Intelligent Health Monitoring of Cable Network Structures Based on Fusion of Twin Simulation and Sensory Data

Abstract

The precise and effective prognosis of safety risks is vital to ensure structural safety. This study proposed an intelligent method for the health monitoring of cable network structures, based on the fusion of twin simulation and sensory data. Firstly, the authors have established a framework that integrate simulation data with sensory data. The authors have established a high-fidelity twin model using genetic algorithm. The mechanical parameters of the structures were obtained based on the twin model. The key components of the structure are captured by using Bayesian probability formula and multiple mechanical parameters. The fusion mechanism of twin simulation and random forest (RF) was established to capture the key influencing factors. The coupling relationship between structural safety state and key factors was obtained, and the safety maintenance mechanism was finally formed. In view of the risk prognosis of the structure, the establishment method for the database of influencing factors and maintenance measures was formed. The authors used the Speed Skating Gymnasium of 2022 Winter Olympic Games (symmetric structure) as the case study for validating the feasibility and effectiveness of the proposed method. The theoretical method formed in this study has been applied to the symmetric structure, which provides ideas for the safety maintenance of large symmetric structures. Meanwhile, this research method also provides a reference for the health monitoring of asymmetric structures.

1. Introduction

Cable net structures have been widely used in large public buildings, due to their advantages of large space, light weight, reasonable force, diversified structural forms, and fast construction speed [1]. Once an accident occurs, it will cause huge casualties, property losses, and adverse social impacts [2,3]. Therefore, it is particularly important to detect the risks affecting structural safety in time and realize the predictive maintenance of structural safety. In the normal service period of cable net structures, the limit states are mainly divided into four categories, that is, cable relaxation, cable breaking, deformation exceeding limit, and strut instability. In the failure mode of the structure, the relaxation and corrosion of the cable members are the main factors leading to the safety accidents of the structure [4].

With the technological development of the construction industry, the research on the safety performance of large-span spatial structures has become a hot spot. In building structures, the establishment of a damage identification model is an important tool to maintain structural safety [5]. Aiming at the mechanical response of long-span space structure under temperature and construction errors, many analytical models and theoretical methods have been formed and have effectively guided the engineering practice. Liu et al. [6] studied the mechanical performance of hub-spoke cable truss structure under the action of cable failure, conducted a reliable evaluation of the structure, and finally, captured the key stress components that affect the safety of the structure. Ruan et al. [7] determined the reasonable error control index by analyzing the influence characteristics of cable length error and external node coordinate error on cable force. In order to improve the stability of the reticulated shell structures, Zhang et al. [8] proposed a new cable layout scheme and pointed out that the best length of strut should be ensured in practical engineering. Rizzuto et al. [9] conducted an experimental study on a honeycomb grid-connected reticulated shell structure. The study found that the stiffness of the components near the bearing has a great influence on the overall stiffness of the structures. Arezki et al. [10] studied the influence of temperature change on cable truss structures and cable safety performance.

Previous studies have performed extensive research for examining the influence of temperature effect and error on structural safety. However, for the integration of various influencing factors, there are few studies on the comprehensive analysis and prediction of the structural safety status. The virtual–real interaction cannot be achieved. According to literature research and engineering practice, the traditional structural health monitoring system has strong dependence on sensing equipment, which leads to the high cost of the safety risk assessment and great lag [11,12]. Thus, the intelligent safety closed-loop maintenance on the structures could hardly be achieved [13]. There is a lack of key factors and corresponding maintenance measure analyses methods for large-span spatial structures. Therefore, the risk prognosis of structural safety cannot be achieved. The traditional structure health monitoring uses a lot of acquisition equipment, which has the problems of high cost and large error [14]. Therefore, in order to solve the above problems, this study conducted the following exploration. A high-fidelity twin model of structure is established based on perceptual data. In this model, the key stress components of the structure are analyzed and the key influencing factors are obtained to reduce the calculation time of safety assessment. In the twin model, the same working conditions are arranged in the key stressed members, based on the actual load of the structure, so as to obtain the degree of relaxation and corrosion of the structure. Finally, maintenance measures of the structure are formed according to two types of factors, and closed-loop control from risk capture to accurate maintenance is realized. Driven by the theoretical method, the twin platform of structural health monitoring is formed. The application of the platform can effectively maintain the safety of the structure and reduce the monitoring cost of the actual structure. Therefore, the goal of this study is to realize the intelligent operation and maintenance of the structure [15]. In this study, digital twins (DTs), artificial intelligence, and remote sensing technology play a key role [16,17,18].

In the current Industry 4.0 scenario, industries are developing customized, cost-effective processes to meet customer needs with the help of the DTs framework, which enables users to monitor, simulate, control, optimize, and identify defects and trends in ongoing processes and reduce the possibility of human error [19]. In the Industry 4.0 paradigm, technological advances in networked physical systems (CPS) and steady improvements in the intelligent manufacturing framework have given birth to the concept of a “DTs” [20]. For structural health monitoring, DTs play an important role in mechanical property evaluation and the parameter analysis of structures. Accurately predicting the service life of structures and equipment plays a crucial role in the predictive maintenance field of Industry 4.0 [21]. The high-fidelity virtual model based on DTs is established to map the real state of the structures. DTs is the link between the physical world and virtual space, which provides the basis for intelligent maintenance of structural safety. [22]. In the safety maintenance of structures, artificial intelligence provides a method for the intelligent analysis and processing of data [23]. In the development of the construction industry, intelligent construction is a hot research topic. For building structures, how to realize intelligent maintenance of structural safety is a difficult problem in civil engineering [24,25]. The integration of DTs and artificial intelligence can effectively improve the accuracy and efficiency of structural safety maintenance. [26]. Singh et al. [27] proposed the use of frequency response function and finite element model updating technology for health monitoring, in order to identify structural damage during normal service period. The results show that the updated model can accurately predict the dynamic response of buildings. Revetria et al. [28] proposed a system that uses augmented reality, the Internet of Things, DTs, and simulation to monitor stress throughout the life cycle. Tahmasebinia et al. [29] analyzed the influence of creep and shrinkage on Sydney Opera House by creating DTs through integrated BIM and finite element model. Lin et al. [30] proposed a method to determine the seismic collapse capacity of Bridges and the collapse vulnerability assessment of vulnerable areas by using DTs. The DTs model can predict the intensity and probability of collapse. Liu et al. [31] proposed a dynamic guidance method for fire evacuation by integrating DTs model with the Dijkstra algorithm for the intelligent security of buildings. Pan et al. [32] integrated DTs and artificial intelligence to model, predict, and optimize problems in the whole life cycle of actual complex projects in a data-driven manner. Lu et al. [33] integrated DTs, machine learning, and data analysis to create a simulation model. The model can accurately map the state of physical space. The proposed method effectively promotes the implementation and development of smart cities. Ma et al. [34] studied the feasibility of using the synthetic point cloud generated from the DTs model to train the deep neural network for the semantic segmentation of the point cloud inside the building. This method can be applied to the completion modeling of buildings containing invisible indoor structures, which improves the intelligent level of building delivery. Acharya et al. [35] proposed a visual positioning method to achieve the real-time and accurate positioning of indoor buildings. The 3D indoor model is used to eliminate the image-based indoor environment reconstruction requirements, and the deep convolution neural network is fused to finetune the image.

Sensory data plays an important role in structural health monitoring. Graves et al. [36] carried out the monitoring of bridge deformation based on remote sensing technology, which provided a reference for health monitoring of other structural types. Zhu et al. [37] developed the helical deployment scheme of distributed fiber optic sensors, which effectively reduced the prestress loss of the structures. Fu et al. [38] collected frequency data and deformation data by sensing technologies (STs) to achieve an accurate assessment of bridge structural safety. Wang et al. [39] developed a machine learning-derived two-stage method for post-earthquake building location and damage assessment, considering the data characteristics of satellite remote sensing (SRS) optical images with dense distribution, small size, and imbalanced numbers. This method has a strong application prospect in the field of disaster prevention and the reduction of buildings. It was found that remote sensing technology, mainly based on image recognition and sensing equipment, provides an important basis for structural health monitoring [40,41]. Perception technology can accurately capture the actual state of the structure and effectively ensure the accuracy of safety risk prediction.

In the aspect of structural health monitoring, how to realize the real-time analysis of safety risks and the accurate implementation of maintenance measures to ensure structural safety is the focus of this study. Combined with the role of DTs and sensory data in the construction industry, this paper puts forward an intelligent health monitoring method of structural safety driven by fusion of twin simulation and sensory data. The fusion mechanism of twin simulation and sensory data is clarified. Driven by sensory data, the twin model can map the real state of the structure more accurately. The sensory data of real structures and simulation data of virtual model form twin data representing structural safety. Pressure gauge, total station, and displacement meter are used to collect the mechanical response of the structure in practice. According to the mechanical change characteristics of the structures, a theoretical model of structural risk prognosis is established. On this basis, a numerical model of risk prognosis is developed. The model includes the establishment of twin model, the analysis of structural safety performance, and the maintenance of structural safety performance. The genetic algorithm is used to iterate the physical parameters to ensure the fidelity of the twin model. The physical parameters are iterated by genetic algorithm to ensure fidelity of model. The working conditions are set in the twin model to obtain the mechanical parameters representing the structural safety performance. The Bayesian probability formula is integrated to analyze the structural safety performance. The key stress components in the structure are obtained. The structural response caused by relaxation and corrosion of key components is analyzed. The fusion mechanism of DTs and random forest (RF) is established to obtain the key influencing factors. The coupling relationship between structural displacement and key factors is obtained. Based on the coupling relationship, efficient maintenance measures are formed to realize the closed-loop control of structural safety. Driven by the theoretical framework, three modules that drive the implementation of intelligent maintenance method are explored. Finally, the theoretical method formed is applied to the structure of the speed skating gymnasium of the 2022 Winter Olympics, and the twin platform of intelligent analysis and maintenance of the structure is established, which effectively improves the information and intelligent level of security maintenance.

The organization of the paper is as follows. Section 2 introduces the proposed predictive maintain framework for safe and effective structure. The proposed framework takes sensory data that captures the dynamic behaviors of structures as input for driving the twin model for risk prediction and maintain it during the service period. Section 2 puts forward the theoretical method of this study, which integrates the sensory data into the twin model. On the one hand, it can correct the twin model and improve its fidelity. On the other hand, the establishment of the twin model can reduce the dependence on structural acquisition equipment. This chapter also clarifies the main process of structural risk prediction.

Section 3 sets up the numerical model of intelligent health monitoring of structural safety driven by fusion of twin simulation and sensory data. The model includes the establishment of the twin model, the analysis of structural safety performance, and the maintenance of structural safety performance. In this section, the theoretical method of Section 2 is further explained, and three key implementation steps are formed. The establishment of numerical model provides method support for the engineering application is in Section 4.

Section 4 illustrates a case study for validating the proposed DTs-based intelligent monitoring method for safety assurance during the cable net structure service period. In this section, the intelligent analysis and prediction of structural safety is carried out by establishing the twin model. The key stress members are obtained, and the key working conditions affecting the structural safety are analyzed. The coupling relationship between the key factors and mechanical parameters is fitted. Finally, the structural safety maintenance measures and intelligent health monitoring cloud platform are formed. The theoretical method is applied in the platform. On the one hand, the accurate and rapid maintenance of structural safety maintenance is realized, and, on the other hand, the dependence on sensing equipment is reduced, and the problems of high cost and hysteresis of traditional structural health monitoring are avoided.

Section 5 concludes and shows future research directions.

2. Framework for Intelligent Monitoring of Structural Safety Driven by DTs and Sensory Data

During the service period, the large-scale structure systems (e.g., cable net structures) have the characteristics of large deformation and large displacement. The probability of the safety risk is high at this stage [42]. It has become an important scientific problem to realize the risk prognosis of structures. There are many uncertain factors in the analysis of the structural safety state, which should refer to the idea of elastic time mechanics. The geometric parameters, physical parameters, and boundary parameters are all functions of time [43,44]. Information technologies, such as DTs integration of the Internet of Things and artificial intelligence, fully consider the spatiotemporal evolution characteristics of the structures to achieve high-fidelity mapping of the structures [45].

2.1. Fusion Mechanism of Twin Simulation and Sensory Data

For the risk prognosis of structures during service life, twin simulation and sensory data are integrated. On the one hand, the high-fidelity twin model of the structure is established by the finite element software. The mechanical parameters of the structures are extracted in real time from the twin model, and the safety state of the structure is visualized. On the other hand, the mechanical parameters of the structures are collected in real time by STs, such as the cable force sensor. The mechanical parameters of the structures are collected to modify the twin model to ensure the rationality of the structural safety analysis. The simulation data and sensing data form twin data representing structural safety. Driven by the twin data, the safety performance of the structure is analyzed to obtain the key stress members. The key factors affecting the safety state of the structure are analyzed based on the key stress members. According to the key factors, the maintenance measures of structural safety are formed to effectively ensure the safety of the structures. The fusion mechanism of the twin simulation and sensory data for structure intelligent health monitoring is shown in Figure 1. In this study, the perception technology and twin simulation technology are integrated to realize the intelligent prediction of the safety risk of the structural service period. The twin model is established in the formation of the structure, and the parameters of the model are modified by genetic algorithm. The actual mechanical parameters of the structure collected by the sensor are used as the index to evaluate the accuracy of the twin model. The validity of the twin model is verified by comparing the perception data with the simulation data. At the same time, the twin data formed by the two are used as the basis of structural safety evaluation. The twin model is used to simulate the mechanical behavior of the actual structure under various working conditions. Driven by Bayesian probability formula, multiple mechanical parameters are integrated to capture the key stress components. By analyzing the degree of relaxation and corrosion of key stressed components, the safety state of the structure can be accurately predicted, and efficient maintenance measures can be formed. Driven by the integration mechanism, a risk perception platform is formed. The platform reduces the time of security risk assessment and ensures the accuracy of analysis. The twin platform can accurately map the safety state of the structure, reduce the dependence of the structure health monitoring on the sensor equipment, and avoid the problems of high cost and great lag of the traditional method. The fusion of sensory data and twin simulation provides ideas for the intelligent operation and maintenance of existing buildings. In this study, the APDL language of ANSYS was integrated with a genetic algorithm to optimize the model parameters, and the twin model established thereby could map mechanical properties of the structure accurately and effectively. Based on the establishment of the high-fidelity twin model, the data relationship of structural security risks is mined. By integrating multiple kinds of mechanical parameters, the key stress components in the structure are obtained, thus reducing the calculation amount of structural safety evaluation. According to the environmental effects of the structure during its normal service period, the key factors affecting the structure safety are analyzed. According to the key security factors, the maintenance measures database is formed to support the health operation and maintenance of the structure.

Driven by the fusion mechanism, the establishment method of structural twin model, structural safety analysis process, and safety maintenance mechanism are formed. The integration of twin simulation and sensory data has two advantages. On the one hand, the sensing data can pick up the mechanical parameters of the real structures, which guarantees the basis for the simulation of the twin model. On the other hand, the structural safety maintenance measures are formed based on the twin model. Finally, the closed-loop control of structural safety is realized by accurate feedback to the actual structures through STs.

2.2. Development of Theoretical Model

Driven by twin simulation and sensory data, the interactions between virtual and real and spatiotemporal evolution are fully considered to realize the intelligent health monitoring of structural service period [46]. Firstly, a high-fidelity twin model with interactive mapping with physical space is established to form twin data for structure risk prognosis. The virtual–real interactive configuration modeling of intelligent safety maintenance realizes the time-history parallel simulation and virtual–real integrated maintenance of multi-element, multi-process, and multi-service processes [47]. The intelligent health monitoring of cable network structures framework based on twin simulation and sensory data is shown in Figure 2.

In the framework, the mechanical parameters of the structures are first collected. In this study, the safety state of the structure is characterized by the change of mechanical parameters. At the same time, a virtual model representing structural state is established. The model is modified to improve fidelity. The corresponding working conditions are set in the model to simulate the safety state of the structure. The fusion of simulation data and measured data forms twin data. The key factors affecting the structural safety are analyzed based on the twin data, in order to obtain efficient maintenance measures based on the safety level of the structure. The feasibility analysis of maintenance measures is carried out in the virtual model to accurately guide the maintenance of the actual structure. The theoretical model of structure risk prognosis is represented by Equation (1).

$$M_{RP}=\left\{\begin{array}{c}{S}_{IF}=\left\{\begin{array}{c}KC=\left(\begin{array}{ccc}K{C}_1,& \cdots & ,K{C}_n\end{array}\right)\\ IF=\left(\begin{array}{ccc}I{F}_1,& \cdots & ,I{F}_n\end{array}\right)\end{array}\right.\\ L_{SS}=\left(\begin{array}{ccc}{L}_{SS1},& \cdots & ,L_{SSP}\end{array}\right)\\ MM=\left(\begin{array}{ccc}M{M}_1,& \cdots & ,M{M}_q\end{array}\right)\end{array}\right.$$

(1)

In Equation (1), $M_{RP}$ is risk prognosis theory model. The theoretical model is composed of three dimensions, namely the influencing factors of space ($S_{IF}$), structure security levels ($L_{SS}$), and maintenance measures ($MM$). $S_{IF}$ consists of structure key stress components ($KC$) and influencing factors ($IF$). The theoretical method is divided into three dimensions, i.e., physical entity, virtual model, and risk prediction. The load and mechanical responses of the actual structure are collected by the sensing equipment, such as the total station and pressure gauge. The collected information is used as the basis for establishing the virtual model. Parameters such as component size and prestress are optimized in the virtual model to ensure the consistency between the simulation data and the collected data. The twin simulation of the normal service period structure is carried out in the virtual model. In order to reduce the time for the simulation calculation, the key stressed components are analyzed in the twin model. At the same time, the key factors affecting structural safety are obtained. According to the literature and engineering practice, the relaxation and corrosion of cable structure occur mainly under the influence of environmental factors in the normal service period. In the twin model, the temperature and other loads of the actual structure are arranged to obtain a degree of relaxation and corrosion of the components. In order to improve the accuracy of the safety assessment, the relaxation and corrosion degree of key components are input into the RF to obtain the safety grade of the structure. Finally, the maintenance measures are formulated according to the safety level of the structure, so as to realize the closed-loop control of the safety risk of the structure.

According to the stress characteristics of the structure during the service period, the effects of cable relaxation and corrosion on the structure are investigated. At the same time, the key stress members of the structure are obtained by integrating various mechanical parameters. In the research process, the combination of influencing factors and key stress components constitutes the factor space that affects the structural safety. According to the influencing factor space, the mechanical response of the structure is obtained. The safety level of the structure is evaluated by the change of mechanical parameters, and the most effective structural safety maintenance measures are finally formed.

The construction of a high-fidelity twin model considering virtual–real interaction and spatial–temporal evolution is the premise to realize the application of intelligent maintenance method. Based on this fusion of twin data, the changes of structural mechanical parameters under the influence of various factors are analyzed, and the most critical components in the structure are accurately found. The safety prediction model is formed under the fusion of twin simulation and sensory data to determine the importance of influencing factors. The key factors are adjusted in the twin model to predict the safety state of the structure. Finally, the intelligent safety maintenance mechanism of structural is constructed. The intelligent safety maintenance method is shown in Figure 3.

3. Numerical Model for Intelligent Health Monitoring of Structures

In this study, structure-oriented intelligent health monitoring forms a numerical model with three aspects. The first is the high-fidelity twin model that characterizes the true state of the structure. Based on the twin model, the structural safety analysis model is formed to obtain the key stress components of the structure. Finally, a structural safety prediction and maintenance model is constructed to form a safety maintenance strategy.

3.1. Establishment of Twin Model

The establishment of a twin model is the first step to realize the intelligent health monitoring of cable network structures. The real-time integration of information and physics is the key to the twin modeling method [48]. In order to strengthen the simulation ability of the twin model, it is necessary to improve the high fidelity of the model and modify the virtual twin model. The genetic algorithm [49] is combined with the monitoring data of the structure and the simulation data of finite element model to correct the virtual model. The basic design parameters of the structure are taken as optimization variables. According to the distribution law of the parameters, the upper and lower limits of their values are determined. This study mainly corrects the initial tension and cross-section area of the cable. According to the literature [50], the two types of parameters are subject to normal distribution and lognormal distribution, respectively. The error ($\Delta \sigma$) between the simulated value and the measured value of the cable force of the structure under self-weight is taken as the optimization objective. The mathematical language of the parameter optimization process is expressed as Equation (2).

$$\left\{\begin{array}{c}\begin{array}{c}{a}_1\le {\omega }_1\le b_1\\ a_2\le {\omega }_2\le b_2\end{array}\\ \begin{array}{c}\cdots \\ a_n\le {\omega }_n\le b_n\end{array}\end{array}\right.\stackrel{\mathrm{iteration}}{\iff }min\left(\Delta \sigma \right)$$

(2)

In the Equation (2), ${\omega }_i$ represents the ith optimization variable, such as the cross-sectional area of the cable. $a_i$ and $b_i$ refer to the upper and lower limits of the ith optimization variable, respectively. By optimizing the basic parameters, the simulation value of the structural cable force is close to the measured value. By modifying the structural–physical properties in the model, a high-fidelity twin model is established. The model can accurately map the state of the actual structure and provide support for the twin simulation of the structural performance.

Cable structures have stiffness and can bear weight only when prestressing is applied. Therefore, the twin model was established in this study. In the actual structure, the cable force measurement of each cable in the shape of the structure is captured. At the same time, prestress and dead weight conditions are arranged in the twin model to obtain the cable force simulation value of each cable. According to formula (2), the optimal parameters of the structure in the twin model are selected to ensure that the error between the simulated value and the measured value is within 5%. In order to verify the high fidelity of the twin model, the construction conditions of the structure were arranged in the model, and the cable forces of each construction step were simulated. The simulation value of cable force during construction is compared with that of actual structure. The error of both is also guaranteed to be within 5%.

Virtual space is a faithful mapping of physical space, and the integration of the two forms a multi-domain and multi-scale twin model [51]. From the perspective of mechanism–data model fusion, this study proposes a mechanism modeling and data modeling and fusion method for intelligent safety analysis maintenance. As shown in Figure 4, this method mainly includes four parts: geometric modeling, mechanism modeling, data modeling, and model fusion and visualization.

In each stage, the elements of spatial dimension are captured, so as to realize the multidimensional information integration of space-time fusion.

(i) In the aspect of geometric modeling, Revit and other three-dimensional modeling software are used to construct the structural unit to realize the accurate expression of key structural parameters, constraints, and positioning relationships between components. The geometric model mainly evaluates its fidelity from the aspects of geometric accuracy, structural integrity, and boundary constraints, so as to ensure that the geometric model is consistent with the real structure visually.

(ii) In terms of mechanism modeling, based on the established geometric model, the structural component subsystem model, the environmental subsystem model, and the subsystem model of service period behavior are constructed by using APDL and other modeling languages or tools, which have the ability to simulate and analyze the structure with high fidelity. The fidelity of the mechanism model of structural safety maintenance is mainly evaluated by the error of the simulation response and the actual response of the same physical quantity under the same working condition.

Firstly, the finite element model is established, driven by the geometric model. For structural members, the material parameters, connection modes, and constraint conditions are set to form the subsystem model of structural members. In order to map the structure of the real world, the corresponding working conditions are set in the finite element model, such as cable relaxation and corrosion, to form the environmental subsystem model. According to the environmental effect, the simulation is carried out in the finite element model to calculate the mechanical parameters of the structure, analyze the state of the structure in real time, and form the behavior subsystem model.

(iii) In terms of data modeling, the sample data set of structural safety analysis and prediction is constructed for the specific application scenarios, such as the change of mechanical parameters and the analysis and prediction of safety performance. Further, data models, such as the safety analysis model, safety prediction model, and safety maintenance model, are constructed by data fusion and machine learning. The fidelity of the data model is evaluated by the accuracy of its prediction analysis.

(iv) In the aspect of model fusion and visualization, based on Java Web technology, taking the geometric model as the visualization carrier and the mechanism model and data model as the kernel, the interoperability and visualization of each model in Web are realized. The mechanism model and data model are the core components of twin model. The specific integration process is as follows: In the data model training stage, the mechanism model provides prior knowledge and simulation data for the data model and improves the generalization ability and accuracy of the data model. The data model constantly feeds back and adjusts the mechanism model in the use process to realize the dynamic updating of structural safety performance and, finally, integrates the functions of high-fidelity simulation analysis and high confidence prediction optimization of supporting structural safety through the mechanism–data model.

3.2. Performance Analysis of Structural Safety

By establishing the high-fidelity twin model, the simulation data in the model can be directly extracted for analysis. By setting the working condition consistent with the real structure, the mechanical parameters of each component of the structure are obtained, and the most critical stress component is intuitively judged. The Bayesian fusion theory can fully consider the changes of multiple factors and realize the damage identification and accurate positioning of the structure [52]. In this paper, it is used to judge the sensitivity of structural mechanics parameters to improve the accuracy of safety maintenance. By integrating the changes of various mechanical parameters, the key components are accurately determined. This method avoids the problem of the inaccurate judgment of single index.

(i) Analysis of mechanical parameters.

In order to accurately reflect the structure, the loads are arranged in the mechanism model as in the real environment. The mechanical parameters, such as cable force and displacement, were obtained in the twin model. By comparing the mechanical parameters under environmental factors with those under standard working conditions, the change rate of the parameters was obtained. The change rate of parameters is the basis of safety analysis. In analyzing the cable force, the probability of the cable force change ($P_i^{*}$) of each node is calculated by Equation (3).

$$P_i^{*}=\frac{\left|C_{fi}^{*}-C_{fi}\right|}{C_{fi}}$$

(3)

In the Equation, $C_{fi}$ represents the cable force value of each node on the cable before the action of various factors, $C_{fi}^{*}$ means to the cable force value of each node on the cable after the action of various factors.

(ii) Analysis of structural safety.

Driven by Bayesian fusion theory, the key stress components and key stress nodes are judged by comprehensively considering the changes of various mechanical parameters. The data fusion avoids the situation of inaccurate judgment due to a single index and significantly improves the accuracy of safety analysis. In the Section 4 of this study, the method is applied.

It is assumed that the index value of the change rate of each mechanical parameter before data fusion is: ${\rho }_1,{\rho }_2,\cdots {\rho }_m$. The identification target (i.e., key nodes and components) is: ${\phi }_1,{\phi }_2,\cdots {\phi }_n$. According to Bayesian probability formula, the conditional probability ($P\left({\rho }_1,{\rho }_2,\cdots ,{\rho }_m|{\phi }_i\right)$) of identifying target (${\phi }_i$) is expressed as Equation (4).

$$P\left({\rho }_1,{\rho }_2,\cdots ,{\rho }_m|{\phi }_i\right)={\displaystyle \prod }_{l=1}^{m}P\left({\rho }_l|{\phi }_i\right)$$

(4)

$P\left({\rho }_l|{\phi }_i\right)$ is the probability of parameter change of ${\phi }_i$ target in the lth test. Taking the change rate of cable force as an example, $P\left({\rho }_l|{\phi }_i\right)$ is expressed as Equation (5).

$$P\left({\rho }_l|{\phi }_i\right)=\frac{P_i^{*}}{{{\displaystyle \sum }}_{i=1}^n{P}_i^{*}}$$

(5)

Finally, the overall change rate of the identified target after the fusion of multiple indicators of the change rate of mechanical parameters is expressed as Equation (6).

$$\begin{gathered} P\left({\phi }_i|{\rho }_1,{\rho }_2,\cdots ,{\rho }_m\right)=\frac{P\left({\rho }_1,{\rho }_2,\cdots ,{\rho }_m|{\phi }_i\right)P\left({\phi }_i\right)}{{{\displaystyle \sum }}_{r=1}^{n}P\left({\rho }_1,{\rho }_2,\cdots ,{\rho }_m|{\phi }_r\right)P\left({\phi }_r\right)} \\ =\frac{\left\{{{\displaystyle \prod }}_{l=1}^{m}P\left({\rho }_l|{\phi }_i\right)\right\}P\left({\phi }_i\right)}{{{\displaystyle \sum }}_{r=1}^n\left\{{{\displaystyle \prod }}_{l=1}^{m}P\left({\rho }_l|{\phi }_r\right)\right\}P\left({\phi }_r\right)} \end{gathered}$$

(6)

In the virtual model with high fidelity, the working conditions corresponding to the site are set, and the mechanical performance index data of the model simulation are extracted to determine the key stress components. The relevant information of the index data is processed by probability, and the comprehensive mechanical parameter sensitivity of each cable is calculated by Bayesian probability formula. According to the sensitivity of the mechanical parameters, the most sensitive component of the mechanical properties of structural members can be judged intuitively, which provides the basis for intelligent safety maintenance. In the component sensitivity analysis, the key mechanical parameters are cable force, stress, and vertical displacement. Through the sensitivity analysis of the mechanical parameters, the process of structural safety performance analysis is shown in Figure 5.

By integrating multiple mechanical parameters, the most sensitive components in the structure can be obtained more accurately. In this study, the component with the highest rate of change of mechanical parameters and the highest stress is taken as the key component. By analyzing the key components in the cable network structure, the safety state of the structure can be accurately characterized. Therefore, the structural safety analysis method based on Bayesian probability formula can accurately capture key components and reduce the calculation time of structural safety assessment.

3.3. Risk Prognosis of Structural Safety

Twin simulation provides an important theoretical basis and technical support for bidirectional connection and real-time interaction between virtual space and physical space [53]. The strong nonlinear fitting ability of artificial intelligence is suitable for the analysis of complex mapping relations. The high-fidelity behavior simulation of online data-driven structural service period is completed by DTs technology, and the real-time state visualization of structure is realized. Based on the intelligent algorithm, the analysis of the safety performance data of the structural service period is realized. The integration of twin simulation and sensory data provides a quantitative analysis basis for risk prognosis.

3.3.1. Fusion Mechanism of Twin Simulation and RF

Based on the above analysis, the STs are integrated into the twin simulation. By combining twin simulation and RF [54], the development trend of structural safety state is obtained, and the adjustment suggestions of structural safety maintenance are given. As shown in Figure 6, the fusion mechanism of twin simulation and RF is constructed. The intelligent prediction of structural safety state is realized, and the key factors affecting structural safety are obtained. Firstly, the structural safety information and the environmental effects are captured in real time [55,56], which provides a basis for the intelligent prediction and maintenance of structural safety. Intelligent prediction and maintenance for structural safety is the core of the integration of twin simulation and RF. Finally, the unsafe condition of the structure is corrected to obtain safety maintenance measures. The feasibility of the correction is analyzed in the twin model, and then the structural safety is accurately guided and maintained. Driven by sensing data, a high-fidelity twin model of the structure is established. The key stressed components are captured in the twin model. The same load is arranged in the twin model, according to the environmental effect of the structure during the normal service period. The relaxation and corrosion of components are the key factors affecting the structure safety. Therefore, the degree of corrosion and relaxation of key components and the corresponding maximum vertical displacement of the structure are obtained in the twin model. According to the change of vertical displacement, the safety level of the structure is divided. The degree of corrosion and relaxation is taken as the input factor, and the degree of safety is taken as the output factor. The importance of each influencing factor was obtained through RF calculation. Various influencing factors are derived from formula (1). Driven by the key factors, the security maintenance measures are formed, and the feasibility analysis of the measures is carried out in the twin model.

By analyzing the fusion mechanism of twin simulation and RF, it can be concluded that the intelligent prediction and maintenance driven by the fusion of twin simulation and RS have the following characteristics:

(i) The key factors of steel structural safety are accurately obtained, and the visual monitoring is realized [57]. Based on the key factors, the intelligent prediction of the development trend of structural safety state is realized.

(ii) According to the development trend of structural safety state, the interaction between mechanical parameters and key factors is fitted. Based on the prediction, the structural safety maintenance measures are given to ensure the safety.

(iii) Using virtual maintenance and real scientific prediction, the data model and intelligent scheme established by RF and other means are simulated by twin model and, finally, fed back to the actual structure to realize the intelligent safety maintenance.

3.3.2. Prediction and Maintenance of Structural Safety Performance

RF is composed of multiple decision trees. Each decision tree is not the same. When constructing the decision tree, the samples are randomly selected from the training data, and all the features of the data will not be used. Each tree uses different samples and features, and the training results are not the same [58]. The mechanical parameters were classified in this study. The key factors affecting structural safety were obtained from RF.

Based on the analysis of the structural safety performance, the key stress members are obtained. Therefore, the cable force variation of the key stress members is taken as the index of structural safety. Different working conditions are set in the twin model to obtain the change of mechanical parameters of each node. Based on the analysis characteristics of RF, the working condition parameters and mechanical parameters are combined to analyze the working condition that has the greatest influence on the structure. The key working conditions are emphatically analyzed to intelligently predict the safety performance of the structure. According to the prediction results, the unsafe state is finally adjusted in the twin model to realize the intelligent safety maintenance. The prediction and maintenance mechanism of structural safety performance is shown in Figure 7.

On the basis of prediction, the key factors affecting structural safety are obtained. When the structure encounters the working conditions of the influencing factors, the maintenance measures of the structure can be obtained directly. The most sensitive components of the structure, regarding environmental factors, can be accurately captured by the fusion of multiple mechanical parameters. In this study, the influence of corrosion and relaxation of key components on structural safety is analyzed. For different types of working conditions, this study formed the main maintenance measures. In order to improve the universality of the research method, different working conditions are classified by influencing factors. According to the influence of different working conditions, the database of structural health monitoring is formed. Driven by the database, the factors affecting the structural safety are maintained accurately and in a timely manner. In the large-scale structure systems (e.g., cable net structures), the mechanical parameters are obtained by cable relaxation and corrosion. The safety state is classified according to the mechanical response of the structure. Finally, the corresponding maintenance measures are formed. The establishment of the database is expressed as Equation (7).

$$W_c=\left\{\begin{array}{c}I{F}_1\\ \begin{array}{c}I{F}_2\\ ⋮\end{array}\\ I{F}_n\end{array}\stackrel{input}{\Rightarrow }\right.M_r=\left(\begin{array}{c}\begin{array}{c}Leve{l}_1\\ Leve{l}_2\end{array}\\ \begin{array}{c}⋮\\ Leve{l}_m\end{array}\end{array}\right)\stackrel{output}{\Rightarrow }{M}_m$$

(7)

$W_c$ denotes the working conditions to which the structure is subjected. The working condition contains a variety of influencing factors $\left(I{F}_i,\left(i=1,2,\dots n\right)\right),$ such as the relaxation and corrosion degree of each cable. Under the action of working conditions, the mechanical response ($M_r$) of the structure is obtained. According to the level of safety $(Leve{l}_i\left(i=1,2,\dots m\right))$, efficient maintenance measures $(M_m)$ are output.

In this study, the output of maintenance measures is based on the change of vertical displacements of the structure. When the working condition is dominated by cable relaxation, the safety performance is mainly maintained by supplementing tension. When the working condition is dominated by cable corrosion, anti-corrosion treatment of components should be carried out. The dominance of the influencing factors in the study means that a certain factor has the greatest influence on the change of the displacement of the structure. Therefore, the maintenance measures of structural safety performance can be output in this study. In order to verify the rationality of the research method, the application of the test case is carried out in Section 4.

In the prediction and maintenance process, the working condition factors are cable relaxation and corrosion. The variation degree of vertical displacement is taken as the index of structural safety. According to the technical regulations [59,60], the maximum displacement of the structure is ensured to be lower than 1/250 of the span. The maximum value of cable force shall not exceed 0.4 times of yield stress.

In the process of structural safety maintenance, twin simulation and sensory data are integrated to form an intelligent maintenance platform. In the platform, the change of environment and the setting of working conditions are taken as the input layers of physical space and virtual space, respectively. Additionally, in the input layer, environment changes and working condition settings are one-to-one mapping. The safety performance is evaluated according to the vertical displacement value to formulate corresponding maintenance measures. According to Equation (7), the structural safety performance and maintenance measures corresponding to the working conditions are formed into a database. The database provides a real-time feedback mechanism for the subsequent safety performance maintenance to ensure that the structural safety level is consistent with normal operation and maintenance status. The structure intelligent safety maintenance platform architecture is shown in Figure 8.

4. Case Study—The Speed Skating Gymnasium

For structural safety maintenance, the real-time dynamic mapping of safety state between the physical entity and virtual model is realized by introducing DTs [61]. Driven by DTs, a virtual model with high fidelity is established, and intelligent algorithms are integrated to analyze and predict mechanical parameters. Finally, the intelligent maintenance mechanism for structural safety is formed. In order to verify the effectiveness of the intelligent maintenance method, the research method is applied to the speed skating gymnasium of 2022 Winter Olympics. For the intelligent health monitoring of the structure, a risk prediction platform is formed to ensure that the structure is in the safe state. The research results provide support for ensuring the safety performance of the structure in the service period.

4.1. Project Profile

The main structure of the speed skating gymnasium in the 2022 Winter Olympic Games is the cast-in-place reinforced concrete construction. The roof of this stadium is the largest single-story and two-way orthogonal saddle-shaped cable network structure in the world. The plan dimension of the cable network is 198 m × 124 m and the building elevation is 15.8 m~33.8 m, which is supported on a circumferential steel ring truss with curtain wall ties on the outside of the truss. The cable network structure consists of load-bearing and stability cables, both of which are closed via high vanadium. Specifically, the load-bearing and stability cables are both double ones. The structural layout plan of the speed skating gymnasium is shown in Figure 9.

The cable net structure in this study is a symmetric structure, so 1/4 of the structure is taken for safety performance analysis. Additionally, 1/4 span structure can represent the whole structure. The cable number of 1/4 span of the structure is shown in Figure 10.

4.2. Construction of Twin Model for the Speed Skating Gymnasium

For structural safety maintenance, the establishment of a high-fidelity twin model is the basis. For the simulation analysis of structural performance, this study uses ANSYS to establish the finite element model. The finite element model is used as a mechanism model to describe the structural safety. The mechanical properties of the structure, such as cable force (C_f), deflection (ω), stress (δ), and strain (ε), can be simulated by setting various functions of the structure, such as cable relaxation and corrosion.

In order to accurately map the state of the structure, the mechanism model is firstly modified. A genetic algorithm, combined with field sensor monitoring data and model simulation data, was used to adjust the cross-sectional area of components and initial tensioning force in the twin model. The consistency of the cable force simulation value and the measured value is improved by the modification rule of the virtual model in Section 3.1. The column tension and pressure sensors are arranged in the structure to collect the cable forces. The displacements of the nodes are collected by the displacement meters. The arrangement of sensing equipment and real-time data collection of the speed skating gymnasium are shown in Figure 11.

In the process of correction, the APDL of ANSYS and genetic algorithm are combined to update the basic parameters of the structure. Under the action of self-weight, the cable force of each node of the radial cable is collected. Meanwhile, the corresponding simulation values are obtained from the model. The error of the two models is taken as the index to evaluate the model fidelity. The mapping of real structure can be realized. The change of the section size and the initial tensioning force of the component will lead to the change of the structural cable force. The fidelity of the mechanism model is improved by modifying the two types of parameters. After modification, the diameter of the load-bearing cable is 62–68 mm, and the diameter of the stability cable is 70–76 mm. The initial tensioning force of each cable after correction is shown in Figure 12. Taking the cable force of the stabilization cable in Figure 10 as the research object, the comparison of the simulation degree of the cable force before and after the model modification is shown in

Table 1. Comparison of the simulation of cable force before and after model modification (unit: kN).

Component Element Number	Simulation Value	Modification Value	Monitoring Value	Simulated Value Error	Correction Value Error
WDS1	2540.73	2185.01	2228.71	14%	2%
WDS2	2888.02	2696.56	2750.49	5%	2%
WDS3	3006.64	2756.36	2783.92	8%	1%
WDS4	2802.13	2803.25	2747.19	2%	2%
WDS5	2620.95	2848.86	2848.86	8%	0%
WDS6	2940.61	2887.20	2800.58	5%	3%
WDS7	2744.65	2922.02	2951.24	7%	1%
WDS8	2656.86	2955.68	2985.24	11%	1%
WDS9	2960.78	2992.50	3052.35	3%	2%
WDS10	3135.21	3018.11	2957.75	6%	2%
WDS11	3475.93	3049.06	3049.06	14%	0%
WDS12	3380.72	3068.92	3130.30	8%	2%
WDS13	2844.16	3094.51	3125.46	9%	1%
WDS14	3400.38	3102.25	3009.18	13%	3%
WDS15	3489.29	3112.38	3143.50	11%	1%

.

According to the establishment method of the twin model in Section 3.1, the measured and simulated values of the structural form cable forces are compared. Driven by the genetic algorithm, the initial tension and cross-sectional area of the cable in the twin model are optimized. Figure 12 shows the parameters finally selected after multiple rounds of optimization. Table 1 compares the simulation errors of the twin model before and after optimization. According to Table 1, it is found that the modified model based on APDL and genetic algorithm can more accurately map the security state of the structure. The error of the model before and after modification can be reduced by up to 14%. Therefore, the twin model establishment method proposed in this study is feasible. Under the action of self-weight, the force of WDS15 is the largest in the stability cables. CZS25 has the largest stress in the load-bearing cables.

In order to verify that the twin model can accurately analyze the mechanical response of the structure under multiple working conditions, this study compared the measured and simulated cable forces at each stage of the construction process. In the construction process of cable net, there are 17 construction steps. At the construction site, the pressure gauge captures the cable force at each stage in real time. According to the construction scheme, the construction process is simulated in the twin model, and the simulation value of each construction step cable force is obtained. As shown in Figure 13, the measured and simulated values of cable force in each construction step are compared.

As shown in Figure 13, the error between the simulated value and the measured value of cable force in each construction step is controlled within 5%. Therefore, it is found that the twin model established in this study can accurately analyze the mechanical response of the structure under various working conditions. The establishment of the twin model provides reliable support for safety analysis and the prediction of structures. The twin model can accurately predict the mechanical response of the structure during the normal service period and can assist in the formulation of maintenance measures.

4.3. Numerical Analysis and Field Measurement of Structural Safety Performance

Based on the high-fidelity twin model, the load action of the structure is adjusted, and the changes of mechanical parameters are analyzed. At the same time, according to the data fusion strategy in Section 3.2, the mechanical properties of structural members are judged. The key stress members are obtained by comparing the changes of mechanical parameters under the self-weight condition and after the arrangement of dead load. According to the load specification, the constant load is equivalent to the node load, which is set at 1.2 kN. The load layout and twinning simulation of the structure are shown in Figure 14.

For the important integral tensioning structure, the internal force and displacement should be controlled. Because the cross-section area of each cable is different, the change rate of cable force and stress is also different. In the twin model, the rate of change of cable force, strain, and displacement, before and after constant load, applied to each cable is obtained. The key stress members of the structure are judged by combining the change rate of three types of mechanical parameters with Bayesian probability formula. In this study, the cable corresponding to the highest rate of change is taken as the key stress member. Taking the displacement change as an example, the vertical displacement of the structure, before and after applying dead load, is shown in Figure 15.

In order to clearly compare the changes of mechanical parameters of each cable, the rate of change was normalized. Taking the displacement change rate of the stability cable as an example, the change rate before normalization is calculated by Equation (8).

$$P_{vdi}^{WD}=\frac{\left|V{D}_i^{*}-V{D}_i\right|}{V{D}_i}$$

(8)

$P_{vdi}^{WD}$ is the stability cable vertical displacement rate, $V{D}_i^{*}$ represents vertical displacement of cable mid-span node after applying dead load, and $V{D}_i$ denotes vertical displacement of cable mid-span node before applying dead load. After normalization, the vertical displacement change rate ($P_{vdi}^{WD*}$) of the ith cable is represented by Equation (9).

$$P_{vdi}^{WD*}=\frac{P_{vdi}^{WD}}{{{\displaystyle \sum }}_{i=1}^{15}{P}_{vdi}^{WD}}$$

(9)

A total of 15 stability cables are analyzed in this study. The change rate of other mechanical parameters of the stability cable and the load-bearing cable is the same as that of the vertical displacement. The variation of three kinds of parameters of each stability cable and load-bearing cable are obtained. Three kinds of parameters are fused by Bayesian fusion theory to obtain the key stability cable and load-bearing cable in the structure. The sensitivity of each cable after data fusion is shown in Figure 16.

In Section 3.2, multiple types of mechanical parameters can be integrated by the Bayesian probability formula to identify key components more accurately. In this study, the change rate of mechanical parameters, before and after loading, is used as the standard to measure the key stressed components. The more obvious the variation of mechanical parameters under load, the more critical the corresponding component. According to Figure 16, the stability cable and load-bearing cable with the most sensitive mechanical parameters in the structure under load are the edge cables, namely WDS1 and CZS1. The change rate of mechanical parameters of these two cables is obviously higher than that of other cables of the same type. Therefore, for the service life of the structure, the members whose mechanical parameters are most easily changed are obtained. The influence of these components on structural safety should be studied during the service period. During the service period of the structure, the influence of the relaxation and corrosion of the key components on the safety state of the structure should be analyzed. The calculation time of safety assessment can be significantly reduced by capturing the key stress components, and the analysis of key components can accurately reflect the safety state of the structure.

4.4. Key Influencing Factors and Maintenance Measures of Structural Safety Performance

By analyzing the safety performance of the structure, the key stress members of the structure are obtained. The key stress components selected in this study are the cable with the largest internal force and the largest variation of mechanical parameters under load. The components selected in this study better represent the influencing factors of the structural safety state during service. The corresponding components in the speed skating gymnasium structure of 2022 Winter Olympics are WDS1, WDS15, CZS1, and CZS25. During the service period, the cable net structure members are prone to corrosion and relaxation, which seriously affects the safety of the structure. The maximum depth of cable corrosion is 0.56 mm [62]. According to the cable diameter of the structure, the maximum corrosion degree is 1%. The maximum relaxation of the cable is 4.26% [63]. The single influencing factors analyzed in this study are shown in

Table 2. Single factors affecting structural safety.

Component	WDS1		WDS15		CZS1		CZS15
Influence factor	corrosion	relaxation	corrosion	relaxation	corrosion	relaxation	corrosion	relaxation
Value range	0–1%	0–4.26%	0–1%	0–4.26%	0–1%	0–4.26%	0–1%	0–4.26%

. According to the captured key stress components, the degree of corrosion and relaxation of each component is randomly arranged in the twin model. For example, for WDS1, the degree of corrosion is 0 to 1%, and the degree of relaxation is 0 to 4.26%. The twin model can accurately simulate the corrosion and relaxation degree of the cable during the normal service period. According to the amount of corrosion and relaxation, the mechanical response of the structure can be obtained for the safety maintenance of the structure.

In order to obtain the key influencing factors, this study combines multiple factors in Table 2. The specific values of corrosion and relaxation are randomly sampled from the corresponding range of values. The corresponding working condition is set in the structure to obtain the maximum vertical displacement. Under the most unfavorable combination of factors, the maximum vertical displacement is −205 mm. During normal operation and maintenance of the structure, the maximum vertical displacement is −130 mm. According to

Table 3. Relationship between vertical displacement and structural safety level.

Vertical displacement (mm)	[−152.5, −130)	[−170, −152.5)	[−187.5, −170)	[−205, −187.5)
Safety level	level a	level b	level c	level d

, the vertical displacement is classified. During operation and maintenance, the vertical displacement changes under the influence of relaxation and corrosion. According to the value of vertical displacement, the safety grade of the structure is divided into a, b, c, and d. Driven by RF, the degree of relaxation and corrosion of key components are taken as the input parameters, and the safety level of the structure is taken as the output parameter. Finally, according to the relationship between input and output variables, the most critical factors affecting structural safety are obtained. Among them, the input parameters include eight influencing factors, namely the degree of corrosion and relaxation of key components. The input and output forms of data in RF are shown in

Table 4. Input and output forms of data in RF.

Input								Output
WDS1 Corrosion	WDS1 Relaxation	WDS15 Corrosion	WDS15 Relaxation	ZCS1 Corrosion	ZCS1 Relaxation	ZCS25 Corrosion	ZCS25 Relaxation	Vertical Displacement (mm)	Safety Level
0.5%	2.3%	0.8%	1.4%	0.16%	1.5%	0.43%	1.84%	−155.3	b

.

The large sample was formed by random sampling. According to the change of safety state, the influence degree of eight factors on structural safety performance was obtained, as shown in Figure 17.

Figure 17 shows that the most influential factor on the structural safety is the relaxation of WDS1. The importance of this influencing factor is 0.329, which is much higher than other influencing factors. Therefore, the influence of WDS1 relaxation on structural safety should be emphasized during the service period. Additionally, based on the mechanical response of the structure, effective maintenance measures are formulated for the most critical influencing factors. The method of supplementing tension has a significant effect on safety accidents caused by cable relaxation. In this study, the vertical displacement of the structure before the relaxation of WDS1 is taken as the basis. According to the different degrees of relaxation, the scheme of supplementing tension is formed to ensure that the vertical displacement of the structure is consistent with that before relaxation. Under the normal service condition, the supplementing tension corresponding to different relaxation degrees of WDS1 is shown in Figure 18. This study clarified the coupling relationship between the key influencing factors and vertical displacement. The development trend of the vertical displacement of the structure under different relaxation degrees of WDS1 is shown in Figure 19. According to the technical rules of cable structures [60], the mechanical properties of the cable are in the stage of linear elasticity under normal use. Therefore, for the relaxation of a single cable, there is a strong linear relationship between the mechanical response and the degree of relaxation. The coupling relationship between the relaxation degree and vertical displacement is expressed as Equation (10). The development trend of vertical displacement and maintenance measures under other influencing factors are also obtained. The resulting results are embedded in the twin platform to guide structural health monitoring.

$$y=-201.72521x-133.73844$$

(10)

4.5. Intelligent Health Monitoring for the Speed Skating Gymnasium

Through the analysis of the safety performance of the structure, the key stress members and the key factors affecting the safety of the speed skating gymnasium structure are found. Additionally, security maintenance measures are formed. Based on the platform architecture in Section 3, the structural safety DTs–STs platform of the speed skating gymnasium is developed, as shown in Figure 20. The twin platform is the link between security analysis and maintenance. Driven by the twin platform, intelligent health monitoring of the structure is realized. In this process, STs provide real data support for the establishment of the twin platform and effectively transmits maintenance measures. In the platform, the establishment of the twin model, the capture of key components, and the analysis of key factors are integrated in Section 4.2, Section 4.3 and Section 4.4. The twin model has been verified by the comparison of the construction process data, so it can accurately analyze the mechanical response under various load conditions. By capturing the key stress components, the calculation time of safety assessment is effectively reduced. Finally, combining the results of twinning simulation with RF, the key factors affecting structural safety are analyzed accurately. According to the mechanical response of the structure under the action of key factors, the maintenance measures are formulated. Finally, the feasibility analysis of maintenance measures is carried out in the twin model, and the intelligent closed-loop control of structural safety risks is realized. The operation process of the platform is based on the mechanical parameters collected by the total station and pressure gauge. The measured values of the parameters ensure the accuracy of the twinning simulation. At the same time, based on the prediction of the platform, the development law of the mechanical parameters of the structure is accurately obtained, and the dependence on the sensing equipment is reduced. Therefore, the cloud platform established in this study can accurately predict the impact of the security risks on the structure and, at the same time, assist in the development of maintenance measures, effectively solving the problems of high acquisition cost, insufficient analysis accuracy, and the hysteresis of the traditional monitoring system.

By embedding the twin model into the maintenance platform, on the one hand, the dynamic behavior of the structure can be obtained in real time to judge the safety state of the structure. On the other hand, for events requiring security maintenance, the feasibility analysis of decision making can be carried out in the mechanical simulation module. According to the analysis in Section 4.3 and Section 4.3, vertical displacements corresponding to cable relaxation and corrosion are mainly captured in the decision and early warning module. In the case of displacement changes, the maintenance measures are output. In the visualization module, the information is fully transformed into 3D models and visualized data. The module includes the structure displacement change, component deformation, and other functions. In the data management module, the working conditions and maintenance measures corresponding to displacement changes are collected. In this process, the displacement meter and other data acquisition equipment sense the change of structural mechanical parameters, which ensures the rationality of the safety maintenance. Driven by the platform, structural maintenance measures are formed. Based on the maintenance measures, the structural safety control system of the speed skating gymnasium is developed, as shown in Figure 21. The system monitors the status of maintenance devices in real time. At the same time, the maintenance equipment parameters are set to ensure the safety performance of the structure.

With the collaboration of platform modules, the corresponding supplementation under different influencing factors is shown in Figure 22. The finite element simulation method is used to verify the measures of the platform. The output value of the platform is highly consistent with the authentication value of the finite element. Finally, a database for intelligent maintenance of structural safety is formed. For the structure of the speed skating gymnasium, some influencing factors, as well as the corresponding most effective maintenance measures, are shown in

Table 5. Partial influencing factors and corresponding most effective maintenance measure.

Working Condition	Maintenance Measure
WDS1 relaxation 2%, WDS15 corrosion 0.2%	supplementation 46.3 kN
WDS15 relaxation 1.6%, CZS25 relaxation 2.3%	supplementation 52.8 kN
CZS1 corrosion 0.98%, WDS1 relaxation 1.2%	preservative treatment
CZS25 corrosion 0.36%, WDS15 relaxation 3.6%	supplementation 62.4 kN

.

In the platform, the change of the vertical displacement of the structure is mainly captured. In order to achieve accurate control, the decision making and feasibility analysis are carried out in the twin model. The main maintenance strategies formed in this study are consistent with the similar structures [64], which further illustrates the feasibility of this research method in structural safety maintenance.

5. Conclusions and Future Works

In this study, considering the problem that the uncertainty of the mechanical parameters of the cable network structure affects the safety maintenance, an intelligent health monitoring of cable network structures method driven by DTs model and fusion of sensory data is proposed. Sensing devices, such as cable force and displacement sensors, extract the mechanical parameters of the structure, which provide data support for this study. The application of STs ensure the rationality of structural safety analysis and maintenance. Oriented to structural intelligent health monitoring, the fusion mechanism of twin simulation and sensory data is formed. A theoretical model of structural health monitoring is established from the perspective of virtual–real interaction and spatial–temporal evolution. A numerical model of structural intelligent health monitoring is developed based on the theoretical model. In this study, the theoretical method is applied to the health monitoring of the Winter Olympic skating gymnasium structure. The theoretical method formed in this study has been applied to the symmetric structure, which provides ideas for the safety maintenance of large symmetric structures. The main conclusions are as follows:

(i) The fusion mechanism of twin simulation and sensory data is established. A theoretical model for structural intelligent health monitoring is formed. The mechanical parameters of the structure are obtained by sensing technology, and the safety state of the structure is mapped in real time in the twin model. Firstly, the influencing factor space of structural safety is constructed. Under the action of factors, the safety state of the structure is obtained. According to the safety state of the structure, efficient maintenance measures are formed. Finally, the maintenance measures are transmitted to the actual structure by the STs.

(ii) Driven by the theoretical model, a numerical model of structural intelligent health monitoring is constructed. The model includes the establishment of a twin model, the analysis of structural safety performance, and the maintenance of structural safety performance. The fidelity of the modified twin model is more than 95%. The function of the structure is set in the twin model to obtain the variation of the mechanical parameters of each component. The key stress components are captured according to the change of parameters. The results show that this research method can accurately analyze the safety state of the structure and effectively reduce the cost of structural monitoring. Based on the fusion of twin simulation and RF, the prediction and maintenance mechanism of structural safety performance is established. The method of establishing the database of influencing factors and maintenance measures is formed.

(iii) The research method is applied to the speed skating gymnasium of 2022 Winter Olympics, which effectively ensures the safety performance of the structure during the service period. The key stress members and the key factors affecting the safety of the structure are captured. The coupling relationship between key factors and structural safety state is formed. The most effective maintenance measures are established according to the safety state of the structure. A structural intelligent health monitoring platform for structural safety is formed based on theoretical methods. In the platform, the database of structural safety influencing factors and corresponding maintenance measures are formed to ensure that the structure is in a safe state during the service period.

Under the background of industry 4.0, the realization of intelligent structure health monitoring is the research focus of civil engineering. In this study, the reliability of the twin model simulation is verified based on the mechanical parameters of the actual structure collected by sensing equipment such as pressure gauge. Driven by the twin model, the key components and the key influencing factors in the structure are captured. On this basis, a security risk prediction platform is established, which effectively solves the problems of high acquisition cost, low evaluation accuracy, and efficiency of traditional operation and maintenance system. This research method provides ideas for intelligent health monitoring of existing buildings. This research method effectively solves the problem of insufficient safety performance analysis and prediction accuracy caused by the real-time change of mechanical parameters and cannot realize the intelligent safety maintenance of the structure. Based on the analysis and prediction of structural safety, the interaction mechanism between load conditions and maintenance measures is formed. The results of this study provide a reliable reference for the structural health monitoring of Olympic venues. This research method also provides a basis for the maintenance of the safety performance of other building structures (truss structures, chord structures, reticulated shell structures). This study analyzes the influence of the corrosion and relaxation of key components on structural safety and forms maintenance measures. In the future, the influence function of random rust field will be established to analyze the influence of cable corrosion on structural safety and form effective maintenance measures. At the same time, the DTs and deep learning will be integrated to consider more influencing factors to achieve a comprehensive assessment of structural safety. Finally, the intelligent operation and maintenance of the structure is realized by STs. Under the theoretical framework of DTs, the intelligent monitoring and maintenance of the whole life cycle of structures by integrating image recognition and other perceptual technologies is the research focus under the background of industry 4.0.