make changes in structural response easy to observe. Furthermore, García-Macías and Ubertini [112] reported that it is needed to combine such techniques with a priori statistics and practical engineering judgment.

For the data-based feature extraction strategy, appropriate time series models should be selected to fit the measured vibration data based on the data’s nature (e.g., linearly smooth or nonlinearly smooth) [113]. Specifically, determining the appropriate order of the time series model is an important consideration, which directly affects the reliability of the damage detection results. The appropriate order is the order that enables the time series model to give uncorrelated residuals [113]. If the order is insufficient, the result of feature extraction will lack accuracy. If the order is too high, the time series model will lack reliability because vibration signals are random in nature and cannot be precisely predicted by a function in the time dimension. In addition, since associating measured data with first-hand observations of degraded systems is one of the most commonly used methods, the results of feature extraction should also be represented in the machine learning process [113]. For physics-based feature extraction, the choice of frequency-domain modal identification techniques requires more practical consideration [56]. When a priori knowledge about the signal source is available, parametric signal processing techniques have more fidelity. Otherwise, modal parameters need to be extracted by nonparametric techniques such as frequency response functions [116]. It is implied that the strategy design and method selection for feature extraction should be analyzed based on the specific and actual situation. Therefore, the close cooperation and frequent communication between professionals such as SHM engineers, structural engineers and technical consultants, the efforts at the organizational level in terms of decision-making process, and knowledge transfer may be important [23].

In addition, the automation and real-time feature extraction process have received much attention in the research field. Ubertini, Gentile, and Materazzi [117] introduced a variety of techniques that enable automated extraction of modal features from structural response data in long-term monitoring. Khoa et al. [118] adopted machine learning techniques for vibration data feature extraction to improve the computation time over the non-use of the method by a factor of 200, while the accuracy can be ensured well, further greatly facilitating the process of real-time feature extraction. Because the flexibility and accuracy of the civil infrastructure SHM feature extraction framework relies on massive data analysis and high-fidelity models with high computational requirements, real-time operations and automation are critical factors to improve the usability of processes with large computational requirements.

5.3.2. Damage Diagnosis

The identification of damage features and the damage diagnostic tools (FEMs/statistical identification algorithms) are major factors in achieving an effective and efficient damage diagnostic process.

For feature identification, effective damage features should be time- or frequency-domain model parameters that are sensitive to changes caused by damage and insensitive to other variations. For physics-based diagnostic methods, a combination of information from multiple sources is often a critical factor to obtain appropriate features [48]. These sources of information may include actual degradation experiments such as induced damage experiments, corrosion growth, temperature cycling, and so on [48]. For data-based diagnostic methods, the process of ranking and selecting valid features from the results of feature extraction is often automated through machine learning algorithms trained by large-scale heterogeneous datasets [112]. In particular, Ying et al. [114] utilized adaptive enhancement algorithms to linearly integrate multiple weighted weak classifiers to enhance the effectiveness of feature selection, and feature libraries have been created that enables machine learning to automatically search for features applicable to different types of damage, leading to improve the relevance of feature selection to different specific situations in practice scenarios.

For damage diagnostic tools, synthesis finds that the factors of such tools contributing to the successful implementation of SHM can be categorized into two groups: the quality of the model used for diagnosis and the performance of the diagnostic process. For physics-based damage detection methods, the main factors mentioned in the literature to enhance the quality of the FEM include the utilization of real engineering damage data [119], the comprehensiveness and validity of the information involved in the establishment of the initial FEM [58], the results of the trait selection layer used for the updating and testing of the FEM, and the ability of the FEM to decouple the confusions between structural parameters [120], etc. Regarding the enhancement of the speed of FEM updating, the factors include the application of genetic algorithms [121], the use of image processing and pattern recognition techniques for data compression prior to the execution of the comparison procedure [122], and the introduction of the Kriging model, which has merit of fast runtime, as a meta-model to directly replace iterative analysis for optimization process [123], etc.

For data-based damage detection methods, efforts to improve the algorithm quality mainly include enhancements of real-time performance, inclusiveness, and applicability. Tibaduiza Burgos et al. [58] reviewed the application of data-driven aspects of damage diagnosis, pointing out the trend toward real-time diagnostic processes and the contribution of automation factors to reducing maintenance costs. Currently, the most commonly used techniques are represented by a convolutional neural network (CNN) because of their powerful ability to revolutionize traditional feature engineering methods by directly processing raw data and automatically extracting hidden patterns. Their convolutional layers are able to automatically extract features from an image, generate feature maps, and activation functions without the need for a feature selection process, and then, the flattened output is fed back to the final multi-layer perceptron to perform the classification task [20]. In the research of Su et al. [124], CNN is incorporated into the Subspace System Identification with Covariance Analysis (SSI-COV) framework to perform image recognition tasks thereby enabling automated operational modal analysis to improve conventional manual procedures. In terms of improving the applicability of algorithms, Silva et al. [125] proposed a novel unsupervised nonparametric genetic algorithm for decision boundary analysis (GADBA) for covering linear and nonlinear anomalies caused by non-damage that are not completely eliminated by the data normalization process. The method demonstrates better classification performance than before in avoiding both false alarms and underreporting. Santos et al. [126] used four Kernel-based damage detection algorithms for different operational and environmental conditions. Compared with other algorithms that are considered reliable, the proposed algorithm proves to have better classification performance. Furthermore, Zhang et al. [127] considered the connection between vibration characteristics and space, such that the measured response may be related to the sensor placement. In the above research, recurrent neural networks (RNN) combined with Long Short-Term Memory methods (LSTM) extract spatial features to create a reliable agent data-driven model which enables prediction of bridge response.

In particular, the integration of data-based and physics-based damage detection methods is an important factor for successful SHM implementation, as it offers the possibility of combining the advantages of the two methods. Ritto and Rochinha [128] proposed an integration of physics-based damage detection method and a machine learning technique based on the idea of the digital twin concept. The physics-based model is used to construct the digital twin model, so that the interpretability and accuracy of the physics-based approach can be combined to the simplicity and real-time advantages of the data-based approach, leading to real-time reporting of engineering damage information. Ozdagli and Koutsoukos [129] reported a physics-knowledge-guided supervised learning method to train the damage identification model, and integrate the physical parameters extracted from the physics-based simulation data into the intermediate layer of the neural network to replace the unsupervised learning. Based on the study by Ozdagli and Koutsoukos [129], Zhang and Sun [130] used the damage identification results from iterative testing of the physics-based FEM to extend the original modal property-based features, overcome the challenges of insufficient data-driven training data for damage diagnosis as well as modeling errors due to idealization of the physics-based model, and the process of damage detection can be more useful and scientific. Based on the integration of the relevant literature, the critical factors for the successful integration of the two types of methods in practice are summarized as follows: First, both physics-based and data-based methods should be calibrated/trained and validated/tested with experimental or field data, and the calibration/training process and validation/testing process should be chosen to use different data, otherwise, the validation/testing process will be meaningless [128]. Second, it is necessary to normalize the data, parameters, and objective function. High-fidelity and interpretable physics-based models can be used to analyze a wide range of new scenarios. In case there are unknown parts of such a model, another mathematical structure is supposed to emerge [128]. Third, the combination of the two types of methods tends to cause numerous calculations, which is related to real time application. Measures to reduce computational costs are a critical factor in improving the usability of the combined method, such as building reduced-order models that act as proxies for the original expensive models [131].

5.4. Prognosis and Decision-Making

The original goal for SHM is to conceptualize and detect structural damage in order to target maintenance and renovation or manage structural RUL in suitable ways [132]. Meanwhile, the incorporation of SHM information has significantly improved the predictive accuracy of life-cycle assessment [133,134]. Therefore, damage information is an important but not the ultimate goal of SHM, but the ultimate significance of SHM is to make sustainable decisions on the building life cycle based on the prediction of changes in subsequent performance and the RUL of the structure. And for owners and other non-structural engineering stakeholders in building lifecycle management, they will not benefit from complex interpretations of structural damage monitoring data, but the most meaningful and important thing is to provide concise decision-support information that can improve their productivity. The synthesis finds that the factors in this phase can be categorized into those that affect life-cycle performance and those that affect the design of the decision framework, as shown in

Table 5. Influencing factor categories for prognosis and decision-making.

Categories	Sub-Categories	Activities	Factor Properties Involved
			Orientation	Sustainability
The performance of life-cycle model	Data quality	P4A1; P2A5	Important consideration factor; organizational factor; decision-making factor; contributing factor	-
	The indicators of structural aging	P4A1; P4A2	Important consideration factor; organizational factor; technical factor; contributing factor	-
	Application of tools or theories	P4A1	Technical factor; important consideration factor; contributing factor	-
	Formation of prognostic result	P4A1; P4A2	Technical factor; critical factor; important consideration factor; contributing factor	-
	Diversity of prognostic result information	P4A1; P4A2	Important consideration factor; contributing factor	Economic factor
The design of decision-making support framework	Decision-making goals	P4A2	Important consideration factor; contributing factor	Economic factor; social factor
	Principles and methods of decision-making	P4A2	Technical factor; critical factor; contributing factor Important consideration factor; critical factor	Economic factor
	Comprehensive decision-making considerations at the practical implementation level	P4A2	Important consideration factor; contributing factor; Risk factor/challenge; critical factor	Social factor

. Life-cycle modeling and decision-making frameworks may be an effective way to transform SHM data into decision-support information.

5.4.1. The Performance of a Life-Cycle Model

The performance of a life-cycle model depends primarily on the provided data, the chosen structural indicators for predictions, the application of tools or theories, the generation of prognostic results, and the diversity of the information it contains. The data used for the life-cycle model consists of the data in the development of the model and the input data for predictions.

For the data used for model development, the essential factors in achieving a high-quality prognosis may have two aspects; namely, the validity and relevance of the structure to be predicted and the representativeness illustrated by the amount of data [135]. Furtner and Veit-Egerer [135] developed a specific life-cycle model for the structure that is monitored based on the state-of-art information in the existing papers, and 50 years of monitoring experience on the same type of infrastructure all over the world. The adopted theoretical curves are the result of the average of a large amount of structural and empirical baseline data, and the inputs include a database provided by the client with a large amount of rich structural information, thus yielding rich, detailed, and valuable information about the life cycle of the structure, including information about the subsequent components that are most likely to demonstrate sustainability. Experience served directly as a source of model relevance and validity in this study. Meanwhile, the study by Wenzel, Veit-Egerer, and Widmann [136] adopted the same idea to propose a strategic idea, called the open life-cycle model. This idea is based on a degradation model that is generic but flexible to suit the future infrastructure to be predicted, and it can provide validity for the future infrastructure. Experience from a variety of sources provides representativeness and reliability to the model based on this idea. Furthermore, the time scale of the input data and the calibration of the model are also important sources for predicting the reliability of the structure [133]. It is worth noting that inputting field data for calibration improves the model accuracy more than standardized data [137].

And the type of input data depends on the structural aging indicators. The indicators considered for prognosis based on life-cycle modeling should be rich and valid [136]. In other words, they should be closely related to structural aging, including structural parameters, and environmental conditions [136]. In particular, Jayathilaka [138] innovatively introduced the state-level residence time as a parameter for predicting the future state of the member, representing each possible state transition of the member to be tested throughout its life cycle based on the nine different types of residence times defined, and demonstrating the importance of this parameter. In addition, the flexibility of the model is an important factor in its ability to freely make changes to the designed parameter library [136]. Then, during model development and utilization, the use of theories/tools may largely determine the ideas and principles/efficiency and functionality of the life-cycle model [133,135]. In the prediction, the life-cycle curve is considered a critical factor in playing the role of a prognostic process information characterization due to the fact that it is often used as a basis for maintenance planning decisions [135]. The construction of a risk-rating system helps to provide detailed information on the subsequent performance of structures at different locations contained in the infrastructure [136].

5.4.2. The Design of Decision-Making Support Framework

The design of the decision-making support framework greatly influences the successful implementation of SHM for sustainable civil infrastructure. This is because it is the decisions and behaviors of stakeholders that illustrate the value of the damage diagnostic information provided by SHM. An effective and reasonable decision-making framework not only provides suitable solutions for real-world situations, but also promotes synergistic communication and agreement among multi-stakeholder groups. Based on a review of the literature on infrastructure life-cycle decision-making in which SHM has been involved, it can be found that the success of a decision support framework depends primarily on the identification and balance of decision goals, the development and adoption of principles and methods, and the detailing and comprehensiveness of decision considerations. When considering decision optimization at the objective level, it is essential to account for multiple objectives and constraints, with particular attention to stakeholder budgets. Frangopol and Liu [139] pointed out that most existing infrastructure maintenance and management decisions are based on the principle of life-cycle cost minimization, but such single consideration often fails to provide long-term performance that meets production and development needs. Therefore, it is suggested to simultaneously consider multiple, often competing decision-making objectives such as cost, budget, safety, and structural condition, then provide a compromise solution [139].

Scientific and available decision-making principles and methods may be an important bridge from research to implementation in the prognostic and decision-making phases of SHM [134]. The introduction of probabilistic methods may be an important consideration because it is a good way to cover the uncertainty of the actual development and maintenance of the structure in the decision-making scheme, as such uncertainty cannot be completely eliminated [134]. Also, the probabilistic approach allows the obtained information to be scalable, so that the limited SHM information can be utilized well, leading to the monitoring of cost reduction [140]. In addition, event trees may be a commonly used decision-making tool [134,141]. From the perspective of effective utilization of SHM information, how to integrate the information with the decision-making process should be considered as a factor [142]. Following this idea, Orcesi and Frangopol [142] presented a methodology for using SHM data for updating the probability density function of the failure time of a structure through a Bayesian process. The results demonstrated that the SHM system is able to update the knowledge of the structure thus improving the effectiveness of the maintenance and refurbishment program.

It is worth noting that SHM-supported decision-making usually solves problems from a technical perspective; however, the change in the fundamental perspective of solving decision-making problems may bring significant, socially desirable improvements to the sustainable infrastructure lifecycle. Regarding this point, Bakker, Volwerk, and Verlaan [143] discussed the decision-making frameworks based on economic principles and methodologies, and the authors pointed out that allocated reasonable budgets may hinder the implementation of economic investment strategies; thus, it is necessary to understand the relationship between decision-making options resulting from different perspectives and incorporating them into deliberations. Furthermore, the design of SHM-supported decision-making frameworks needs to be comprehensive and detailed in its consideration of matters at the level of practical implementation. Most of the literature emphasized the issue of time delay and human preference. For the time factor, Orcesi and Frangopol [134] incorporated the time delay between the assessment and intervention schedule into the SHM-supported bridge life-cycle management decision framework, allowing stakeholders to plan management strategies using a long-term time horizon. For the human preference factor, Verzobio et al. [144] mathematically described the biased judgments of decision-makers in the SHM-based decision-making process by constructing three models to represent distorted individual behaviors, which are critical in realistically formulating infrastructure maintenance policies. The study by Valkonen and Glisic [145] creates a risk appetite assessment tool based on a domain-specific risk-taking scale to assess differences in stakeholders’ attitudes toward risk, since such differences may make the negative value of the SHM system [146]. Based on the description and quantification of human preferences in the above studies, Chadha et al. [147] proposed a methodology to accommodate and mitigate the risk associated with various factors. This involves considering the risk profile of the decision maker’s behavior, combined with the decision maker’s acceptable risk intensity. The approach incorporates a risk profile model for maintenance decisions based on a set of predefined choices, leading to incorporate a layer of human psychology into the design of the decision-making framework.

6. Sustainable Civil Infrastructure SHM Influencing Factor Correlation Scheme

Based on the comprehensive review of the previous literature, this section establishes a scheme through selective coding to describe the relationships among influencing factors at the implementation level of SHM in sustainable civil infrastructure. The factors may impact not only directly relevant activities but also activities in different phases. The proposed scheme can promote stakeholders to make forward-looking or iterative decisions and behaviors during the implementation process, which may reduce rework and improve resource efficiency. As shown in Figure 2, sub-categories of factors directly affecting the activities identified in the previous sections and encapsulated within rectangular boxes representing the corresponding activities. Different colors are used to classify influencing factors in the same activity. For overall consistency, the influencing factors mentioned in this section are shown as labels in Figure 2 with the corresponding indexes (i.e., F1–F139) in the tables of Appendix A. Indexes mentioned in Figure 2 are explained in

Table 6. List of acronyms of influencing factors involved in Figure 2.

Index (Acronym)	Influencing Factor
F17	Decision-making on the scope of monitoring
F21	Knowledge of wired and wireless monitoring networks and the characteristics of the related devices
F22	Organizational learning about characteristics of vision-based and vibration-based sensors
F27	Validity of the information provided by the technology (environmental interference)
F28	Adaptability, flexibility, and scalability of sensors
F41	Introducing one or more different technologies or concepts and planning for the most resource-efficient implementation scheme
F42	Applications in real-world scenarios show less accuracy than experimental scenarios with small-scale benchmark structures
F43	Collaboration of interdisciplinary organizations and consultants
F44	Integration of intelligent sensing and data processing functions, technologies, and components
F46	Emphasize the needs of administrators and end-users as well as the flexibility and availability of technology
F48	Orderly organization of the work of documentation in the field
F55	Data flow issue in sensor failure detection
F56	Trade-off between redundancy scope of sensor network and total cost
F59	Adoption of time synchronization error resilience algorithm
F67	Damage-sensitive features extracted from the data are at risk of being masked by the data normalization process
F72	Considerations related to the location of data processing execution
F78	Design and application of dynamic code migration strategies
F88	Combining robust signal processing techniques with a priori statistics and engineering judgment
F100	The suitable features usually come from some combination of information obtained from several sources
F101	The accuracy of the initial finite element model (FEM)
F108	Size and heterogeneity of datasets used for training in machine learning
F109	Automated selection and feedback of damage-sensitive features
F110	Improving machine learning algorithm performance with effective feature libraries
F119	The available input data time period is long enough
F121	The high quality of input damage detection data
F122	Diversity and validity of types and sources of structural information for judgments

. Arrows without factor labels represent the overall impact of sub-categories instead of specific factors inside. Arrows point from sub-categories/activities of influencing factors to sub-categories/activities being influenced by these factors. Because the direct Influence of sub-categories on their associated activities is easy to be understood, this section mainly focuses on the explanations of the relationships between cross-activity influencing factors.

Specifically, in the process of P2A1, the decision on monitoring strategies (F17) and main sensor network types (F21) influence the design of data communication schemes through decisions related to monitoring range and the choice between wired or wireless sensor networks, respectively. Understanding the characteristics of sensors with different fundamental properties and principles (F22) is crucial for leveraging their strengths during the selection of primary sensor network types in the P2A2 process. For instance, when vision-based sensors are used to measure the infrastructure with a suspended cable structure, small lights and other targets are often placed on the structure. This on-site measurement method eliminates the advantages of cameras as non-contact sensors [148]. The communication of information and knowledge about sensor strengths, weaknesses, and other characteristics between SHM system designers, sensor suppliers, and on-site installation and measurement personnel can help the selected sensor maximize its resource advantages, as sensors with certain performance advantages often incur significant economic costs [75,149]. In addition, when making decisions on objective indicators representing sensor performance, the effectiveness of the information provided by sensors (F27) often directly affects the workload of data cleaning and normalization in the P2A3 process. Similarly, the fault-tolerant design in the overall process of sensor placement of P2A1 and on-site calibration and self-monitoring of sensors (F56) in P2A2 affect and balance each other at the demand and workload levels [150]. Moreover, sensor network innovations and advanced technologies often directly influence the design requirement of raw data preprocessing, data transmission, and data storage. Because intelligent sensing technology tends to integrate functions (F44), well-designed intelligent solutions (F41) often require the coordination of multiple activities [76]. The intelligence of solutions is also reflected in the black-box nature and the ultimate user-friendly experience (F46). This is particularly evident in data storage, significantly affecting the intelligence requirements of database management system designs.

In the feature extraction activity (P3A1) of the data assessment process, design of feature extraction strategies involves understanding the structural information of the actual project, sensing technology, and site conditions, making its implementation dependent on the completeness of factors related to pre-assessment and on-site documentation (F48). Similarly, the emphasis on combination of robust signal processing and prior engineering judgments (F88) also derives from the results of pre-assessment work [96]. Furthermore, a contribution of the feature extraction process to the overall sensors network synchronization problem (P2A2) is allowing the use of asynchronous data for modal parameter identification through certain algorithms (P3A1) (F59) [151]. As for P3A2, both physical and data-based damage detection methods involve the process of feature selection from extracted features, which is closely linked to cross-activity influencing factors. In physical-based detection methods, this process is based on information obtained through multiple channels (F100). In addition to actual degradation experiments, the richness and diversity of information obtained from sensor networks in P2A1 activities also make significant contributions. For data-based detection methods, the feature selection process mainly relies on the automatic selection process of adaptive enhancement algorithms, forming an excellent feature library that can in turn train the feature extraction layer (P3A1) to directly filter out invalid features and inspire P2A1 sensor network design to measure more effective information (F109). Furthermore, F109 can inspire the iterative design for sensor networks in P2A1 based on flexibility and scalability of sensor networks. As for damage diagnosis process in P3A2, the accuracy is closely related to cross-activity factors. The accuracy of physics-based methods is largely influenced by the accuracy of the FEM (F101), which is closely related to the quantity and quality of construction project information provided. Therefore, the results of pre-assessment work are crucial [58]. However, the accuracy of data-based methods depends on the quality of machine learning, which depends on the large-scale heterogeneous dataset used for training (F108) and the performance of algorithms based on effective feature pools (establishing a pool containing a large number of potential good features, in which the algorithm automatically searches for the most suitable features for a specific task [114] (F110)). The former benefits from the achievements of sensor network design in P2A1, and the latter benefits from the feature extraction process and forms a virtuous cycle with the factor of F109.

Finally, the role of influencing factors in the prognosis and decision-making stage (P4) is less affected by cross-activity factors compared to other stages. In the development and application of life-cycle models (P4A1), the factors that are susceptible to cross-activity factors are mainly related to the basis used to predict the subsequent life-cycle performance changes of structures, including damage detection results data and structural aging parameters. On the one hand, the time range of damage detection result data (F119) is related to the long-term archiving ability of data storage (P2A5); the quality of damage detection data (F121) depends on the diversity and accuracy of damage detection P3A2 results. On the other hand, the optimal structural aging parameters should be diversity and effectiveness (F122), which also depend on these two characteristics of P3A2 results [135]. Subsequently, the formation of P4A1 prognostic outcomes (including life-cycle curves and risk-rating systems, etc.) often forms the foundation of P4A2 (decision-making activities), and the diversity of information it contains promotes more reasonable decision-making and smoother and lower-cost decision-making processes.

7. Discussion

In this review, three research objectives set in Section 1 have been progressively accomplished; firstly, according to the comprehensive review of the existing literature, the involved implementation steps of SHM in construction projects have been divided into several phases and the corresponding activities summarized, resulting in the SHM implementation framework. Then, based on the proposed framework, a literature review of the SHM sub-categories has been performed to comprehensively analyze the influencing factors during each logical stage of SHM implementation for sustainable civil infrastructure. Finally, taking the activities as the bridge, the relationships among different influencing factors have been examined; consequently, an influencing factor relationship scheme of SHM in sustainable civil infrastructure has been proposed based on the opinions of the existing literature. The findings and outcomes will be discussed in the subsequent sub-section.

7.1. SHM Implementation Framework

The management of the SHM process is not a novel concept. The previous literature often delineates SHM’s implementation stages, modules, and steps based on varied logical grounds posited by different authors or specific project situations. The distinctions between the outcomes of this review and the previous research can be illustrated in two aspects. Firstly, two aspects have been considered in this review, one is comprehensiveness, another is to avoid the task overlaps induced by inconsistency in specialized terminology standards. Based on two considerations, and regarding the division of SHM stages and implementation steps, opinions from previous literature have been integrated, instead of emphasizing the design of SHM implementation processes. A similar endeavor can be observed in Wong [53], who introduced a modular concept, segmenting large-span bridge SHM systems into six integrated modules and two phases. In the classification of influencing factors, modular method is also adopted, rather than categorizing based directly on related activities, promoting more comprehensive reviews and valuable findings. However, a distinction from Wong [53] is that it partitioned the monitoring and evaluation processes into real-time and non-real-time, and such an idea and criteria could not be suitable for the current practical situations based on the findings of our comprehensive literature review. For instance, advancements in smart sensing systems have paved the way for on-site, real-time data feature extraction and even damage monitoring processes in infrastructure SHM [152], achieving energy-saving goals by reducing data transmissions. Moreover, the basic idea involved division of SHM in the other literature can be summarized as modularization based on item type, phasing based on logical sequences, or a combination of both.

Secondly, mainstream practices for managing the SHM implementation process generally only stay on the levels of phase and module. There is no systematic exploration of the involved specific activities. This poor situation could be owed to the diversity of technical and organizational aspects involved in SHM, making it difficult to describe the specific activities involved in the process through uniform terminology, standards, and processes [23].

7.2. Influencing Factors

It should be pointed out that the analysis of influencing factors in SHM implementation of sustainable civil infrastructure still stays in the initial stage. Referring to the general classification method for influencing factors in project management, this review aims at the influencing factors’ properties to conduct the axial coding. The results show that the technical factor accounts for most percentages, followed by the organizational factor. Technical factors may be with the greatest potential to contribute to sustainable civil infrastructure SHM, which is consistent with the mainstream view in the current literature, which posits SHM as a technology-driven field. Nevertheless, organizational factors, decision-making factors, and important consideration factors also encompass a significant number of influencing factors. Particularly, important consideration factors are often critical factors in result orientation. The realization of these factors relies on endeavors directed away from technical fronts. It is worth mentioning that, based on the combined findings of different literatures, when a factor serves as both a technical and organizational, information and knowledge transfer and/or personnel training in organizations may be a suggested direction to make efforts. Consequently, the implementation of SHM in sustainable civil infrastructure may necessitate a collaborative advance from research, education, and technological demonstrations. This implies effective integration and coordination of multi-stakeholder groups from the academic community, industry associations, and government [23]. Regarding this point, Farrar and Worden [3] also emphasized that multidisciplinary research efforts are needed for SHM, and few attempts have been made to integrate these techniques with a specific focus on developing SHM solutions.

7.3. Influencing Factor Relationship Scheme of SHM in Sustainable Civil Infrastructure

The mainstream method for research on influencing factors in construction projects is grounded on the results of questionnaires and workshops, and then conducting SNA, whose validity of data is constrained by the professional knowledge of the respondents [153]. However, this review, based on findings from the literature review, employs selective coding to form a scheme on the relationships of influencing factors, holding promise in overcoming the limitations of the aforementioned mainstream approach in terms of data quality.

The derived theoretical outcomes emphasize insights in two main areas. On one hand, the interactions between sensor deployment (P2A1) and other tasks might be among the most prevalent across all activities. Consequently, it is crucial for the strategic design of sensor deployment to actively take into account the subsequent tasks involved, and influencing factors associated with these tasks should be considered in advance. For instance, in practice, attention should be given to the synergy between decisions making during sensor network design and on-site installations in the activity of P2A2. Meanwhile, the flexibility and scalability of the sensor network become crucial due to the demands of iterative design implementation. For example, wireless contact-free sensors might cater to these specific requirements. Furthermore, innovative technologies in sensor networking might fundamentally alter strategies for subsequent tasks, and then, the completion assurance of these subsequent tasks should be emphasized when the technologies are adopted.

On the other hand, the environmental sustainability of the SHM implementation process significantly relies on endeavors in data transmission, because it is not only a significant source of energy consumption but also offers good potential for energy conservation. Energy consumption or conservation is achieved through harmonization with the sensor network and factors related to data preprocessing. Thus, it is suggested that data management at the sensor level might be imperative as a distinct module for designing energy strategies. This is consistent with the views presented by Boyle et al. [93], who advocate for the decoupling of radio frequencies when the optimization of energy strategies is performed.

8. Conclusions

The main outcomes of this review include: (1) an SHM implementation framework, (2) synthesis tables of influencing factors in the four phases of the framework (see Appendix A), and (3) a scheme of influencing factor relationships for sustainable civil infrastructure SHM. These outcomes sequentially fulfill the three research goals of this review. The main conclusions of the synthesis are summarized as follows: (1) the pre-assessment phase of SHM is an important but an easily overlooked matter; (2) sensor work scheduling and data transmission are promising endeavors for balancing economic and environmental sustainability, while social sustainability is mainly in terms of security and user experience; (3) the success of SHM for sustainable civil infrastructures requires concerted efforts on the technical and organizational fronts; and (4) since the influencing factors of different phases may interact with each other, the implementation process should emphasize forward-looking and holistic thinking.

However, despite the range and reliability of data sources that have been improved compared to the mainstream methods, some limitations also exist in this review. On the one hand, qualitative research cannot determine the level of importance of different kinds of factors and the ranking of the likelihood that it happens in practice but can only make judgments about its potential to contribute to sustainable SHM implementation. This is because the number of influencing factors covered by the category cannot represent the importance, but only indicates that there are more ways for it to contribute, i.e., more potential directions of endeavor could be investigated based on it. On the other hand, compared to the field of construction project management, there is very little literature that directly examines the influencing factors of SHM implementation process. Regarding this point, it may result in the comprehensiveness of the identified influencing factors being limited by the size of the literature data.

There exist some limitations, but this literature review is of some significance. On the one hand, the review demonstrates the current state-of-the-art SHM data processing, damage identification and prognostic decision-making techniques, and how they can be applied in order to achieve the three dimensions of sustainability. On the other hand, the review proposed a more detailed scheme in terms of management by identifying and integrating different ways of categorizing and semantically expressing SHM processes and factors in the existing literature, thus contributing to the successful implementation of SHM for sustainable civil infrastructures.

Based on an analysis of the current situation, the findings provide an inspiration for future research directions. For example, on the technical side, one of the research directions suggested is the development of techniques to improve the accuracy of identifying and differentiating data anomalies, and the integration of intelligent techniques, concepts, and functions of sensor networks are identified as a dominant trend in sustainable infrastructure SHM research. At the non-technical level, it suggests simultaneously considering environmental and economic aspects of sustainability through research on the topic of scheduling strategies; also, the need for knowledge and information transfer within the organization during SHM implementation inspires the integration of the organizational design and the SHM process management in the research field.