Distributed sensor network deployment and data fusion method for towering steel structure

Abstract

The invention discloses a distributed sensor network deployment and data fusion method for a towering steel structure. The method relates to the technical field of network deployment and data fusion, and comprises the following steps: key monitoring part identification and sensor deployment, monitoring data processing and adaptive fusion, and digital twin driving and early warning emergency. According to the method, key monitoring parts of all stages are determined based on a finite element model and construction simulation, and a heterogeneous sensor network capable of being dynamically adjusted is deployed; collecting data, binding the data to a construction stage after edge preprocessing, and generating a monitoring characteristic value through three-stage adaptive fusion; and finally, constructing a digital twin model to update and display in real time, and matching an emergency plan according to whether the characteristic value exceeds the early warning interval, thereby improving the distributed sensor network deployment and data fusion accuracy, and solving the problem that in the prior art, due to construction environment interference and network topology optimization insufficiency, the data fusion accuracy is poor. And therefore, the problem of low accuracy of distributed sensor network deployment and data fusion can be solved.

Description

Technical Field

[0001] This invention relates to the field of network deployment and data fusion technology, and in particular to a method for deploying and fusioning distributed sensor networks for tall steel structures. Background Technology

[0002] Tall steel structures have a large height-to-section ratio, high overall flexibility, and concentrated weak points (such as the bottom stress zone, the top displacement-sensitive zone, and the variable cross-section stress concentration zone). The existing deployment does not fully integrate the structural dynamics characteristics and modal analysis results, resulting in either insufficient sensor deployment in key areas, creating monitoring blind spots, or redundant deployment on symmetrical or similar members, leading to resource waste. The root cause of this problem is the lack of a systematic deployment strategy that takes into account structural characteristics, monitoring objectives, and cost-effectiveness, and the failure to quantify the contribution of measurement points to mode identification through scientific methods such as modal confidence criteria.

[0003] Meanwhile, the strong winds, high temperatures, electromagnetic interference, and shielding from reinforced concrete components in the construction environment of tall steel structures, coupled with the inherent limitations of sensor nodes relying on battery power, make signal attenuation, delays, and even packet loss easy to occur during data transmission. The root cause is that the impact of environmental interference on communication links was not fully considered during deployment, and the data transmission path and network topology were not optimized, resulting in some nodes being overloaded and consuming energy too quickly, thereby damaging network stability.

[0004] At the data fusion level, due to the lack of targeted selection of sensor types (failure to match and adapt sensors according to different monitoring parameters such as vibration and displacement), and the significant nonlinear characteristics of the structural dynamic response during construction, existing fusion algorithms have failed to effectively integrate multi-source heterogeneous data and have difficulty distinguishing between environmental noise and the actual structural response. The root cause is that the fusion model has not fully incorporated the structural nonlinear effects and the coupling effects of multiple loads, and cannot achieve accurate mapping between monitoring data and structural health status. Ultimately, this leads to both data redundancy and information loss. There are also problems with low accuracy in distributed sensor network deployment and data fusion due to construction environment interference and insufficient network topology optimization. Summary of the Invention

[0005] To address the technical problems of low accuracy in distributed sensor network deployment and data fusion caused by construction environment interference and insufficient network topology optimization in existing technologies, this invention provides a method for distributed sensor network deployment and data fusion for tall steel structures. The technical solution is as follows: On the one hand, a method for deploying distributed sensor networks and fusing data for tall steel structures is provided. This method includes: S100, based on the finite element model of the tall steel structure to be monitored and the simulation of the entire construction process, identifying key monitoring parts at each construction stage, and distributing the key monitoring parts through a heterogeneous sensor network to adapt to dynamic adjustments during construction, allowing for location migration and supplementation as construction stages progress; S101, acquiring the original parameters of the key monitoring parts, preprocessing the original data at the edge computing gateway, binding the preprocessed data with the current construction stage, and performing three levels of adaptive fusion. The value output after adaptive fusion is recorded as the monitoring feature value. The three levels of adaptive fusion include data-level fusion, feature-level fusion, and decision-level fusion; S102, constructing a digital twin model of the tall steel structure, using the adaptive fusion features and evaluation results to drive the digital twin model to update its status and visualize it in real time, determining whether the monitoring feature value is within the warning critical range output by the digital twin model, and if so, providing a corresponding emergency plan based on the warning level.

[0006] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. By achieving dynamic adaptation and precise coverage of key monitoring throughout the entire construction cycle, the pain points of traditional monitoring deployments—fixed deployment and poor adaptability—are effectively addressed. Key monitoring locations at each stage are identified in advance through finite element models and full-process construction simulation. Coupled with a heterogeneous sensor network deployment mode that supports location migration and addition, the sensor positions and quantities can be flexibly adjusted according to the construction progress, ensuring that core areas such as main load-bearing component nodes and temporary support connection points are always within the effective monitoring range. Simultaneously, the non-intrusive / semi-intrusive installation method and the LoRa and NB-IoT low-power wide-area network design ensure the stability and low energy consumption of data transmission in complex construction environments while avoiding damage to the steel structure's inherent performance. Furthermore, the mechanism of dynamically adjusting monitoring priorities based on strain increment weights further improves the efficiency of precise allocation of monitoring resources and reduces resource waste from ineffective monitoring points.

[0007] 2. The three-level adaptive data fusion system significantly improved the quality and feature representation capabilities of monitoring data, providing high-precision and high-reliability core data support for safety assessment. In data-level fusion, the interpolation step size was dynamically adjusted based on the time coding granularity of the construction stage, and the interpolation error tolerance window was optimized by combining the filtering cutoff frequency. This enabled accurate repair of missing data under different construction conditions, ensuring both the reconstruction accuracy of stable-stage data and the retention of key features of transient signals. Feature-level fusion effectively integrated core information from multiple sources such as vibration, strain, and deformation through multi-dimensional feature extraction and principal component analysis dimensionality reduction, combined with dynamic optimization of the load path anomaly judgment threshold using the overall tilt synthesis vector. This eliminated redundant interference and improved the sensitivity to identify local dangerous deformation patterns. Decision-level fusion, through the combination of construction stage labels and deep learning models, coupled with dynamic adjustment of the serialization window length based on the state evolution rate, achieved accurate assessment of safety status at different construction stages, enhanced the real-time capture capability of transient risks, and avoided misjudgments and omissions caused by rigid assessment standards.

[0008] 3. A full-process visualized safety management and control system was constructed through the real-time driving and hierarchical early warning mechanism of the digital twin model, significantly improving the timeliness and scientific nature of construction safety decisions. By synchronizing the fused features and assessment results to the digital twin model in real time, the visualized dynamic update of the construction status of the tall steel structure was realized, allowing managers to intuitively grasp the overall and local health status of the structure. Based on the early warning thresholds dynamically adjusted according to the construction stage, design specifications, and environmental loads, combined with a two-level early warning judgment logic, the first-level early warning promptly alerts potential risks, while the second-level early warning is automatically pushed and highlighted in the 3D model to quickly locate the risk areas and coordinate with management personnel for handling. At the same time, the time-series archiving and traceability design of the monitoring data not only provides complete data support for the safety review of the construction process, but also forms a closed-loop safety management and control model of "monitoring-assessment-early warning-handling-traceability". Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating a distributed sensor network deployment and data fusion method for tall steel structures provided in this application embodiment; Figure 2 This is a feature-level fusion flowchart of a distributed sensor network deployment and data fusion method for tall steel structures provided in an embodiment of this application. Detailed Implementation

[0011] The technical solution provided in this application will now be described with reference to the accompanying drawings.

[0012] To facilitate understanding of the embodiments of this application, the following points will be explained first: First, in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but it does not exclude the possibility of indicating an "and" relationship; the specific meaning can be understood in context. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Here, a, b, and c can be single or multiple.

[0013] Second, the use of prefixes such as "first" and "second" in this application is solely for the purpose of distinguishing and describing different things belonging to the same category, and does not constrain the order, size, or quantity of things. For example, "first message" and "second message" are simply different messages, and there is no chronological, size, or priority relationship between them.

[0014] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0015] like Figure 1 The diagram shown is a flowchart of a distributed sensor network deployment and data fusion method for tall steel structures provided in an embodiment of this application. The method includes the following steps: S100, based on the finite element model of the tall steel structure to be monitored and the simulation of the entire construction process, identifies the key monitoring parts at each construction stage. It then distributes the key monitoring parts through a heterogeneous sensor network to adapt to dynamic adjustments during construction, allowing for location relocation and supplementation as construction progresses.

[0016] It should be noted that the sensor nodes communicate and network with each other via a low-power wide-area network, using the LoRa or NB-IoT protocol; the nodes are installed in a non-intrusive or semi-intrusive manner, including by attaching them to the steel structure surface via magnetic bases, special clamps, or structural adhesive.

[0017] It should be understood that key monitoring areas include, but are not limited to: installation nodes of main load-bearing components that change dynamically with the construction progress, connection points of temporary support systems, and areas of high-altitude welding and bolted connections; heterogeneous sensor networks include, but are not limited to, fiber optic grating sensors and resistance strain gauges for monitoring component stress, total stations or electronic theodolites for monitoring three-dimensional deformation and verticality, piezoelectric accelerometers for monitoring structural vibration, and environmental sensors for collecting ambient wind speed, temperature, and humidity.

[0018] Furthermore, the specific steps to allow for relocation and addition during the construction phase are as follows: By installing several strain gauges on key monitoring parts of the structure, the original strain signals of each monitoring point at a predefined time are synchronously acquired at a pre-set sampling frequency. The original strain signals are filtered and temperature compensated to obtain the effective strain value, and the real-time strain increment within a preset time window at the predefined time is obtained. Within a preset time window, all monitoring points are traversed to obtain the real-time strain increment with the largest absolute value. The strain increment normalization weight factor for each monitoring point is calculated, and the real-time force ratio of the monitoring point at a predefined time is obtained. The strain increment normalization weight factor is dynamically adjusted to the force ratio warning threshold for each monitoring point to generate a dynamic force ratio warning threshold.

[0019] The specific steps for generating the dynamic force ratio early warning threshold are as follows: If the strain increment normalization weight factor is greater than the strain increment weight benchmark value, it indicates that the monitoring point is experiencing relatively drastic stress changes. In this case, the stress ratio warning threshold of the monitoring point is relaxed to obtain the dynamic stress ratio warning threshold, so as to avoid triggering false alarms due to normal load redistribution. If the strain increment normalization weight factor is less than or equal to the strain increment weight benchmark value, the current dynamic stress ratio warning threshold will be maintained. The real-time stress ratio of each monitoring point is compared with the dynamic stress ratio warning threshold. If the real-time stress ratio is greater than the dynamic stress ratio warning threshold, an early warning signal is immediately triggered, and the monitoring point number and strain increment normalization weight factor are identified. If the real-time force ratio is less than or equal to the dynamic force ratio warning threshold, no signal needs to be triggered.

[0020] In this embodiment, firstly, several strain gauges are deployed at key monitoring locations of the structure. A data acquisition system synchronously acquires the raw strain signals of each monitoring point at a predefined time using a pre-set sampling frequency. This step, through the design of "synchronous acquisition" and "preset sampling frequency," ensures the timeliness and consistency of the raw data, effectively avoiding subsequent analysis errors caused by asynchronous data acquisition from different monitoring points, and laying a data foundation for accurately obtaining the structural stress state. Subsequently, the acquired raw strain signals are filtered and temperature-compensated: filtering effectively removes noise signals caused by environmental interference (such as electromagnetic interference, vibration interference, etc.), improving the signal-to-noise ratio of the strain signals; temperature compensation eliminates the influence of environmental temperature changes on the measurement results, ensuring that the obtained effective strain values ​​accurately reflect the actual stress and deformation of the structure, avoiding misjudgments of the stress state due to temperature errors. Based on the processed effective strain values, the real-time strain increment within a predefined time window is further calculated. This real-time strain increment directly reflects the stress change trend of each monitoring point within a specific time range, providing core parameters for subsequent judgment of the structural stress stability. Next, within a preset time window, the real-time strain increments of all monitoring points are traversed and filtered to determine the real-time strain increment with the largest absolute value. Based on this, the normalized weighting factor of the strain increment of each monitoring point is calculated, thereby obtaining the real-time force ratio of each monitoring point at a predefined time. This step, through the introduction of the "normalized weighting factor", can quantify the contribution of each monitoring point to the overall structural force, clearly locate the main stress-bearing parts of the structure, and provide a clear direction for targeted early warning and control. Subsequently, based on the relationship between the strain increment normalized weight factor and the preset strain increment weight benchmark value, the stress proportion warning threshold for each monitoring point is dynamically adjusted: if the strain increment normalized weight factor is greater than the strain increment weight benchmark value, it indicates that the monitoring point is experiencing relatively drastic stress changes. In this case, the stress proportion warning threshold for the monitoring point is relaxed to generate a dynamic stress proportion warning threshold. This dynamic adjustment mechanism can effectively avoid false alarms caused by normal load redistribution of the structure (such as the dynamic transfer of loads in different parts of the bridge during vehicle travel, the redistribution of loads on the building structure under wind loads, etc.), thus improving the practicality and reliability of the warning system. If the strain increment normalized weight factor is less than or equal to the strain increment weight benchmark value, the current dynamic stress proportion warning threshold is maintained to ensure that strict warning standards are maintained for monitoring points with stable stress states, avoiding the risk of missed alarms due to excessive relaxation of the threshold.Finally, the real-time stress ratio of each monitoring point is compared with the corresponding dynamic stress ratio warning threshold one by one: if the real-time stress ratio of a monitoring point is greater than the dynamic stress ratio warning threshold, it means that the stress state of the monitoring point has exceeded the safe and controllable range. The system immediately triggers a warning signal and simultaneously identifies the corresponding monitoring point number and strain increment normalization weight factor. By accurately identifying the fault location and key parameters, it can provide clear basis for maintenance personnel to quickly locate the fault point and judge the severity of the fault, shorten the fault handling response time, and reduce the risk of structural damage caused by excessive stress. If the real-time stress ratio is less than or equal to the dynamic stress ratio warning threshold, it means that the stress state of the monitoring point is within the safe range and there is no need to trigger a warning signal, ensuring that system resources are not occupied by invalid warnings and improving the operating efficiency of the monitoring and warning system.

[0021] Furthermore, to allow for relocation and addition as construction progresses, it also includes: A mapping relationship between displacement change and stress early warning threshold adjustment coefficient is established. This is achieved by acquiring the initial values ​​of support displacement at each monitoring point and the corresponding initial stress values. Support displacement values ​​are collected in real-time over a predefined monitoring period to obtain the displacement change relative to the initial values. The displacement change is then compared with a reference range. Specifically: If the displacement change is within the reference range, it is determined to be in the plastic development zone. The current displacement change is input into the mapping relationship between the displacement change and the stress warning threshold adjustment coefficient. The warning threshold adjustment coefficient is output. The warning threshold adjustment coefficient is combined with the current stress warning threshold to obtain the target stress warning threshold, so as to lower the stress warning threshold and prevent missed alarms. The reference range represents the open interval formed by the lower limit of the reference range and the upper limit of the reference range. If the displacement change is less than or equal to the lower limit of the change reference, it is determined to be in the elastic stability zone, and the current stress warning threshold is maintained. If the displacement change is greater than or equal to the upper limit of the change benchmark, it is determined to be a warning danger zone, and an early warning message is sent to the preset relevant personnel.

[0022] In this embodiment, firstly, a mapping relationship between displacement change and stress warning threshold adjustment coefficient is pre-established. This mapping relationship is the core foundation for achieving dynamic adaptation of the stress warning threshold. By quantifying the correlation between displacement change, a structural state characterization parameter, and the adjustment range of the warning threshold, a scientific basis is provided for subsequent precise adjustment of the warning standard, avoiding subjectivity and blindness in threshold adjustment. Subsequently, through deployed displacement monitoring equipment and stress monitoring equipment, the initial values ​​of support displacement and corresponding stress at each monitoring point are simultaneously collected. The accurate collection of initial values ​​provides a benchmark for subsequent calculation of displacement change and judgment of structural state, effectively eliminating misjudgment of state due to lack of benchmark and ensuring the accuracy of subsequent analysis. During structural operation, the support displacement values ​​of each monitoring point are continuously collected in real time according to a predefined monitoring cycle. By calculating the difference between the real-time support displacement value and the previously collected initial support displacement value, the displacement change of each monitoring point relative to the initial value is obtained. This displacement change can intuitively reflect the deformation state of the structural support parts and is a key indicator for judging whether the structure is in a stable operating state. Its real-time collection can ensure timely capture of structural state changes and avoid risk omissions due to data lag. Next, the calculated displacement change is compared one by one with the preset reference interval (an open interval consisting of the lower and upper limits of the reference interval). This allows for accurate determination of the operating area of ​​each monitoring point on the structure, and then the corresponding early warning threshold adjustment strategy or early warning operation is executed. If the displacement change falls within the reference interval, the structural part corresponding to that monitoring point is determined to be in the plastic development zone. At this time, the current displacement change is input into the pre-established mapping relationship between displacement change and stress early warning threshold adjustment coefficient. The corresponding early warning threshold adjustment coefficient is accurately output through the mapping relationship. Then, this early warning threshold adjustment coefficient is combined with the current stress early warning threshold to obtain the target stress early warning threshold. This target stress early warning threshold is lower than the current threshold. The plastic development zone indicates that the structure has undergone a certain degree of inelastic deformation. Lowering the early warning threshold can improve the efficiency of stress early warning. The system triggers an early warning response in advance, effectively preventing missed risk reports due to excessively high thresholds and allowing time for early intervention during the plastic development stage of the structure. If the displacement change is less than or equal to the lower limit of the baseline, the structure is determined to be in the elastic stability zone. In this case, maintaining the current stress warning threshold is sufficient. This strategy avoids unnecessary threshold adjustments during the stable operation phase of the structure, ensuring the stability of the early warning system while reducing ineffective computational resource usage and improving system operating efficiency. If the displacement change is greater than or equal to the upper limit of the baseline, the structure is determined to be in the warning danger zone. In this case, the system immediately triggers the highest level of response, directly sending warning information to the relevant pre-set operation and maintenance and control personnel. This design enables rapid response to dangerous structural conditions, ensuring that relevant personnel can obtain risk information and initiate emergency response procedures as soon as possible, minimizing the probability of serious accidents such as structural collapse due to excessive deformation.

[0023] S101. Obtain the raw parameters of key monitoring locations. Preprocess the raw data at the edge computing gateway. Preprocessing includes, but is not limited to: filtering out high-frequency noise using wavelet transform or digital filters; identifying and removing outliers caused by construction interference using statistical distribution-based algorithms (such as the 3σ criterion) or isolated forest algorithms; and imputing missing data due to transmission interruptions using time series prediction models or K-nearest neighbor algorithms. Bind the preprocessed data to the current construction stage and perform three levels of adaptive fusion. The value output after adaptive fusion feature fusion is recorded as the monitoring feature value. The three levels of adaptive fusion include data-level fusion, feature-level fusion, and decision-level fusion.

[0024] It should be further explained that the specific steps of data-level fusion are as follows: For multi-sensor data from the same key monitoring location, a sliding window weighted average algorithm is used for spatiotemporal alignment and data optimization, outputting a time series. The specific steps for spatiotemporal alignment and data optimization are as follows: A three-level granularity standard for time coding of the construction stage is preset. The three-level granularity standard includes macro granularity, meso granularity and micro granularity. Each level of granularity corresponds to a unique coding identifier. A mapping relationship between the three-level time coding granularity and the interpolation step size reference value is established. When data loss occurs within the macro granularity, the data loss is considered as a brief interruption under the same stable construction stage. The interpolation step size is set to the preset first short step size to achieve high-precision local waveform reconstruction. When data loss occurs at the meso-granular level, the data loss involves the transition between different states within the same construction phase cycle. The interpolation step size is set to a preset second medium step size to identify and avoid crossing the boundaries of critical events within the sub-phase. When data loss occurs at the micro-granular level, it means that the construction status and load have changed significantly. The interpolation step size is set to the preset third longest step size to maintain the integrity of the data sequence.

[0025] Data-level fusion also includes: Establish a mapping relationship between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient. If the filter cutoff frequency is less than or equal to the critical lower limit of the cutoff frequency, input the current filter cutoff frequency into the mapping relationship between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient, output the corresponding tolerance window adjustment coefficient, and combine the tolerance window adjustment coefficient with the current interpolation error tolerance window to obtain the target tolerance window, so that the allowable instantaneous absolute error can be relatively larger, ensuring that the trend and low-frequency phase characteristics of the interpolated signal remain correct. If the filter cutoff frequency is within the critical cutoff frequency range, the current interpolation error tolerance window is maintained. The critical cutoff frequency range refers to the open interval formed by the lower critical cutoff frequency limit and the upper critical cutoff frequency limit. If the filter cutoff frequency is greater than or equal to the critical upper limit of the cutoff frequency, the current filter cutoff frequency is input into the mapping relationship between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient, and the corresponding tolerance window adjustment coefficient is output. The tolerance window adjustment coefficient is combined with the current interpolation error tolerance window to obtain the target tolerance window, so as to limit the instantaneous absolute error and preserve the transient characteristics and peak information of the signal.

[0026] In this embodiment, for multi-sensor data collected at the same key monitoring location, a sliding window weighted average algorithm is first used to perform spatiotemporal alignment and data optimization processing, and finally output a standardized time series. The specific implementation steps and technical effects of spatiotemporal alignment and data optimization are as follows: First, a three-level granularity standard for construction stage time coding is preset, clarifying the division rules of macro-granularity, meso-granularity and micro-granularity, and configuring a unique coding identifier for each level of granularity. At the same time, a mapping relationship between the three-level time coding granularity and the interpolation step size reference value is established. The establishment of this mapping relationship realizes the differentiated adaptation of the interpolation step size, providing a core basis for the accurate completion of data missing in different construction stages, and avoiding the problem of insufficient completion accuracy or loss of key information caused by a single interpolation step size. Based on this, targeted interpolation strategies are implemented according to the temporal granularity range of the data loss: If the data loss occurs at the macro-granularity, since the macro-granularity corresponds to the stable construction stage, such data loss is judged as a brief interruption under the stable stage. In this case, the interpolation step size is set to the preset first short step size according to the mapping relationship. Short step interpolation can restore the local details of the missing data to the greatest extent, achieve high-precision local waveform reconstruction, and ensure the local consistency between the supplemented data and the original data, providing reliable data support for subsequent accurate analysis of the structural state of the stable construction stage; if the data loss occurs at the meso-granularity, which involves the transition of different states within the same construction stage cycle, such loss can be... To cover key nodes in state transitions, the interpolation step size is set to a preset second medium step size. This medium step size ensures data continuity while effectively identifying and avoiding crossing critical event boundaries within sub-stages, preventing the masking of key state transition information due to excessive smoothing. If data loss occurs at the micro-granular level, corresponding to significant changes in construction status and load, this data loss signifies a break in core monitoring information. In this case, the interpolation step size is set to a preset third long step size. Long-step interpolation aims to maintain the integrity of the data sequence, avoiding the generation of false data due to forced fine interpolation, and ensuring the overall integrity of the time series to support subsequent trend analysis. After the above data processing is completed, further data-level fusion processing is carried out. The core step is to dynamically adjust the interpolation error tolerance window based on the filter cutoff frequency. The specific steps and technical effects are as follows: First, a mapping relationship is established between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient. This mapping relationship enables adaptive adjustment of the error tolerance window, allowing the error control standard to accurately match the signal filtering requirements and improving the adaptability of data fusion.The current filter cutoff frequency is then compared with the preset lower cutoff frequency threshold and the critical cutoff frequency range (the open range formed by the lower and upper cutoff frequency thresholds). If the filter cutoff frequency is less than or equal to the lower cutoff frequency threshold, it indicates that the low-frequency characteristics of the signal need to be preserved. In this case, the current filter cutoff frequency is input into the mapping relationship, and the corresponding tolerance window adjustment coefficient is output. This coefficient is combined with the current interpolation error tolerance window to obtain the target tolerance window. The instantaneous absolute error allowed by the target window is relatively larger. Its core technical effect is to prioritize ensuring the correct trend and low-frequency phase characteristics of the interpolated signal, avoiding the distortion of the low-frequency trend due to strict error limits. If the filter cutoff frequency is within the critical cutoff frequency range... If the current high and low frequency characteristics of the signal are balanced, there is no need to adjust the error tolerance standard; maintaining the current interpolation error tolerance window is sufficient. This strategy can balance the retention requirements of high and low frequency signals, ensuring the stability and universality of data fusion. If the filter cutoff frequency is greater than or equal to the critical upper limit of the cutoff frequency, it indicates that the high-frequency transient characteristics of the signal need to be captured. At this time, the current filter cutoff frequency is input into the mapping relationship to output the corresponding tolerance window adjustment coefficient. The target tolerance window is obtained by combining the coefficient with the current window. The target window strictly limits the instantaneous absolute error and accurately retains the transient characteristics and peak information of the signal. This information often corresponds to critical working conditions such as sudden changes in structural stress and load impact, providing core data support for the early identification of structural dangerous states.

[0027] It should be understood that, such as Figure 2 The diagram shown is a feature-level fusion flowchart of the distributed sensor network deployment and data fusion method for tall steel structures provided in this application embodiment. The specific process is as follows: First, the overall tilt synthesis vector is calculated from the multi-source time series data. Then, the structural state is determined by comparing the magnitude of this vector with the reference magnitude. If the magnitude of the vector is greater than the reference magnitude, it is determined that the overall tilt is significant, and the judgment threshold adjustment coefficient is output through mapping relationship A (relaxed threshold). If the magnitude of the vector is less than or equal to the reference magnitude, it is determined that the overall attitude is normal, and the judgment threshold adjustment coefficient is output through mapping relationship B (tightened threshold). Then, the target threshold is calculated using this coefficient and the original threshold. Finally, the load path anomaly judgment logic is updated to achieve dynamic adaptive adjustment of the load path anomaly judgment threshold.

[0028] It should be further explained that the specific steps of feature-level fusion are as follows: Feature parameters related to construction quality and safety status are extracted from various time-series data, including but not limited to extracting the characteristic frequency and damping ratio of the construction stage from vibration signals through fast Fourier transform; obtaining the real-time stress ratio and stress concentration factor from strain signals; obtaining verticality deviation, inter-story displacement and cumulative settlement from deformation data; and using principal component analysis to reduce the dimensionality and remove redundancy from high-dimensional feature vectors. By obtaining the verticality deviation reference vector of each vertical component control point, the weighted average value of the verticality deviation of all monitoring points in the overall coordinate system of the structure is calculated and denoted as the overall tilt synthesis vector. The overall tilt synthesis vector is used to characterize the geometric tilt state and direction of the overall structure and to establish the mapping relationship between the overall tilt synthesis vector and the load path anomaly judgment threshold. If the overall tilt synthesis vector is greater than the reference vector magnitude, the current overall tilt synthesis vector is input into the mapping relationship between the overall tilt synthesis vector and the judgment threshold adjustment coefficient, and the judgment threshold adjustment coefficient is output. The judgment threshold adjustment coefficient is combined with the current load path anomaly judgment threshold to obtain the target load path anomaly judgment threshold, so as to relax the load path anomaly judgment threshold, maintain high sensitivity to anomalies caused by dangerous deformation modes such as local torsion, and reduce invalid interference while ensuring the accuracy of early warning. If the overall tilt composite vector is less than or equal to the reference vector magnitude, the current overall tilt composite vector is input into the mapping relationship between the overall tilt composite vector and the judgment threshold adjustment coefficient, and the judgment threshold adjustment coefficient is output. The judgment threshold adjustment coefficient is combined with the current load path anomaly judgment threshold to obtain the target load path anomaly judgment threshold, so as to tighten the load path anomaly judgment threshold and improve the sensitivity to local anomalies in the load path.

[0029] In this embodiment, firstly, core feature parameters directly related to construction quality and safety status are accurately extracted from various time-series data that have undergone prior optimization. The specific extraction methods and technical effects are as follows: For vibration time-series signals, Fast Fourier Transform (FFT) technology is used to convert the time-domain vibration signal into a frequency-domain signal, efficiently extracting the characteristic frequency and damping ratio of the construction stage. These two parameters are key indicators reflecting the dynamic mechanical properties of the structure. The offset of the characteristic frequency can intuitively characterize the degree of structural stiffness degradation (such as stiffness reduction caused by loose component connections or material damage), while abnormal fluctuations in the damping ratio can reflect changes in the structural energy dissipation mechanism, providing accurate data support for early identification of structural dynamic safety hazards during construction. For strain time-series signals, based on the previously filtered and temperature-compensated data... The effect variation value is calculated to obtain the real-time stress ratio and stress concentration factor. The real-time stress ratio can clearly determine the difference between the current stress state of the structure and the design limit stress, and directly assess the structural stress safety. The stress concentration factor can quickly locate the local stress concentration area of ​​the structure (such as the corner of the component, the joint of the node, etc., which are prone to stress concentration), and avoid structural failure caused by local stress overload. For deformation time series data, verticality deviation, inter-story displacement and cumulative settlement are extracted through time series trend analysis. These parameters are the core indicators of structural geometric stability. Excessive verticality deviation may cause eccentric stress in the structure, abnormal inter-story displacement is prone to inter-story structural failure, and excessive cumulative settlement directly threatens the overall stability of the structure, providing a comprehensive basis for construction quality control (such as component installation accuracy) and structural safety assessment. Meanwhile, considering that the feature vectors formed by the extracted feature parameters have high dimensionality and there may be information redundancy between different parameters, principal component analysis is used to reduce the dimensionality and remove redundancy of the high-dimensional feature vectors. This process can greatly simplify the data dimension while preserving the core feature information to the greatest extent, reduce the complexity and cost of subsequent analysis and calculation, and avoid misjudgment of security status caused by redundant information interference, thereby improving analysis efficiency and judgment accuracy.After feature extraction and dimensionality reduction, the overall tilt state of the structure is analyzed based on the verticality deviation, thereby realizing the dynamic adjustment of the load path anomaly judgment threshold. The specific steps and technical effects are as follows: First, a verticality deviation benchmark vector is constructed using the verticality deviation data collected from the control points of each vertical component. This benchmark vector serves as a reference standard for the normal geometric state of the structure, providing an objective basis for the subsequent judgment of the overall tilt state and avoiding misjudgment due to the lack of a unified benchmark. Then, the weighted average of the verticality deviation of all monitoring points in the overall coordinate system of the structure is calculated and recorded as the overall tilt synthesis vector. This vector, by weighted fusion of the local deviation information of each monitoring point, can comprehensively and accurately characterize the overall geometric tilt state and specific tilt direction of the structure, breaking through the limitation that the local deviation of a single monitoring point cannot reflect the overall geometric state of the structure. This provides a reliable core indicator for the overall structural safety assessment. At the same time, a mapping relationship between the overall tilt synthesis vector and the load path anomaly judgment threshold is established, providing scientific model support for the dynamic adaptation of the threshold. Based on the relationship between the magnitude of the overall tilt composite vector and the reference vector, a differentiated load path anomaly judgment threshold adjustment strategy is implemented: if the overall tilt composite vector is greater than the reference vector magnitude, it indicates that the structure currently has a certain degree of overall tilt. In this case, the current overall tilt composite vector is input into a preset mapping relationship, and the corresponding judgment threshold adjustment coefficient is output. This coefficient is then combined with the current load path anomaly judgment threshold to obtain the target load path anomaly judgment threshold. This adjustment focuses on relaxing the threshold, and its technical effect is to avoid frequent invalid warnings caused by overly strict thresholds when the structure is under overall tilt, while maintaining stability. The high sensitivity to anomalies caused by dangerous deformation modes such as partial torsion ensures that the early warning system does not miss key safety hazards while reducing interference. If the overall tilt synthesis vector is less than or equal to the magnitude of the reference vector, it indicates that the overall geometric state of the structure is stable. At this time, the current overall tilt synthesis vector is also input into the mapping relationship to output the judgment threshold adjustment coefficient. The target load path anomaly judgment threshold is obtained by combining the coefficient with the current threshold. This adjustment focuses on tightening the threshold, which significantly improves the sensitivity to local anomalies in the load path and can promptly capture subtle safety hazards such as local force path deviations in the structure, providing a guarantee for refined safety management during construction.

[0030] It should be further explained that the specific steps of decision-level integration are as follows: The reduced feature vectors and the current construction stage label are input into a pre-trained deep learning model to output the safety assessment level of the current construction status. For different construction stages, the temporal evolution characteristics of the reduced low-dimensional feature vectors are analyzed, and a mapping relationship between the state evolution rate and the optimal serialization window length is established. If the state evolution rate is greater than the baseline value of the evolution rate, it indicates that the structure is in a rapid change and transient response stage. Then, the current state evolution rate is input into the mapping relationship between the state evolution rate and the optimal serialization window length, and the adjusted optimal serialization window length is output to enhance the real-time capture capability and early warning timeliness of transient risks and sudden events. If the state evolution rate is less than or equal to the baseline value, it indicates that the structure is in a relatively stable state, and the current optimal serialization window length is maintained.

[0031] In this embodiment, firstly, the low-dimensional feature vectors, after dimensionality reduction and redundancy removal through principal component analysis, are input along with the current construction stage label into a pre-trained deep learning model. Through the model's precise learning and inference of the matching relationship between the feature vectors and the construction stages, the safety assessment level of the current construction status is output. The low-dimensional feature vectors ensure the efficiency and integrity of the model's input data, avoiding the drag on model computational efficiency caused by high-dimensional data. Integrating the construction stage label allows the model to conduct assessments based on the stress characteristics and risk focus differences of different construction stages, significantly improving the relevance and accuracy of the safety assessment level and providing a direct and clear decision-making basis for construction safety management. Subsequently, considering the differentiated characteristics of different construction stages, the temporal evolution characteristics of the dimensionality-reduced low-dimensional feature vectors are analyzed in depth. By quantifying the magnitude and trend of feature vector changes over time, the evolution rate of the structural construction state is obtained. Simultaneously, a mapping relationship between the state evolution rate and the optimal serialization window length is established, providing a scientific basis for the dynamic adaptation of the window length. Based on the comparison between this mapping relationship and the baseline value of the state evolution rate, a differentiated optimal serialization window length adjustment strategy is implemented: If the state evolution rate is greater than the baseline value, it indicates that the structure is in a rapid change and transient response phase (such as the peak period of concrete pouring, the stage of hoisting large components, etc.). In this case, the current state evolution rate is input into the preset mapping relationship, and the adjusted optimal serialization window length is output. Shortening the serialization window length can reduce the time required for data accumulation, accelerate the model's response speed to the latest data, and thus significantly enhance the real-time capture capability and early warning timeliness of transient risks (such as stress mutations caused by hoisting impact loads) and sudden events, buying time for rapid handling of dangerous conditions; if the state evolution rate is less than or equal to the baseline value, it indicates that the structure is in a relatively stable construction state (such as the construction interval period, the component curing stage, etc.). In this case, the current optimal serialization window length can be maintained. This strategy can ensure the stability and reliability of model evaluation, avoid unnecessary window adjustments that occupy computational resources, and ensure the efficient operation of decision-level fusion.

[0032] S102. Construct a digital twin model of the towering steel structure, adaptively integrate features and evaluation results, drive the digital twin model to update its status and visualize its display in real time, determine whether the monitoring feature value is within the warning critical range output by the digital twin model, and if so, provide a corresponding emergency plan according to the warning level.

[0033] It should be understood that the monitored characteristic values ​​are compared with the early warning thresholds calculated based on the twin model in real time. The thresholds are dynamically adjusted according to the construction stage, design specifications, and real-time environmental loads. When the characteristic value exceeds the threshold, the system triggers multi-level early warnings. The early warnings are divided into at least two levels: Level 1 is the attention level, which prompts attention through the data platform; Level 2 is the alarm level, which automatically pushes alarm information to relevant construction management personnel and links the 3D model to highlight the risk areas. Based on the warning level and type, preliminary construction decision suggestions are provided, including suspending local operations, conducting verification measurements, adjusting the construction sequence, or activating emergency plans.

[0034] It should be noted that the specific steps for determining whether the monitored feature value is within the warning critical range output by the digital twin model are as follows: If the monitored feature value is within the warning critical interval output by the digital twin model, it is determined to be a level one warning, and attention is alerted through the data platform. The warning critical interval represents the closed interval formed by the lower and upper warning critical limits. If the monitored feature value is less than the lower limit of the warning threshold output by the digital twin model, the adaptive fusion feature and evaluation result will be stored in the time series database to form a traceable construction digital archive. If the monitored characteristic value is greater than the upper limit of the warning threshold output by the digital twin model, it is determined to be a level two warning, and the alarm information is automatically pushed to the relevant construction management personnel, and the risk area is highlighted in the 3D model.

[0035] In this embodiment, the definition of the warning critical interval is first clarified. This interval is a closed interval formed by the lower and upper warning critical limits generated by the digital twin model based on structural design parameters, construction condition simulation, and historical safety data. Its accuracy provides a scientific and reliable benchmark for subsequent graded warning determination. Then, the monitoring feature values, after preliminary feature extraction and fusion processing, are precisely compared with the aforementioned warning critical interval: if the monitoring feature value falls within the warning critical interval, it is determined to be a Level 1 warning. At this time, a attention reminder is issued to relevant personnel through the data platform. The technical effect of this graded response is to provide accurate warnings when the structure is in a "critical safety state," avoiding the waste of resources caused by excessive warnings due to slight fluctuations, and timely reminding personnel to pay attention to changes in the structural state, achieving early intervention and prediction of risks. If the monitoring feature value is less than the lower warning critical limit, it indicates that the structure is currently in a safe and stable state. At this time, the adaptive fusion features obtained from the previous processing and the safety assessment results are synchronously stored in the time-series database to construct a complete and traceable construction digital archive, which not only provides a basis for subsequent construction stages... Safety assessments provide historical data support and verifiable original evidence for project completion acceptance and subsequent operation and maintenance, ensuring traceability and accountability throughout the construction process. If the monitored characteristic value exceeds the warning threshold, it is determined to be a Level II warning. At this time, the system immediately activates a high-level response mechanism. On the one hand, it automatically and accurately pushes alarm information to relevant construction management personnel (including on-site operation and maintenance personnel, project control personnel, etc.) to ensure that key risk information quickly reaches the responsible persons. On the other hand, it links with the digital twin 3D model to highlight the corresponding risk areas, achieving "precise positioning + rapid response" of risks. This helps management personnel to locate the risk location as soon as possible, shorten the emergency response time, minimize the probability of safety accidents, and protect the lives of construction personnel and the safety of project property.

[0036] The various features and processes described above can be used independently of each other or can be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of this disclosure. Furthermore, certain method or process blocks may be omitted in some embodiments. The methods and processes described herein are not limited to any particular order, and the blocks or states associated with them may be performed in other suitable orders. For example, the described blocks or states may be performed in an order different from the order specifically disclosed, or multiple blocks or states may be combined in a single block or state. Example blocks or states may be performed serially, in parallel, or in some other manner. Blocks or states may be added to or removed from the disclosed example embodiments. The exemplary systems and components described herein may be configured differently from those described. For example, elements may be added to, removed from, or rearranged compared to the disclosed example embodiments.

[0037] The various operations of the example methods described herein can be performed at least in part by an algorithm. This algorithm can be contained in program code or instructions stored in memory (e.g., the aforementioned non-transitory computer-readable storage medium). Such an algorithm may include a machine learning algorithm. In some embodiments, the machine learning algorithm may not be explicitly programmed into the computer to perform the function, but can learn from training data to create a predictive model that performs the function.

[0038] The various operations of the example methods described herein can be performed, at least in part, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors can constitute the engine of a processor implementation that operates to perform one or more of the operations or functions described herein.

[0039] Similarly, the methods described herein can be implemented at least in part by a processor, where one or more specific processors are examples of hardware. For example, at least some operations of a method can be performed by one or more processors or an engine implemented by a processor. Furthermore, one or more processors can also be operated to support the performance of related operations in a “cloud computing” environment or as “Software as a Service” (SaaS). For example, at least some operations can be performed by a set of computers (as an example of a machine including processors), where these operations are accessible via a network (e.g., the Internet) and via one or more suitable interfaces (e.g., application programming interfaces (APIs)).

[0040] The performance of certain operations can be distributed across processors, residing not only within a single machine but also deployed across multiple machines. In some example embodiments, the processor or processor-implemented engine may reside in a single geographic location (e.g., within a home environment, office environment, or server cluster). In other example embodiments, the processor or processor-implemented engine may be distributed across multiple geographic locations.

[0041] In this specification, multiple instances may implement components, operations, or structures described as single instances. Although individual operations of one or more methods are shown and described as separate operations, one or more of the separate operations may be performed simultaneously and do not need to be performed in the order shown. Structures and functions presented as separate components in the example configuration may be implemented as composite structures or components. Similarly, structures and functions presented as single components may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of this document.

[0042] While an overview of the subject matter has been described with reference to specific example embodiments, various modifications and changes can be made to these embodiments without departing from the broader scope of embodiments of this disclosure. Such embodiments of the subject matter are referred to herein, individually or collectively, by the term "invention," and are used for convenience only and are not intended to limit the scope of this application to any single disclosure or concept, should more than one disclosure or concept be disclosed in fact.

[0043] The embodiments described herein have been described in sufficient detail to enable those skilled in the art to practice the disclosed teachings. Other embodiments may be used and derived therefrom, such that structural and logical substitutions and changes may be made without departing from the scope of this disclosure. Therefore, the detailed description should not be construed as limiting, and the scope of the various embodiments is defined only by the appended claims and the full scope of their equivalents.

Claims

1. A method for deploying and fusing distributed sensor networks for tall steel structures, characterized in that, Includes the following steps: S100, based on the finite element model of the tall steel structure to be monitored and the simulation of the entire construction process, identifies the key monitoring parts at each construction stage. It distributes the key monitoring parts through a heterogeneous sensor network to adapt to dynamic adjustments during construction, allowing for location migration and supplementation as construction stages progress. S101, Obtain the original parameters of key monitoring parts, preprocess the original data at the edge computing gateway, bind the preprocessed data with the current construction stage, and perform three-level adaptive fusion. The value output after adaptive fusion feature fusion is recorded as the monitoring feature value. The three-level adaptive fusion includes data-level fusion, feature-level fusion and decision-level fusion. S102, construct a digital twin model of the tall steel structure, adaptively integrate features and evaluation results, drive the digital twin model to update its status and visualize it in real time, determine whether the monitoring feature value is within the warning critical range output by the digital twin model, and if so, provide a corresponding emergency plan according to the warning level.

2. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 1, characterized in that, The key monitoring areas include, but are not limited to: the installation nodes of the main load-bearing components that change dynamically with the construction progress, the connection points of the temporary support system, and the areas of high-altitude welding and bolt connection. The heterogeneous sensor network includes, but is not limited to, fiber optic grating sensors and resistance strain gauges for monitoring component stress, total stations or electronic theodolites for monitoring three-dimensional deformation and perpendicularity, piezoelectric accelerometers for monitoring structural vibration, and environmental sensors for collecting ambient wind speed, temperature, and humidity.

3. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 1, characterized in that, The specific steps to allow for location relocation and addition during construction phases are as follows: By installing several strain gauges on key monitoring parts of the structure, the original strain signals of each monitoring point at a predetermined time are synchronously collected at a predetermined sampling frequency. The original strain signals are filtered and temperature compensated to obtain the effective strain value, and the real-time strain increment within a predetermined time window is obtained. Within the preset time window, all monitoring points are traversed to obtain the real-time strain increment with the largest absolute value. The strain increment normalization weighting factor of each monitoring point is calculated, and the real-time force ratio of the monitoring point at a predefined time is obtained. The strain increment normalization weighting factor is dynamically adjusted to the force ratio warning threshold of each monitoring point to generate a dynamic force ratio warning threshold.

4. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 3, characterized in that, The specific steps for generating the dynamic force ratio early warning threshold are as follows: If the strain increment normalization weight factor is greater than the strain increment weight benchmark value, it indicates that the monitoring point is experiencing relatively drastic stress changes. In this case, the stress ratio warning threshold of the monitoring point is relaxed to obtain the dynamic stress ratio warning threshold, so as to avoid triggering false alarms due to normal load redistribution. If the strain increment normalization weight factor is less than or equal to the strain increment weight benchmark value, the current dynamic stress ratio warning threshold will be maintained. The real-time stress ratio of each monitoring point is compared with the dynamic stress ratio warning threshold. If the real-time stress ratio is greater than the dynamic stress ratio warning threshold, an early warning signal is immediately triggered, and the monitoring point number and strain increment normalization weight factor are identified. If the real-time force ratio is less than or equal to the dynamic force ratio warning threshold, no signal needs to be triggered.

5. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 3, characterized in that, The provision allowing for relocation and addition during construction phases also includes: A mapping relationship between displacement change and stress early warning threshold adjustment coefficient is established. This is achieved by acquiring the initial values ​​of support displacement at each monitoring point and the corresponding initial stress values. Support displacement values ​​are collected in real-time over a predefined monitoring period to obtain the displacement change relative to the initial values. The displacement change is then compared with a reference range. Specifically: If the displacement change is within the reference range, it is determined to be in the plastic development zone. The current displacement change is input into the mapping relationship between the displacement change and the stress warning threshold adjustment coefficient, and the warning threshold adjustment coefficient is output. The warning threshold adjustment coefficient is combined with the current stress warning threshold to obtain the target stress warning threshold, so as to lower the stress warning threshold and prevent missed alarms. The reference range of the change refers to the open interval formed by the lower limit of the reference range and the upper limit of the reference range. If the displacement change is less than or equal to the lower limit of the change reference, it is determined to be in the elastic stability zone, and the current stress warning threshold is maintained. If the displacement change is greater than or equal to the upper limit of the change benchmark, it is determined to be a warning danger zone, and an early warning message is sent to the preset relevant personnel.

6. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 1, characterized in that, The specific steps of the data-level fusion are as follows: For multi-sensor data of the same key monitoring location, a sliding window weighted average algorithm is used for spatiotemporal alignment and data optimization, outputting a time series. The specific steps for spatiotemporal alignment and data optimization are as follows: A three-level granularity standard for time coding of the construction stage is preset. The three-level granularity standard includes macro granularity, meso granularity and micro granularity. Each level of granularity corresponds to a unique coding identifier. A mapping relationship between the three-level time coding granularity and the interpolation step size reference value is established. When data loss occurs within the macro granularity, the data loss is considered as a brief interruption under the same stable construction stage. The interpolation step size is set to the preset first short step size to achieve high-precision local waveform reconstruction. When data loss occurs at the meso-granular level, the data loss involves the transition between different states within the same construction phase cycle. The interpolation step size is set to a preset second medium step size to identify and avoid crossing the boundaries of key events within the sub-phase. When data loss occurs at the micro-granular level, it means that the construction status and load have changed significantly. The interpolation step size is set to the preset third longest step size to maintain the integrity of the data sequence.

7. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 6, characterized in that, The data-level fusion also includes: Establish a mapping relationship between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient. If the filter cutoff frequency is less than or equal to the critical lower limit of the cutoff frequency, input the current filter cutoff frequency into the mapping relationship between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient, output the corresponding tolerance window adjustment coefficient, and combine the tolerance window adjustment coefficient with the current interpolation error tolerance window to obtain the target tolerance window, so that the allowable instantaneous absolute error can be relatively larger, ensuring that the trend and low-frequency phase characteristics of the interpolated signal remain correct. If the filter cutoff frequency is within the critical cutoff frequency range, the current interpolation error tolerance window is maintained. The critical cutoff frequency range refers to the open interval formed by the lower critical cutoff frequency limit and the upper critical cutoff frequency limit. If the filter cutoff frequency is greater than or equal to the critical upper limit of the cutoff frequency, the current filter cutoff frequency is input into the mapping relationship between the filter cutoff frequency and the interpolation error tolerance window adjustment coefficient, and the corresponding tolerance window adjustment coefficient is output. The tolerance window adjustment coefficient is combined with the current interpolation error tolerance window to obtain the target tolerance window, so as to limit the instantaneous absolute error and preserve the transient characteristics and peak information of the signal.

8. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 1, characterized in that, The specific steps of the feature-level fusion are as follows: Feature parameters related to construction quality and safety status are extracted from various time-series data, including but not limited to extracting the characteristic frequency and damping ratio of the construction stage from vibration signals through fast Fourier transform; obtaining the real-time stress ratio and stress concentration factor from strain signals; obtaining verticality deviation, inter-story displacement and cumulative settlement from deformation data; and using principal component analysis to reduce the dimensionality and remove redundancy from high-dimensional feature vectors. By obtaining the verticality deviation reference vector of each vertical component control point, the weighted average value of the verticality deviation of all monitoring points in the overall coordinate system of the structure is calculated and denoted as the overall tilt synthesis vector. The overall tilt synthesis vector is used to characterize the geometric tilt state and direction of the overall structure and to establish the mapping relationship between the overall tilt synthesis vector and the load path anomaly judgment threshold. If the overall tilt synthesis vector is greater than the reference vector magnitude, the current overall tilt synthesis vector is input into the mapping relationship between the overall tilt synthesis vector and the judgment threshold adjustment coefficient, and the judgment threshold adjustment coefficient is output. The judgment threshold adjustment coefficient is combined with the current load path anomaly judgment threshold to obtain the target load path anomaly judgment threshold, so as to relax the load path anomaly judgment threshold, maintain high sensitivity to anomalies caused by dangerous deformation modes such as local torsion, and reduce invalid interference while ensuring the accuracy of early warning. If the overall tilt composite vector is less than or equal to the reference vector magnitude, the current overall tilt composite vector is input into the mapping relationship between the overall tilt composite vector and the judgment threshold adjustment coefficient, and the judgment threshold adjustment coefficient is output. The judgment threshold adjustment coefficient is combined with the current load path anomaly judgment threshold to obtain the target load path anomaly judgment threshold, so as to tighten the load path anomaly judgment threshold and improve the sensitivity to local anomalies in the load path.

9. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 1, characterized in that, The specific steps of the decision-level fusion are as follows: The reduced feature vectors and the current construction stage label are input into a pre-trained deep learning model to output the safety assessment level of the current construction status. For different construction stages, the temporal evolution characteristics of the reduced low-dimensional feature vectors are analyzed, and a mapping relationship between the state evolution rate and the optimal serialization window length is established. If the state evolution rate is greater than the baseline value of the evolution rate, it indicates that the structure is in a rapid change and transient response stage. Then, the current state evolution rate is input into the mapping relationship between the state evolution rate and the optimal serialization window length, and the adjusted optimal serialization window length is output to enhance the real-time capture capability and early warning timeliness of transient risks and sudden events. If the state evolution rate is less than or equal to the baseline value, it indicates that the structure is in a relatively stable state, and the current optimal serialization window length is maintained.

10. The method for deploying and fusing distributed sensor networks for tall steel structures as described in claim 1, characterized in that, The specific steps for determining whether the monitored feature value is within the critical range of the early warning output by the digital twin model are as follows: If the monitored feature value is within the warning critical interval output by the digital twin model, it is determined to be a level one warning, and attention is alerted through the data platform. The warning critical interval represents the closed interval formed by the lower warning critical limit and the upper warning critical limit. If the monitored feature value is less than the lower limit of the warning threshold output by the digital twin model, the adaptive fusion feature and evaluation result will be stored in the time series database to form a traceable construction digital archive. If the monitored characteristic value is greater than the upper limit of the warning threshold output by the digital twin model, it is determined to be a level two warning, and the alarm information is automatically pushed to the relevant construction management personnel, and the risk area is highlighted in the 3D model.

Images