1. Introduction
In recent years, the thriving advancement of infrastructure construction has significantly bolstered China’s economy, evolving into a fundamental pillar of its economic structure. Correspondingly, safety incidents in construction projects have garnered increased scrutiny and have emerged as a prominent societal issue [
1]. Nearshore tunnels provide distinct advantages over alternative cross-sea transportation methods due to their expediency, speed, minimal environmental footprint, and high traffic volume. However, the enduring submersion of tunnel structures in seawater exposes them to high water pressure and creates substantial technical challenges during construction. The intricate geological conditions, coupled with potential encounters with fault zones and weathered deep grooves, can lead to water ingress and seepage, thereby introducing significant safety risks [
2]. Consequently, it becomes imperative to scrutinize and analyze safety risk factors in the construction of nearshore tunnels. This helps enhance nearshore tunnel construction technology, optimizes the construction management process, minimizes accident-induced losses, and provides robust foundations for future tunnel selection, design, and construction [
3].
Existing construction safety risk assessment practices suffer from three primary shortcomings. First, reliance on expert analysis and scoring introduces subjectivity, leading to reduced accuracy in evaluation outcomes [
4,
5,
6]. Second, conducting risk assessments prior to construction prevents on-site operators from proposing timely, appropriate risk avoidance measures based on actual engineering issues. Lastly, the unique risk indicator systems of different construction environments hinder the applicability of insights from one project to another [
1,
7,
8,
9,
10,
11]. Implementing multi-source heterogeneous data fusion technology in underground engineering construction safety risk assessment can substantially enhance the precision of evaluation results.
Owing to the intricate environments in underground engineering and tunnel construction, there are numerous hazardous sources throughout the construction process, necessitating the establishment of additional monitoring points. Consequently, the project has to manage a significant volume of real-time monitoring data. Currently, manual monitoring, despite its inefficiencies and delayed risk warning capabilities, remains the primary method for overseeing underground engineering and tunnel construction. This also augments the likelihood of risk assessment and early warning inaccuracies due to human error [
12]. Utilizing building information modeling (BIM) technology, a 3D model can be constructed and analyzed using 3D roaming, animation demonstrations, and simulation construction. This allows for the timely identification of potential hazards during construction based on safety risk factors. The BIM model can facilitate intelligent detection of monitoring points in the construction area, precise positioning of safety risk zones during the monitoring process, and implementation of construction safety risk controls. Marking of unsafe locations permits digital management of the construction site [
13,
14,
15,
16]. Integrating BIM technology into safety risk management of underground engineering construction can effectively mitigate issues such as delayed data collection, transmission, and analysis, and non-intuitive and delayed risk warnings. Collins et al. [
17] examined the development of an underground construction safety risk early warning system based on BIM and Internet of Things (IoT) technologies. This study established a comprehensive early warning and control system that enables real-time dynamic monitoring by consistently recording and assessing process metrics via a BIM management platform. Ding et al. [
18] integrated BIM with semantic web technology to create a construction safety risk management framework within a BIM environment and elaborated on the entire workflow of construction safety risk management, encompassing risk factor identification, risk path analysis, and risk mitigation strategies. Similarly, Lou et al. [
19] investigated the construction of an urban complex project, combining BIM and augmented reality (AR) technologies. They realized real-time dynamic monitoring and control system by regularly capturing and assessing key performance indicators through the BIM management platform. Moreover, AR technology was employed across three distinct phases—pre-accident, during, and post-accident—to bolster construction quality and productivity. Lu et al. [
20] developed a software plug-in bridging BIM technology with safety risk data, which can automatically calculate building safety risks and help architects and structural designers quickly select the design scheme and verify the feasibility of the method through case studies.
In addition to these technological innovations, resilience plays a significant role in safety risk assessment. It refers to a system or organization’s capacity to anticipate, absorb, adapt, and recover from unforeseen disruptions or events. Within the scope of safety risk assessment, resilience serves to identify system vulnerabilities, mitigate risks, and adapt to evolving conditions. It also aids in planning for recovery and in nurturing a culture of continuous learning to refine safety protocols. According to Cimellaro et al. [
21], the concept of catastrophe resilience is viewed as a synthesis of knowledge from organizational and technological disciplines, ranging from social sciences and economics to seismology and earthquake engineering. Numerous presumptions and interpretations are necessary for the study of catastrophe resilience. However, the ultimate goal is to combine data from multiple fields into a single framework, resulting in a free of erroneous assumptions or preconceived concepts of risk. The authors also provided a framework that was based on Californian hospital structures and provided an in-depth explanation and implementation of a streamlined recovery model. Within the building system itself, as well as the losses suffered by the people the system serves, this model considers both direct and indirect losses. Aven [
22] made a connection between resilience and risk, positing that risk assessments could offer valuable insights into resilience evaluations, particularly by accounting for the uncertainty of potential disruptions. Yang et al. [
23] delved into the quantitative facets of resilience, proposing a triple resilience definition framework founded on perturbations, functionality, and performance. This framework also accommodates the handling of uncertainties. Resilience emphasizes a system’s ability to predict, absorb, adapt, and recover from disruptive situations, providing a significant concept that encompasses reliability and risk-based thinking to guarantee the safety of these complex systems. Guo et al. [
24] applied the resilience theory to the safety management of three subway construction sites through analyzing the resilience essence and assessing the system’s resilience using a resilience index. They employed cloud and element extension theories to establish an analytic network process (ANP) extended the cloud comprehensive model, aiming to tackle the inherent randomness and fuzziness encountered during resilience assessments at these sites. San-gaki et al. [
25] established a probabilistic framework and model for determining the probability distribution of earthquake recovery indices in order to account for various uncertainties and produce recovery curves. Additionally, they suggested a probability model that was consistent with dependability techniques and included the ground motion intensity of earthquakes, building responses, structural damage, loss of functioning of buildings, recovery, and resilience. By creating elastic response curves for a typical four-story reinforced concrete moment-resisting frame building and contrasting the findings with those using conventional techniques, the authors proved the validity of their framework and model. Forcellini et al. [
26] employed a probability-based vulnerability curve approach to estimate losses and introduced a novel definition of recovery models. The study also discussed the application of the proposed framework through case studies in both fixed-base and base-isolated structural systems. The existing body of research has made substantial contributions to exploring the advantages and applications of resilience in disaster management. Some studies have focused on the concept and definition of resilience, as well as its practical implementation in risk management. Others have delved into specific domains such as natural disasters and supply chain management, examining the impact of resilience on various types of risks and corresponding mitigation strategies. These studies facilitate an enhanced understanding and application of resilience concepts in the relevant fields and the evaluation of their practical benefits in diverse settings.
In this paper, multi-source information fusion technology enables the integration of on-site data, design data, and environmental data to ascertain the dynamic safety risk level of underground engineering construction. Through Revit secondary development technology, the risk level can be real-time warned on the BIM model, thereby improving the efficiency of underground engineering construction safety risk management. The feasibility of this approach is validated using a specific nearshore tunnel in Ningbo as an example. Research findings confirm that this approach can identify and assess risks throughout the engineering process, provide early warnings, and prevent accidents during nearshore tunnel construction. This study addresses critical limitations in traditional construction safety risk assessment methods, such as the reliance on expert subjectivity and the underutilization of real-time monitoring data. By employing data fusion technology, we obtain a dynamic safety risk profile at specific monitoring points, offering a more accurate depiction of safety risk levels on the construction site. Specifically, the data fusion approach collects data from various sources and types, yielding a more holistic view of safety risks. This facilitates accurate predictive analyses by uncovering hidden correlations and patterns, thereby reducing bias and errors.
Our proposed real-time monitoring and early warning visualization methodology, based on BIM, enhances the practical utility of monitoring data within BIM platforms. This tackles issues such as the lack of intuitive early warning systems and delays in early warnings. The approach significantly augments the efficiency of safety management throughout the construction process. It also provides a robust foundation for construction safety management that benefits all project stakeholders, diminishing the likelihood of safety incidents and elevating the project’s overall safety standards.
6. Discussion
In today’s increasingly specialized, mechanized, and integrated construction sector, the critical issue of how to enhance construction precision and avoid recurrent safety incidents urgently demands resolution. Risk and hazard monitoring and assessment have always been at the forefront of the construction industry. Relying solely on manual labor for construction safety risk evaluation and management proves costly, inefficient, and is susceptible to errors or oversights. BIM technology, renowned for its digital and visual capabilities, has been extensively deployed across the entire life cycle of design, construction, operation, and other stages in recent years. Nevertheless, its application in construction monitoring and early warning remains under-researched.
To further enhance safety management during construction and increase efficiency, this paper proposes an intelligent evaluation and real-time warning system of safety risks in a nearshore tunnel construction based on BIM, considering the current application status in practical engineering. Taking a specific underwater tunnel in Ningbo as an example, we verified the feasibility and applicability of the evaluation method and software. The primary research outcomes of this article are as follows:
Assuming the comparison of six methods for measuring the degree of conflict, namely, conflict factor K, cosine value of the included angle, Pignistic probability distance, Josselme distance, BJS divergence, and confidence Hellinger distance, it has been concluded that the confidence Hellinger distance provides the most accurate reflection of the conflict measurement value between two evidence chains. Through calculations involving the degree of conflict, credibility, and evidence weight, the conflict monitoring data chain undergoes modification. Subsequently, the modified evidence chain is fused using the Dempster rule. When a high conflict monitoring data chain is present, the improved DS evidence theory method can reasonably allocate conflicts, overcoming traditional method limitations and achieving more accurate fusion results.
Using the data from five monitoring points of a nearshore tunnel in Ningbo as an example, we conduct a dynamic safety risk assessment of construction on the monitoring data. The results demonstrate that monitoring point K7+860 has a risk assessment level of II, while the other monitoring points are in a normal state. The dynamic safety risk level of monitoring points, derived from construction monitoring data through data fusion technology, offers a better reflection of the construction site’s safety risk level compared to monitoring scattered data. Based on the safety risk assessment level of monitoring point K7+860, the data fusion-based safety risk assessment process is embedded into a computer through the C sharp programming language to achieve automatic monitoring data fusion functionality. The backend transmits the fusion results to the sensors in the BIM model. These sensors, in turn, utilize the fusion results to determine the dynamic safety risk level of the respective monitoring points. If the risk level exceeds the warning limit, the system will automatically sound an alarm and guide engineering personnel to address the situation.